Keywords: Brain-computer/machine interface, Brain functional imaging - EEG, Brain functional imaging - NIR
Abstract: Brain-computer Interfaces (BCIs) provide a direct pathway between the brain and the outward environment. Specifically, motor imagery (MI)-based BCI controlling functional electric stimulation (FES) is a promising approach for disabled patients with intact mind to restore or rehabilitate their motor functions. This study probed for the improvement of cortical activation for motor imagery during the closed-loop BCI-FES training. We used electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) to inspect the cortical activation for four different training strategies, i.e. MI-BCI-FES, MI-FES, MI and FES. Compared with the other three training conditions, the MI-BCI-FES could achieve significantly stronger cortical activation viewing from the event-related desynchronization (ERD) and the blood oxygen response. The results demonstrate that the closed-loop MI training using BCI-FES can effectively increase the cortical activation of motor areas.

Keywords: Brain-computer/machine interface, Sensory neuroprostheses - Visual, Brain functional imaging - EEG
Abstract: Steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs) have potential to realize high-speed communication between human brain and external devices. Recently, we proposed an intermodulation frequency-based stimulation approach to increase the number of visual stimuli that can be presented on a computer monitor. Although our recent studies have demonstrated that this approach can encode more visual stimuli by only one flickering frequency, the performance of the intermodulation frequency-based SSVEP BCI remains poor and needs further improvement. This study aims to incorporate filter bank analysis and individual SSVEP calibration data into canonical correlation analysis (CCA) to improve the detection of SSVEPs with intermodulation frequencies. Results on classification accuracy and information transfer rate (ITR) suggest that the employment of individual calibration data can significantly improve the performance of the intermodulation frequency-based SSVEP BCI.

Keywords: Brain-computer/machine interface, Brain functional imaging - EEG, Brain functional imaging - Classification
Abstract: Due to the lack of electroencephalography (EEG) data, it is hard to build an emotion recognition model with high accuracy from EEG signals using machine learning approach. Inspired by generative adversarial networks (GANs), we introduce a Conditional Wasserstein GAN (CWGAN) framework for EEG data augmentation to enhance EEG-based emotion recognition. A Wasserstein GAN with gradient penalty is adopted to generate realistic-like EEG data in differential entropy (DE) form. Three indicators are used to judge the qualities of the generated high-dimensional EEG data, and only high quality data are appended to supplement the data manifold, which leads to better classification of different emotions. We evaluate the proposed CWGAN framework on two public EEG datasets for emotion recognition, namely SEED and DEAP. The experimental results demonstrate that using the EEG data generated by CWGAN significantly improves the accuracies of emotion recognition models.

Keywords: Brain-computer/machine interface, Neural signals - Nonlinear analysis, Neural signal processing
Abstract: Changes in the functional mapping between neural activities and kinematic parameters over time poses a challenge to current neural decoder of brain machine interfaces (BMIs). Traditional decoders robust to changes in functional mappings required many day’s training data. The decoder may not be robust when it was trained by data from only few days. Therefore, a decoder should be trained to handle a variety of neural-to-kinematic mappings using limited training data. We proposed an evolutionary neural network with error feedback, ECPNN-EF, as a neural decoder, that considered the previous error as an input to the decoder in order to improve the robustness. The decoder was validated to reconstruct rat’s forelimb movement in a water-reward lever-pressing task. Two days of data were only used to train the decoder while ten days of data were used to test the decoder. The results showed that the performance of ECPNN-EF was significantly higher than that of standard recurrent neural network without error feedback, which was commonly used in BMI. This suggested that ECPNN-EF trained with few days of training data can be robust to changes in functional mappings.

Keywords: Brain-computer/machine interface, Brain functional imaging - Classification, Brain functional imaging - Evoked potentials
Abstract: P300-based brain-computer interfaces (BCIs) are often trained per-user and per-application space. Training such models requires ground truth knowledge of target and non-target stimulus categories during model training, which imparts bias into the model. Additionally, not all non-targets are created equal; some may contain visual features that resemble targets or may otherwise be visually salient. Current research has indicated that non-target distractors may elicit attenuated P300 responses based on the perceptual similarity of these distractors to the target category. To minimize this bias, and enable a more nuanced analysis, we use a generalized BCI approach that is fit to neither user nor task. We do not seek to improve the overall accuracy of the BCI with our generalized approach; we instead demonstrate the utility of our approach for identifying target-related image features. When combined with other intelligent agents, such as computer vision systems, the performance of the generalized model equals that of the user-specific models, without any user specific data.

Keywords: Brain-computer/machine interface, Motor neuroprostheses, Neural signal processing
Abstract: Brain Machine Interfaces (BMIs) mostly utilise spike rate as an input feature for decoding a desired motor output as it conveys a useful measure to the underlying neuronal activity. The spike rate is typically estimated by a using non-overlap binning method that yields a coarse estimate. There exist several methods that can produce a smooth estimate which could potentially improve the decoding performance. However, these methods are relatively computationally heavy for real-time BMIs. To address this issue, we propose a new method for estimating spike rate that is able to yield a smooth estimate and also amenable to real-time BMIs. The proposed method, referred to as Bayesian adaptive kernel smoother (BAKS), employs kernel smoothing technique that considers the bandwidth as a random variable with prior distribution which is adaptively updated through a Bayesian framework. With appropriate selection of prior distribution and kernel function, an analytical expression can be achieved for the kernel bandwidth. We apply BAKS and evaluate its impact on offline BMI decoding performance using Kalman filter. The results reveal that BAKS can improve the decoding performance compared to the binning method. This suggests the feasibility and the potential use of BAKS for real-time BMIs.

Keywords: Neural networks and support vector machines in biosignal processing and classification
Abstract: Premature ventricular contraction (PVC) is usually considered as a benign arrhythmia in the absence of structural heart diseases. However, frequent premature beats may significantly increase the risk of heart failure and even death by arrhythmia-induced cardiomyopathy. Therefore, high PVC counts have been considered as an index to predict the risk of severe arrhythmias. Wearable devices can measure ECG signals and therefore are convenient tools for the detection of premature contraction in casual life. Considering the huge quantities of data recorded by wearable devices, reliable and low-cost data analysis programs should be developed for real time PVC detection. In this research, we use recurrent neural networks with, long short-term memory to detect PVC. Through validating with MIT-BIH arrhythmia database, the detection accuracy of this method is 96%-99%.

Keywords: Neural networks and support vector machines in biosignal processing and classification, Signal pattern classification, Data mining and processing - Pattern recognition
Abstract: We propose a novel electrocardiogram(ECG) beat classification algorithm using a combination of Bidirectional Recurrent Neural Network(BiRNN) and Convolutional Neural Network(CNN) named as BiRCNN. Our model is an end-to-end model. The morphological features of each ECG beat is extracted by CNN. Then the features of each beat are considered in the context via BiRNN. The experimental results on MIT-BIH Arrhythmia Database(MITDB) demonstrate that our proposed system achieves superior beat classification performance.

Keywords: Neural networks and support vector machines in biosignal processing and classification, Data mining and processing - Pattern recognition
Abstract: Detection of ECG characteristic points serves as the first step in automated ECG analysis techniques. We propose a novel end-to-end deep learning scheme called Region Aggregation Network (RAN) for ECG characteristic points detection. A 1D Convolutional Neural Network (CNN) is adopted to automatically process ECG signals. A novel strategy of Region Aggregation is proposed to replace the conventional fully connected layer as regressor. Our work provides robust and accurate detection performance on public ECG database. The evaluation results of our method on QT database show comparable detection accuracy compared with state-of-the-art works.

Keywords: Neural networks and support vector machines in biosignal processing and classification, Signal pattern classification, Physiological systems modeling - Signal processing in physiological systems
Abstract: Ballistocardiography (BCG) is a revamped technology for cardiac function monitoring. Detecting individual heart beats in BCG remains a challenging task due to various artifacts and low signal-to-noise ratio, which are not well addressed by conventional approaches based on intuitive observations of BCG waveforms. In this paper, we propose to employ deep learning networks to capture the characteristics of the variations of BCG waveforms within and across individual subjects. Particularly, we design a neural network that combines Convolutional-Neural-Network (CNN) and Extreme Learning Machine (ELM). We test the new learning method on a real BCG data set and show better detection result compared with a state-of-the-art method. We demonstrate how the advanced machine learning technology can learn and detect BCG waveforms robustly.

Keywords: Signal pattern classification, Neural networks and support vector machines in biosignal processing and classification
Abstract: Cardiac arrhythmia is known to be one of the most common causes of death worldwide. Therefore, development of efficient arrhythmia detection techniques is essential to save patients' lives. In this paper, we introduce a new real-time cardiac arrhythmia classification using memristor neuromorphic computing system for classification of 5 different beat types. Neuromorphic computing systems utilize new emerging devices, such as memristors, as a basic building block. Hence, these systems provide excellent trade-off between real-time processing, power consumption, and overall accuracy. Experimental results showed that the proposed system outperforms most of the methods in comparison in terms of accuracy and testing time, since it achieved 96.17% average accuracy and 34 ms average testing time per beat.

Keywords: Neural networks and support vector machines in biosignal processing and classification, Signal pattern classification, Data mining and processing in biosignals
Abstract: Processing temporal sequences is central to a variety of applications in health care, and in particular multi-channel Electrocardiogram (ECG) is a highly prevalent diagnostic modality that relies on robust sequence modeling. While Recurrent Neural Networks (RNNs) have led to significant advances in automated diagnosis with time-series data, they perform poorly when models are trained using a limited set of channels. A crucial limitation of existing solutions is that they rely solely on discriminative models, which tend to generalize poorly in such scenarios. In order to combat this limitation, we develop a generative modeling approach to limited channel ECG classification. This approach first uses a textit{Seq2Seq} model to implicitly generate the missing channel information, and then uses the latent representation to perform the actual supervisory task. This decoupling enables the use of unsupervised data and also provides highly robust metric spaces for subsequent discriminative learning. Our experiments with the Physionet dataset clearly evidence the effectiveness of our approach over standard RNNs in disease prediction.

Keywords: Image analysis and classification - Machine learning / Deep learning approaches, Image classification
Abstract: Skin cancer is among the deadliest variants of cancer if not recognized and treated in time. This work focuses on the identification of this disease using an ensemble of state-of-the-art deep learning approaches. More specifically, we propose the aggregation of robust convolutional neural networks (CNNs) into one neural net architecture, where the final classification is achieved based on the weighted output of the member CNNs. Since our framework is realized within a single neural net architecture, all the parameters of the member CNNs and the weights applied in the fusion can be determined by backpropagation routinely applied for such tasks. The presented ensemble consists of the CNNs AlexNet, VGGNet, GoogLeNet, all of which have been won in subsequent years the most prominent worldwide image classification challenge ImageNet. For an objective evaluation of our approach, we have tested its performance on the official test database of the IEEE International Symposium on Biomedical Imaging (ISBI) 2017 challenge on Skin Lesion Analysis Towards Melanoma Detection dedicated to skin cancer recognition. Our experimental studies show that the proposed approach is competitive in this field. Moreover, the ensemble-based approach outperformed all of its member CNNs.

Keywords: Image analysis and classification - Machine learning / Deep learning approaches, Image classification, CT imaging
Abstract: Lung cancer overall survival analysis using computed tomography (CT) images plays an important role in treatment planning. Most current analysis methods involve hand-crafted image features for survival time prediction. However, hand-crafted features require domain knowledge and may lack specificity to lung cancer. Advanced self-learning models such as deep learning have showed superior performance in many medical image tasks, but they require large amount of data which is difficult to collect for survival analysis because of the long follow-up time. Although data with survival time is difficult to acquire, it is relatively easy to collect lung cancer patients without survival time. In this paper, we proposed an unsupervised deep learning method to take advantage of the unlabeled data for survival analysis, and demonstrated better performance than using hand-crafted features. We proposed a residual convolutional auto encoder and trained the model using images from 274 patients without survival time. Afterwards, we extracted deep learning features through the encoder model, and constructed a Cox proportional hazards model on 129 patients with survival time. The experiment results showed that our unsupervised deep learning feature gained better performance (C-Index = 0.70) than using hand-crafted features (C-Index = 0.62). Furthermore, we divided the patients into two groups according to their Cox hazard value. Kaplan-Meier analysis indicated that our model can divide patients into high and low risk groups and the survival time of these two groups had significant difference (p < 0.01).

Keywords: Image analysis and classification - Machine learning / Deep learning approaches, X-ray imaging applications
Abstract: Breast cancer is the most prevalent cancer among women. The most common method to detect breast cancer is mammography. However, interpreting mammography is a challenging task that requires high skills and is time-consuming. In this work, we propose a computer-aided diagnosis (CAD) scheme for mammography based on transfer representation learning using the Inception-V3 architecture. We evaluate the performance of the proposed scheme using the INBreast database, where the features are extracted from different layers of the architecture. In order to cope with the small dataset size limitation, we expand the training dataset by generating artificial mammograms and employing different augmentation techniques. The proposed scheme shows great potential with a maximal area under the receiver operating characteristics curve of 0.91.

Keywords: Image segmentation, Image analysis and classification - Machine learning / Deep learning approaches, Optical imaging
Abstract: Dermoscopic imaging is an established technique to detect, track, and diagnose malignant melanoma, and one of the ways to improve this technique is via computer-aided image segmentation. Image segmentation is an important step towards building computerized detection and classification systems by delineating the area of interest, in our case, the skin lesion, from the background. However, current segmentation techniques are hard pressed to account for color artifacts within dermoscopic images that are often incorrectly detected as part of the lesion. Often there are few annotated examples of these artifacts, which limits training segmentation methods like the fully convolutional network (FCN) due to the skewed dataset. We propose to improve FCN training by augmenting the dataset with synthetic images created in a controlled manner using a generative adversarial network (GAN). Our novelty lies in the use of a color label (CL) to specify the different characteristics (approximate size, location, and shape) of the different regions (skin, lesion, artifacts) in the synthetic images. Our GAN is trained to perform style transfer of real melanoma image characteristics (e.g. texture) onto these color labels, allowing us to generate specific types of images containing artifacts. Our experimental results demonstrate that the synthetic images generated by our technique have a lower mean average error when compared to synthetic images generated using traditional binary labels. As a consequence, we demonstrated improvements in melanoma image segmentation when using synthetic images generated by our technique.

Keywords: Multiscale image analysis, Image analysis and classification - Machine learning / Deep learning approaches, Image enhancement
Abstract: Monitoring of heart rate can be used in many medical and sports applications. Lack of portability and connection problems make traditional monitoring methods difficult to use outside of clinical environments. The computer vision techniques have been shown that some physiological variables as heart rate can be measured without contact. Video magnification is one of these approach used for the detection of the pulse signal. In this paper we propose a new strategy to magnify motion in a video sequence using the Hermite transform. In addition a deep learning technique is implemented to estimate the beat by beat pulse signal. We trained the system and validated our results using an electronic pulse monitoring device. Our approach is compared with the classical video magnification using a Gaussian pyramid. The results show a better enhancement of spectral information from the colour changes allowing an accurate estimation of the instantaneous beat by beat pulse than the Gaussian approach.

Keywords: General and theoretical informatics - Security and authentication, General and theoretical informatics - Pattern recognition, Sensor Informatics - Wearable systems and sensors
Abstract: As a continuation from a previous work [1], this paper shows the results obtained when merging ECG information with fingerprints, for biometric authentication. Two approaches have been considered, and results reported.

Keywords: General and theoretical informatics - Security and authentication, Sensor Informatics - Multi-sensor data fusion, Sensor Informatics - Wearable systems and sensors
Abstract: Recently, user authentication technology using fingerprint have been used for mobile payment services like as Apple Pay and Samsung Pay, but some of the world population difficult to obtain fingerprint patterns due to fingerprint damage. So, development of bio-signal authentication technology using behavioral biometrics such as EEG (Electroencephalogram), and ECG (Electrocardiogram) is accelerating. Hence, in this paper, we develop the telebiometric authentication technologies using bio-signals can provide robust authentication.

Keywords: Sensor Informatics - Sensors and sensor systems, Sensor Informatics - Body sensor networks, Sensor Informatics - Wearable systems and sensors
Abstract: Biometric technologies have become widely used in daily life. However, biometric technologies using existing physical features can be concerned about the forgery such as fake fingerprints. For next-generation biometric technologies, behavioral biometrics of a living person such as bio-signals and walking become important for personal identification and healthcare purposes. In this paper, we present sensor requirements and harmonized standard data format for personal authentication based on several bio-signal measurements.

Keywords: General and theoretical informatics - Security and authentication, General and theoretical informatics - Pattern recognition, Sensor Informatics - Wearable systems and sensors
Abstract: In this study, we developed wrist-worn wearable system that can identify individuals using electrocardiogram (ECG) signal based on wrist-worn wearables. We examined the capability of heart signal as personal authentication. Two hundred six volunteers were recruited and were measured ECG twice from their wrists in resting and post-exercise state. We used discrete wavelet transform method to extract features of ECG and classified test data for 1:1 matching authentication by Euclidean distance-based threshold method. The best results was 9.71% of equal error rate (EER) in resting state condition.

Keywords: Sensor Informatics - Wearable systems and sensors, Health Informatics - Quality of service, trust, security, Sensor Informatics - Sensor-based mHealth applications
Abstract: This document provides a method for secure transmission protocol and power optimization for EGG (electrocardiogram) and PPG (photoplethysmography) wireless applications. The focus of the topic is implementation of security and power management by the application protocol for Fintech and medical applications.

Keywords: Fetal and Pediatric Imaging, Image registration, segmentation, compression and visualization - Machine learning / Deep learning approaches, Image analysis and classification - Machine learning / Deep learning approaches
Abstract: Machine learning approaches for image analysis require large amounts of training imaging data. As an alternative, the use of realistic synthetic data reduces the high cost associated to medical image acquisition, as well as avoiding confidentiality and privacy issues, and consequently allows the creation of public data repositories for scientific purposes. Within the context of fetal imaging, we adopt an auto-encoder based Generative Adversarial Network for synthetic fetal MRI generation. The proposed architecture features a balanced power of the discriminator against the generator during training, provides an approximate convergence measure, and enables fast and robust training to generate high-quality fetal MRI in axial, sagittal and coronal planes. We demonstrate the feasibility of the proposed approach quantitatively and qualitatively by segmenting relevant fetal structures to assess the anatomical fidelity of the simulation, and performing a clinical verisimilitude study distinguishing the simulated data from the real images. The results obtained so far are promising, which makes further investigation on this new topic worthwhile.

Keywords: Image analysis and classification - Machine learning / Deep learning approaches, Image classification, Image feature extraction
Abstract: Accurate estimation of food’s macronutrient content for people with Diabetes Mellitus (DM) is of great importance, as it determines postprandial insulin dosage. This paper introduces a classification system for food images that is adjusted to the nutritional needs of people with DM. A two-level image classification scheme, exploiting Convolutional Neural Networks (CNNs), is proposed, in order to classify an image in one of eight broad food categories with similar macronutrient content and then assign it to a specific food within that category. To this end, a visual dataset, namely NTUA-Food 2017, has been designed, consisting of 3248 images organized in eight broad food categories of totally 82 different foods. Moreover, a novel evaluation metric is proposed, which penalizes classification errors proportionally to the discrepancy in postprandial blood sugar levels between the actual and predicted class. The proposed system achieves 84.18% and 85.94% classification accuracy at the first and second level, respectively, of classification on the NTUA-Food 2017 dataset. The algorithm developed for the first level of classification on the NTUA-Food 2017 dataset improves classification accuracy on the benchmark Food Image Dataset (FID) to 97.08% outperforming previous approaches. The algorithm’s mean error in terms of carbohydrate content estimation on the NTUA-Food 2017 dataset is less than 2 g per food serving.

Keywords: Image feature extraction, Image classification
Abstract: Automatic analysis of complex rodent social behavior, especially aggressive ones, is of important scientific interest. In this paper we analyze the properties of the data created as a result of aggressive rodent social behavior. Detection of specific aggressive behaviors is based on the event of leaving traces of saliva on the fur of the attacked individual, which are clearly visible in the thermal imaging. The traces change temperature in time in a specific way. After bite, saliva is cooled and then heated to the body temperature. Usage of this method in social behavior analysis ensures detection and tracking aggressive behaviors even if the event itself is invisible.

Keywords: Magnetic resonance imaging - Interventional MRI, Magnetic resonance imaging - MR angiographic imaging, Image enhancement
Abstract: Susceptibility Weighted Imaging (SWI) is a method extensively studied for its application to improve contrast in MR imaging modality. The method enhances the visualization of metallic content such as iron, calcium and zinc in the tissues by using the susceptibility differences in tissues to generate a unique image contrast. In this study, we propose an SWI based approach to improve the visualization of interventional devices in MRI data. Results obtained from two datasets (biopsy needle and brachytherapy seeds), indicate SWI to be suitable for visualization of the interventional devices, while also being computationally faster when compared with QSM.

Keywords: X-ray radiography, Image reconstruction and enhancement - Machine learning / Deep learning approaches, Iterative image reconstruction
Abstract: Dental panoramic radiography (DPR), a widely used medical examination method, has its intrinsic weakness in high requirement to the positioning of patient. Although positioning devices like chin support can provide a relatively stable and guaranteed environment for exposure, problems including morphological differences of jaw between patients and their improper standing postures still put the reconstructed image at high risk of getting blurred, especially in the anterior segment of dental arch. This paper proposes a novel method based on convolutional neural network (CNN) to estimate the positioning error of patient’s dental arch, and thereby reconstruct the panoramic image with the corrected dental curvature, so that the blur gets reduced. Experiment results demonstrate the method’s effectiveness in providing reconstructed images of stable quality for further diagnosis.

Keywords: Neural signals - Blind source separation (PCA, ICA, etc.), Brain physiology and modeling - Sensory-motor, Neurological disorders - Diagnostic and evaluation techniques
Abstract: Eye movements reveal a great wealth of information about the visual system and the brain. Therefore, eye movements can serve as diagnostic markers for various neurological disorders. For an objective analysis, it is crucial to have an automatic and robust procedure to extract relevant eye movement parameters. An essential step towards this goal is to detect and separate different types of eye movements such as fixations, saccades and smooth pursuit. We have developed a model-based approach to perform signal detection and separation on eye movement recordings, using source separation techniques from sparse Bayesian learning. The key idea is to model the oculomotor system with a state space model and to perform signal separation in the neural domain by estimating sparse inputs which trigger saccades. The algorithm was evaluated on synthetic data, neural recordings from rhesus monkeys and on manually annotated human eye movement recordings with different smooth pursuit paradigms. The developed approach shows a high noise-robustness, provides saccade and smooth pursuit parameters, as well as estimates of the position, velocity and acceleration profiles. In addition, by estimating the input to the oculomotor system, we obtain an estimate of the neural inputs to the oculomotor muscles.

Keywords: Neural signals - Blind source separation (PCA, ICA, etc.), Motor neuroprostheses - Epidural stimulation, Motor learning, neural control, and neuromuscular systems
Abstract: Muscle synergies encode motor activity as a linear superposition of multiple motor units composed of a temporal command exciting a specific network of muscles. This study examines muscle synergies derived from simple standing studies of a complete spinal cord injury (SCI) patient under epidural spinal stimulation. A popular technique for extracting these synergies from EMG data is non-negative matrix factorization (NNMF). However, standard NNMF algorithms do not allow for physiological delays for a neural signal to reach different muscles. These delays are prevalent in SCI patients under spinal stimulation, and so we propose a new algorithm (regularized ShiftNMF) to extract muscle synergies which account for signal delays. We find muscle synergies extracted by the regularized ShiftNMF algorithm are significantly better at reconstructing EMG activity, and the resulting features are physiologically consistent and more useful in describing patient behavior.

Keywords: Neural signals - Blind source separation (PCA, ICA, etc.), Neural interfaces - Microelectrode technology, Brain physiology and modeling - Neuron modeling and simulation
Abstract: In neural electrophysiology, spike sorting allows to separate different neurons from extracellularly measured recordings. It is an essential processing step in order to understand neural activity and it is an unsupervised problem in nature, since no ground truth information is available. There are several available spike sorting packages, but many of them require a manual intervention to curate the results, which makes the process time consuming and hard to reproduce. Here, we focus on high-density Multi-Electrode Array (MEA) recordings and we present a fully automated pipeline based on Independent Component Analysis (ICA). While ICA has been previously investigated for spike sorting, it has never been compared with fully automated state-of-the-art algorithms. We use realistic simulated datasets to compare the spike sorting performance in terms of complexity, signal-to-noise ratio, and recording duration. We show that an ICA-based fully automated spike sorting approach can be a viable alternative approach due to its precision and robustness, but it needs to be optimized for time constraints and requires sufficient density of electrodes to cover active neurons in the proximity of the MEA.

Keywords: Neural signal processing, Brain physiology and modeling
Abstract: Simultaneous recordings of spikes and fields could enable analyses of functional connectivity in the brain at multiple spatiotemporal scales. However, these analyses require developing novel methods to assess causality between binary-valued spikes and continuous-valued fields, which have fundamentally different statistical profiles and time-scales. Thus classical measures of causality cannot be directly applied in multiscale networks. We develop a novel parametric method to assess causality for multiscale spike-field activities by computing directed information. Directed information is an information theoretic measure of causality but is in general hard to estimate. Our method estimates the causality in two steps. First, we construct point process generalized linear models (GLM) for each neuron's spiking activity to estimate its firing rate using the history of both spikes and fields and compute the directed information to spike nodes from any node. Second, we construct regression models for fields using the history of the estimated firing rates and the history of fields, and then compute the directed information to each field node from any node. In both steps, we estimate model parameters using maximum likelihood and devise statistical tests to assess the significance of the causality. Using simulated data from basic three-node structures and a ten-node network, we show that our method can asymptotically identify the true causality. This method could help uncover functional connectivity in the brain at multiple spatiotemporal scales.

Keywords: Neural signal processing, Brain physiology and modeling, Motor neuroprostheses
Abstract: Behavior is encoded across spatiotemoral scales of brain activity, from small-scale spikes to large-scale local field potentials (LFP). Identifying the functional dependence between spikes and LFP networks during behavior can help understand neural encoding and improve future neurotechnologies, but is difficult to achieve. First, spikes and LFP have different statistical characteristics (binary spikes vs. continuous LFPs) and time-scales. Second, given the prohibitively large number of spike channels and LFP features recorded in today's experiments, learning dependencies between all recorded signals is challenging and prone to overfitting. To solve this challenge, we present a model-based approach to estimate the functional dependence between high-dimensional field features and neuronal spikes. We model the binary time-series of spikes for each neuron as a point process dependent on the behavioral states and LFP features across the network. Given the prohibitively large number of possible spike-LFP dependency parameters, we first employ an L1-regularization technique to learn the point process model during both supervised and unsupervised learning to ease detection of significant dependency parameters. We then use the Akaike information criterion (AIC) to enforce model sparsity by incorporating only a minimum number of non-zero dependency parameters into the point process model based on a trade-off between model complexity and its prediction power. Using extensive numerical simulations, we show that our method (i) can correctly identify the functional dependencies and thus improve the prediction of spiking activity and (ii) can improve the prediction of spiking activity with significantly fewer number of parameters compared to when regularization is not enforced. Our approach may serve as a tool to investigate brain connectivity patterns across spatiotemporal scales.

Keywords: Neural signal processing
Abstract: Wearable measurement for electroencephalogram (EEG) is expected to enable brain-computer interfaces, biomedical engineering, and neuroscience studies in real environments. When wearable devices are in practical use, only the user (subject) can take care of measurement, unlike laboratory-oriented experiments where experimenters are always with the subject. As a result, measurement troubles such as high impedance or electrode impairment cannot be easily corrected, and EEG recordings will become incomplete, including many missing values. If the missing values are imputed (interpolated) and complete data without missing entries are available, we can employ existing signal analysis techniques that assume compete data. In this paper, we propose an EEG signal imputation method based on multivariate autoregressive (MAR) modeling and iterative imputation, inspired by the multiple imputation procedure. We evaluated the proposed method with real datasets with artificial missing entries. Experimental results shows that the proposed method outperforms popular baseline interpolation methods. Our iterative scheme is simple yet effective, and can be the foundation for many extensions.

Keywords: Signal pattern classification, Physiological systems modeling - Signal processing in physiological systems, Data mining and processing - Pattern recognition
Abstract: Wearable sensors for upper limbs enable the use of myoelectric control systems in real environments. An important issue in the practical use of myoelectric control is how to deal with the variations of electromyograms (EMGs); the distribution of EMGs changes over days and device (electrode) positions. The amount of training data is usually limited, as data are collected at the beginning of the system use. Transfer learning has been employed to make full use of the limited-amount training data by compensating for the difference of EMGs over time and device placement. However, it was unclear how transfer learning improved the motion recognition accuracy over long-term use with varying device positions. In this paper, we evaluated transfer learning algorithms on one-month long data with three different device positions. We found that transfer learning was able to compensate for the variations over long period and also over different electrode placements, suggesting the practical efficacy of transfer learning.

Keywords: Physiological systems modeling - Signal processing in physiological systems, Time-frequency and time-scale analysis - Empirical mode decomposition in biosignal analysis, Physiological systems modeling - Signals and systems
Abstract: A motor unit potential (MUP) template, which represents the shapes of the MUPs within a MUP train, provides information related to the morphology and physiology of the sampled motor unit. This work presents an improved MUP template estimation technique that uses local time warping and kernel weighted ensemble averaging. An analysis of the algorithm, and a description of the improvements compared with spike triggered averaging is given. MUP template estimates were evaluated using simulated EMG signals with a known gold standard template for each motor unit potential train. Statistically significant reduction in template estimation error is shown, both within the baseline and duration portions of a MUP.

Keywords: Signal pattern classification, Time-frequency and time-scale analysis - Wavelets, Physiological systems modeling - Signal processing in physiological systems
Abstract: An electromyogram (EMG) signal acquisition system capable of real time classification of several facial gestures is presented. The training data consist of the facial EMG collected from 10 individuals (5 female/5 male). A custom-designed sensor interface integrated circuit (IC) consisting of an amplifier and an ADC, implemented in 65nm CMOS technology, has been used for signal acquisition [1]. It consumes 3.8nW power from a 0.3V battery. Feature extraction and classification is performed in software every 300ms to give real-time feedback to the user. Discrete wavelet transforms (DWT) are used for feature extraction in the time-frequency domain. The dimensionality of the feature vector is reduced by selecting specific wavelet decomposition levels without compromising the accuracy, which reduces the computation cost of feature extraction in embedded implementations. A support vector machine (SVM) is used for the classification. Overall, the system is capable of identifying several jaw movements such as clenching, opening the jaw and resting in real-time from a single channel EMG data, which makes the system suitable for providing biofeedback during sleeping and awake states for stress monitoring, bruxism, and several orthodontic applications such as temporomandibular joint disorder (TMJD).

Keywords: Signal pattern classification, Time-frequency and time-scale analysis - Time-frequency analysis, Data mining and processing - Pattern recognition
Abstract: Increasing performance while decreasing the cost of sEMG prostheses is an important milestone in rehabilitation engineering. The different types of prosthetic hands that are currently available to patients worldwide can benefit from more effective and intuitive control. This paper presents a real-time approach to classify finger motions based on surface electromyography (sEMG) signals. A multichannel signal acquisition platform implemented using components off the shelf is used to record forearm sEMG signals from 7 channels. sEMG pattern classification is performed in real time, using a Linear Discriminant Analysis approach. Thirteen hand motions can be successfully identified with an accuracy of up to 95.8% and of 92.7% on average for 8 participants, with an updated prediction every 192 ms.

Keywords: Physiological systems modeling - Signal processing in physiological systems, Nonlinear dynamic analysis - Biomedical signals, Nonlinear dynamic analysis - Deterministic chaos
Abstract: Muscle fatigue is the inability to exert the required force. Surface Electromyography (sEMG) is a technique used to study the muscle’s electrical property. These generated signals are complex and nonstationary in nature. In this work, an attempt is made to utilize graph signal processing methods such as Sequential Visibility motif for the analysis of muscle fatigue condition. The sEMG signals of 41 healthy adult volunteers are acquired from the biceps brachii muscle during isometric contraction with a 6 Kg load. The subjects are asked to perform the exercise until they are unable to continue. The signals are preprocessed, and the first and last 500 ms of the signal are considered for analysis. The segmented signals are subjected to sequential visibility graph algorithm. Further, the number of motifs for a subgraph of four is calculated. The results show that the signals are unique for each subject. The frequency of higher degree motif is more in the case of fatigue. The frequency of each unique motif is capable of differentiating nonfatigue and fatigue conditions. Nonparametric statistical test result indicates all features are significant with p<0.05. This method of analysis can be extended to other varied neuromuscular conditions.

Keywords: Physiological systems modeling - Signal processing in physiological systems, Nonlinear dynamic analysis - Biomedical signals, Nonlinear dynamic analysis - Deterministic chaos
Abstract: In this work, an attempt has been made to analyze the preterm (gestation period ≤ 37 weeks) condition using uterine electromyography (EMG) signals and multifractal detrended fluctuation analysis (MFDFA). The signals recorded using surface electrodes placed on the abdomen is used for this study and these are obtained from a publically available online database. These signals are preprocessed using 4-pole digital Butterworth filter. The preprocessed signals are subjected to MFDFA to extract multifractal features namely maximum singularity exponent, peak singularity exponent, strength of multifractality and exponent index. Generalized Hurst exponent extracted from the signals indicate that uterine EMG signals show multifractal behavior in preterm condition. Among the extracted features the coefficient of variation is found to be lower for peak singularity exponent. This indicates that this feature have lower inter-subject variability. Hence, it appears that the multifractal features can help in the assessment of uterine EMG signals for preterm detection.

Keywords: Neural stimulation, Neural interfaces - Microelectrode technology, Neural interfaces - Tissue-electrode interface
Abstract: Investigating learning in networks of spinal cord neurons can provide insight into the dynamics of connectivity in human spinal cords. It may also hold implications for developing neural prosthetics and neurocomputers. Culturing neural networks on microelectrode arrays (MEAs) allows for the repeated observation and stimulation of electrophysiological activity in vitro. Here we used MEAs to demonstrate learning in networks of spinal cord neurons. This was done by exposing E17 mouse spinal cord cultures to high frequency artificial spike trains, or tetanization. Unexpectedly, when comparing the networks’ responses to low-frequency probing stimulations before and after tetanization, the cultures were found to demonstrate long-term depression (LTD). LTD was most significantly observed between 500-1000 ms after low-frequency probing. These results indicate that periodic high-frequency excitation of spinal cord networks can result in decreased synaptic efficacy.

Keywords: Neural stimulation, Neuromuscular systems - Peripheral mechanisms, Neuromuscular systems - EMG processing and applications
Abstract: High-Frequency alternating current (HFAC) nerve block has great potential for neuromodulation-based therapies. However no precise measurements have been made of the time needed for nerves to recover from block once the signal has been turned off. This study aims to characterise time to recovery of the rat sciatic nerve after 30 seconds of block at varying amplitudes and frequencies. Experiments were carried out in-vivo to quantify recovery times as well as recovery completeness within 0.7s from the end of block. The sciatic nerve was blocked with an alternating square wave signal of amplitude and frequency ranging from 2 to 9 mA and 10 to 50 kHz respectively, then stimulated at 100 Hz during recovery to reduce error to within 10 ms for measurements of recovery dynamics. The electromyogram (EMG) signals were measured from the gastrocnemius medialis and tibialis anterior muscles during trials as indicators of nerve function. Recovery times ranged from 20 to 430 milliseconds and final recovery values at 0.7 seconds spanned 0.2 to 1 approximately. Higher blocking signal amplitudes increased recovery time and decreased recovery completeness. These results suggest that blocking signal properties affect nerve recovery dynamics, which could help improve neuromodulation therapies and allow more precise comparing of results across studies using different blocking signal parameters.

Keywords: Neural stimulation, Neural stimulation - Deep brain, Brain-computer/machine interface
Abstract: Non-invasive brain stimulation of small animals plays an important role in neuroscience especially in understanding fundamental mechanisms of brain disorders. Here, we report a miniaturized ultrasound transducer array designed for non-invasive brain stimulation of mouse for the first time. We designed and fabricated a Capacitive Micromachined Ultrasonic Transducer (CMUT) ring array that operates at 183 kHz in immersion. The fabricated transducer ring array exhibited a focal length of 2.25 mm and a maximum intensity of 175 mW/cm2. In addition, the focus region reached a depth of 4 mm. Because the transducer array is small and lightweight, a compact packaging with minimum surgical procedures was sufficient to perform in vivo mouse experiments. Using the developed micromachined transducer array and simple packaging, we successfully induced the motor responses of a mouse. The success rate of ultrasound stimulation was quantified by recording the electromyography (EMG) signal during the stimulation. While the current ultrasound neuromodulation system is limited to acute experiments, this proposed lightweight, miniaturized transducer (< 1 g) with a natural focus would enable chronic ultrasound neuromodulation experiments on freely-moving mice for the first time.

Keywords: Neural stimulation - Deep brain, Brain-computer/machine interface
Abstract: Ultrasound neuromodulation is a promising stimulation modality because of its non-invasiveness, focusing and steering capability, and relatively high spatial resolution compared to the other stimulation modalities. However, despite the high lateral resolution, the ultrasound beam in the axial direction is relatively long, especially when compared to the small size of the mouse brain. Here, we report a new ultrasound focusing technique for small animal in vivo experiments where a high spatial resolution in both lateral and axial directions is achieved by crossing two ultrasound beams. The focal volume of a full width half maximum (FWHM) of our proposed system is only 0.161 mm3 and the focal diameter in the axial direction is about 1 mm, which is ten times smaller than the previously reported ultrasound neuromodulation system. Thus, the proposed system enables targeting a sub-region of a mouse brain using ultrasound for the first time. We also demonstrate successful stimulation of the motor cortex through in vivo mice experiments where the movement of forepaw of the mouse was observed using the double-crossed ultrasound transducers. Moreover, by sweeping the focal point in the z-axis and measuring the success rate of stimulated movements, we show that our double-transducer system targeted a region with 2 mm-resolution in the dorsal-ventral (DV) coordinates. The success rate of the double-crossed ultrasound stimulation was quantified by recording the electromyography (EMG) signals during the stimulation. Our results show that the double-crossed ultrasound transducer system with a ten times higher spatial resolution enables highly specific and noninvasive stimulation of small animals and thus enables versatile in vivo experiments to study functional connectivities of brain circuits.

Keywords: Neural stimulation - Deep brain, Brain physiology and modeling - Cognition, memory, perception, Neurological disorders - Epilepsy
Abstract: There is a great need for an electrical stimulation therapy to treat medication-resistant, surgically ineligible epileptic patients that successfully reduces seizure incidence with minimal side effects. Critical to advancing such therapies will be identifying the trade-offs between therapeutic efficacy and side effects. One novel treatment developed in the tetanus toxin rat model of mesial temporal lobe epilepsy, asynchronous distributed microelectrode stimulation (ADMETS) in the hippocampus has been shown to significantly reduce seizure frequency. However, our results have demonstrated that ADMETS has a negative effect on spatial memory that scales with the amplitude of stimulation. Given the high dimensional space of possible stimulation parameters, it is difficult to construct a mapping from variations in stimulation to behavioral effect. In this project, we present a novel, principled approach using closed-loop Bayesian optimization to tune stimulation that successfully maximize a desired behavioral objective - performance on a spatial memory assay.

Keywords: Neural stimulation, Brain physiology and modeling, Neural signal processing
Abstract: Motor-evoked potentials (MEP) are one of the most important responses to brain stimulation, such as suprathreshold transcranial magnetic stimulation (TMS) and electrical stimulation. The understanding of the neurophysiology and the determination of the lowest stimulation strength that evokes responses requires the detection of even smallest responses, e.g., from single motor units, but available detection and quantization methods are rather simple and suffer from a large noise floor. The paper introduces a more sophisticated matched-filter detection method that increases the detection sensitivity and shows that activation occurs well below the conventional detection level. In consequence, also conventional threshold definitions, e.g., as 50 µV median response amplitude, turn out to be substantially higher than the point at which first detectable responses occur. The presented method uses a matched-filter approach for improved sensitivity and generates the filter through iterative learning from the presented data. In contrast to conventional peak-to-peak measures, the presented method has a higher signal-to-noise ratio (≥14 dB). For responses that are reliably detected by conventional detection, the new approach is fully compatible and provides the same results but extends the dynamic range below the conventional noise floor. The underlying method is applicable to a wide range of well-timed biosignals and evoked potentials, such as in electroencephalography.

Keywords: Signal pattern classification, Time-frequency and time-scale analysis - Wavelets, Data mining and processing - Pattern recognition
Abstract: Detection of Electrocardiogram (ECG) characteristic points can provide critical diagnostic information about heart diseases. We propose a novel feature extraction and machine learning scheme for ECG delineation. A new feature, termed as randomly selected wavelet transform (RSWT), is proposed to effectively represent ECG morphology. With the RSWT feature pool, a regression tree is trained to estimate the probability distribution to the direction toward the target point, relative to the current position. The continual random walk through 1D space will eventually produce a reliable region from which the final position of the target point is derived. The evaluation results on QT database show better detection accuracy compared with other studies while providing real-time processing capability.

Keywords: Parametric filtering and estimation, Time-frequency and time-scale analysis - Empirical mode decomposition in biosignal analysis, Signal pattern classification
Abstract: The electrocardiogram (ECG) is commonly used to monitor or diagnose adverse heart conditions. While general ECG recordings are widely available, parametric ECG models have been proposed to generate ECG-like signals. Such ECG generators can create extended segments of specific beat morphology or cardiac rhythm, especially in disease states, which can be used to validate cardiac devices or evaluate ECG processing algorithms. Furthermore, if the parameters can be fit to a variety of ECGs, these models are valuable tools in ECG compression and modeling. In this paper we propose a framework to fit parameter values of an ECG generator such that the generated signal is similar to a reference signal. We first design a parametric ECG generator with relatively minimal assumptions of single beat waveform morphology. We then use Particle Swarm Optimization to find ideal values for parameters of our ECG generator which minimize the percent root mean square difference (PRD) between the reference and generated signals. We were able to capture waveform morphologies of normal, idioventricular, and ventricular flutter rhythms with Pearson correlation coefficients above 0.9 between generated and pre-recorded signals from the MIT-BIH database.

Keywords: Time-frequency and time-scale analysis - Time-frequency analysis, Nonlinear dynamic analysis - Biomedical signals, Data mining and processing in biosignals
Abstract: Physical, environmental and psychological stressors can lead to chronic diseases. Changes in physiological parameters during stress can be evaluated using heart rate variability (HRV). A study is conducted with one participant on a daily basis over six months. Bathtub ECG is measured during his daily bathing and used for HRV analysis. The HRV stress index (SI) is computed to assess the stress level based on time-, frequency-domain and nonlinear features. Daily SI is computed and found to be significantly higher in either mental stress or physiological stress. The variations of SI show in high accordance with the work schedules of the participant.

Keywords: Time-frequency and time-scale analysis - Time-frequency analysis
Abstract: Photoplethysmography (PPG) signals collected from wearable sensing devices during physical exercise are easily corrupted by motion artifact (MA), which poses great challenge on heart rate (HR) estimation. This paper proposes a new framework to accurately estimate HR using two leads of PPG signals in combination with accelerometer (ACC) data in the presence of MA. A moving time window is first used to segment PPG signals and ACC signals. Then, MA is attenuated by joint sparse spectrum reconstruction in each time window, where maximum spectrum frequencies of ACC are subtracted from the spectrum frequency of PPG signals. Further, HR for each cleansed PPG is estimated from the frequency with maximum amplitude in the sparse spectrum. The actual HR is determined using spectral band powers calculated from each reconstructed PPG signals. The proposed method was validated using the 2015 IEEE Signal Processing Cup dataset. The average absolute error is 1.15 beats per minutes (BPM) (standard deviation: 2.00 BPM), and the average absolute error percentage is 0.95% (standard deviation: 1.86%). The proposed method outperforms the previously reported work in terms of accuracy.

Keywords: Time-frequency and time-scale analysis - Empirical mode decomposition in biosignal analysis, Neural networks and support vector machines in biosignal processing and classification, Signal pattern classification
Abstract: Heart disease classification based on electrocardiogram(ECG) signal has become a priority topic in diagnosis of heart diseases because it can be obtained with a simple diagnostic tool of low cost. Since early detection of heart disease can enable us to ease the treatment as well as save people's lives, accurate detection of heart disease using ECG is very important. In this paper, we propose a classification method for heart diseases based on ECG by adopting a machine learning method, called Long Short-Term Memory (LSTM), which is a state-of-the-art technique analyzing time series sequences in deep learning. As suitable data preprocessing, we also use symbolic aggregate approximation (SAX) to improve the accuracy. Our experiment results show that our approach not only achieves significantly better accuracy, but also classifies heart diseases correctly in a smaller response time, than baseline techniques.

Keywords: Time-frequency and time-scale analysis - Empirical mode decomposition in biosignal analysis, Physiological systems modeling - Signal processing in physiological systems, Physiological systems modeling - Signals and systems
Abstract: recent research has shown that each apnea episode results in a significant rise of the beat-to-beat blood pressure followed by a drop to the pre-episode levels when patient resumes normal breathing. While the physiological implications of these repetitive and significant oscillations are still unknown, it is of interest to quantify them. Since current array of instruments deployed for polysomnography studies does not include beat-to-beat measurement of blood pressure, but includes oximetry which can supply pulsatile photoplethysmography (PPG) signal, in addition to percent oxygen saturation. Hence, we have investigated a new method for continuous estimation of systolic (SBP), diastolic (DBP), and mean (MBP) blood pressure waveforms from PPG. Peaks and troughs of PPG waveform are used as input to a 5th order autoregressive moving average model to construct estimates of SBP, DBP, and MBP waveforms. Since breath hold maneuvers are shown to faithfully simulate apnea episodes, we evaluated the performance of the proposed method in 7 subjects (4 F; 32±4 yrs., BMI 24.57±3.87 kg/m2) in supine position doing 5 breath holding maneuvers with 90s of normal breathing between them. The modeling error ranges were (all units are in mmHg) -0.88±4.87 to -2.19±5.73 (SBP); 0.29±2.39 to -0.97±3.83 (DBP); and -0.42±2.64 to -1.17±3.82 (MBP). The cross validation error ranges were 0.28±6.45 to -1.74±6.55 (SBP); 0.09±3.37 to -0.97±3.67 (DBP); and 0.33±4.34 to -0.87±4.42 (MBP). The overall level of estimation error, as measured by the root mean squared of the model residuals, was less than 7 mmHg.

Keywords: Optical imaging - Coherence tomography, Ophthalmic imaging and analysis, Image analysis and classification - Machine learning / Deep learning approaches
Abstract: Accurate detection of diabetic macular edema (DME) is an important task in optical coherence tomography (OCT) images of the eye. A relatively simple and practical approach is proposed in this paper. A pre-trained convolutional neural network (CNN) is fine tuned for a classification of DME versus normal cases. The fine-tuned Inception-Resnet-v2 CNN model can effectively identify pathologies in comparison to classical learning. Experiments were carried out on the publicly available data set of the Singapore Eye Research Institute (SERI). The developed model was also compared to other fine tuned models, such as Resnet-50 and Inception-v3. The proposed method achieved 100% classification accuracy with the Inception- Resnet-v2 model using a leave-one-out cross-validation strategy. For robustness, the model trained on the SERI dataset was tested on another dataset provided by the Chinese University HongKong (CUHK), also with 100% accuracy. The proposed method is a potentially impactful tool for accurately detecting DME vs. normal cases.

Keywords: Image classification, Image feature extraction
Abstract: Abstract— This study investigates the possible existence of different natural corneal shape categories. This is important to better describe cornea for both diagnostic and therapeutic assessments. We started by describing corneal shape of different populations as a function of influencing clinical data i.e. age, ametropia and gender. This was done by averaging Zernike polynomial (ZP) decomposition of the anterior surfaces in each subgroup. The results showed small but significant differences of shape that are supported by the literature. This motivated us to examine the feasibility of characterizing the normal corneal shape with an automatic method of clustering independent of any clinical a priori knowledge. Since we did not know beforehand the number of corneal categories, agglomerative hierarchical clustering was applied on ZP coefficients for a large database. The dendrogram based on the Ward’s distance was evaluated with two different clustering validity indexes (coefficient of determination R^2 and semi partial R^2,〖SPR〗^2). The optimal number of categories was around four showing corneal shapes ranging from flatter to steeper.

Keywords: Ophthalmic imaging and analysis, Image analysis and classification - Machine learning / Deep learning approaches
Abstract: Diabetic Retinopathy (DR) is a non-negligible eye disease among patients with Diabetes Mellitus, and automatic retinal image analysis algorithm for the DR screening is in high demand. Considering the resolution of retinal image is very high, where small pathological tissues can be detected only with large resolution image and large local receptive field are required to identify those late stage disease, but directly training a neural network with very deep architecture and high resolution image is both time computational expensive and difficult because of gradient vanishing/exploding problem, we propose a Multi-Cell architecture which gradually increases the depth of deep neural network and the resolution of input image, which both boosts the training time but also improves the classification accuracy. Further, considering the different stages of DR actually progress gradually, which means the labels of different stages are related. To considering the relationships of images with different stages, we propose a Multi-Task learning strategy which predicts the label with both classification and regression. Experimental results on the Kaggle dataset show that our method achieves a Kappa of 0.841 on test set which is the 4-th rank of all state-of-the-arts methods. Further, our Multi-Cell Multi-Task Convolutional Neural Networks (M2CNN) solution is a general framework, which can be readily integrated with many other deep neural network architectures.

Keywords: Image segmentation, Image feature extraction, Image analysis and classification - Digital Pathology
Abstract: More than 8% of world population have diabetes which causes long term complications such as retinopathy, neuropathy, nephropathy and foot ulcers. Growing patient numbers has prompted large scale screening methods to detect early symptoms of diabetes (rather than elevated blood glucose levels which is a late symptom). Vascular tortuosity (twisted and curved nature of blood vessels) in retinal fundus images has proven to reflect the effect of diabetes on macrovasculature. However, large scale patient screening using retinal fundus images has limitations due to the requirement of a retinal camera. Therefore, we hypothesize that the vasculature of superior bulbar conjunctiva which could be captured using a regular camera could be used to measure tortuosity instead of retinal fundus images enabling mass screening. To test this hypothesis, a total of 168 scleral images were acquired from 50 healthy subjects and 34 diabetic patients using a digital camera. The sclera region was segmented using Chan-Vese algorithm and macrovasculature of superior bulbar conjunctiva was segmented using B-COSFIRE filters. Results revealed that the superior bulbar conjunctival macrovascular tortuosity of diabetic patients was significantly less than that of non-diabetic group (p-value = 0.015). A similar result was yielded (p-value = 0.049) from a group of participants who were less than 40 years old which excluded the age related variation of tortuosity.

Keywords: Image segmentation, Image analysis and classification - Machine learning / Deep learning approaches, Image registration, segmentation, compression and visualization - Machine learning / Deep learning approaches
Abstract: Automatic segmentation of vascular network is a critical step in quantitatively characterizing vessel remodeling in retinal images and other tissues. We proposed a deep learning architecture consists of 14 layers to extract blood vessels in fundoscopy images for the popular standard datasets DRIVE and STARE. Experimental results show that our CNN characterized by superior identifying for the foreground vessel regions. It produces results with sensitivity higher by 10% than other methods when trained by the same data set and more than 1% with cross training (trained on DRIVE, tested with STARE and vice versa ) . Further, our results have better accuracy > 0.95 % compared to state of the art algorithms.

Keywords: Neurological disorders - Psychiatric disorders, Neural signal processing, Brain functional imaging - Classification
Abstract: Adolescent Major Depressive Disorder (MDD) is a common and serious mental illness that could lead to tragic outcomes including chronic adult disability and suicide. In this paper, we explore anatomical features and apply machine learning approaches to identify responsive biomarkers distinguishing MDD patients from healthy subjects. The features of interest include metrics in two categories: a) anatomical connectivity defined by diffusion tensor imaging measurements between a pair of brain regions, and b) topological measurements from anatomical networks. A combination of p-value based filtering and minimum redundancy maximum relevance method is performed to select features for optimal classification accuracy. A leave-one-out cross-validation method is used for the classification performance evaluation. The proposed methodology achieves an improved accuracy of 78%, 90.39% sensitivity, and 79.66% precision for 79 subjects. The most distinguishing features are the betweenness centrality of the right lingual gyrus of the ADC network at 12% sparsity, the participation coefficient of the right lateral occipital sulcus of the ADC network at 22% sparsity, the participation coefficient of the right pars opercularis of the AD network at 16% sparsity, and the participation coefficient of the right lateral orbitofrontal cortex in the ADC network at 10% sparsity. Those network measures reflect the change of connectivity between the regions and their associated anatomical subnetworks.

Keywords: Neurological disorders - Stroke
Abstract: Photothrombosis is a technique that can induce ischemic cortical infarcts using the photodynamic effect of anionic xanthene dyes, typically Rose Bengal, to cause occlusion of cerebral blood circulation. The ability to quantitatively predict the scale of the lesion in photothrombotic procedures can offer crucial insight in the development and implementation of light-induced stroke models in animals. In this article, we introduced a quantitative model that could estimate the normalized light intensity distribution in tissue which scatters photons from a collimated beam. We simulated the penetration and scattering profile of light of Rose Bengal’s characteristic absorption wavelengths in mouse cortex. We further illustrated that our model could estimate the spatial extent of effective region under photothrombotic protocols, and how this model can be used to titrate the intensity and geometry of light beams used to generate infarcts of desired dimensional characteristics.

Keywords: Neurological disorders - Mechanisms, Human performance - Sensory-motor, Brain physiology and modeling - Cognition, memory, perception
Abstract: Parkinson's Disease (PD) is typically classified by the onset of motor impairments, however, non-motor symptoms are also present in all disease stages. Vision abnormalities contribute to the non-motor PD deficits, yet little research has studied how PD affects visual perceptions with no produced motor responses. This provides motivation for the current study which focuses on examining allocentric visual displacement perception -- information used for object identification -- in PD patients. To study this PD participants OFF and ON Levodopa therapy, and age-matched healthy control participants were tested. A modular graphics toolbox was implemented to carry out the perceptual testing. Individuals with PD were shown to have impairments in displacement perception of the larger tested magnitudes when both OFF and ON Levodopa compared to control participants, suggesting impairments in visual displacement processing pathways. These abnormalities could contribute to difficulties some PD patients have with visual recognition and visuospatial navigation. Furthermore, the study validated the graphical tool as a means of quantifying perceptual abilities that can be expanded to many perceptual modalities and paired with robotic devices.

Keywords: Neurological disorders - Mechanisms, Brain functional imaging - Connectivity and information flow, Brain functional imaging - EEG
Abstract: Difficulties in Facial Emotion Recognition (FER) are commonly associated with individuals diagnosed with Autism Spectrum Disorder (ASD). However, the mechanisms underlying these impairments remain inconclusive. While atypical cortical connectivity has been observed in autistic individuals, there is a paucity of investigation during cognitive tasks such as FER. It is possible that atypical cortical connectivity may underlie FER impairments in this population. Electroencephalography (EEG) Imaginary Coherence was examined in 22 autistic adults and 23 typically developing (TD) matched controls during a complex, dynamic FER task. Autistic adults demonstrated reduced coherence between both short and long range inter-hemispheric electrodes. By contrast, short range intra-hemispheric connectivity was increased in frontal and occipital regions during FER. These findings suggest altered network functioning in ASD.

Keywords: Neurological disorders - Diagnostic and evaluation techniques, Brain functional imaging - EEG, Neurological disorders - Epilepsy
Abstract: In hospitals, physicians diagnose brain-related disorders such as epilepsy by analyzing electroencephalograms (EEG). However, manual analysis of EEG data requires highly trained clinicians or neurophysiologists and is a procedure that is known to have relatively low inter-rater agreement (IRA). Moreover, the volume of the data and rate at which new data is acquired makes interpretation a time-consuming, resource hungry, and expensive process. In contrast, automated analysis offers the potential to improve the quality of patient care by shortening the time to diagnosis, reducing manual error, and automatically detecting debilitating events. In this paper, we focus on one of the early decisions made in this process which is identifying whether an EEG session is normal or abnormal. Unlike previous approaches, we do not extract hand-engineered features but employ deep neural networks that automatically learn meaningful representations. We undertake a holistic study by exploring various pre-processing techniques and machine learning algorithms for addressing this problem and compare their performance. We have used the recently released “TUH Abnormal EEG Corpus” dataset for evaluating the performance of these algorithms. We show that modern deep gated recurrent neural networks achieve 3.47% better performance than previously reported results.

Keywords: Neurological disorders - Stroke, Human performance - Activities of daily living
Abstract: We have been developing a facial wearable robot to support the eyelid movements of patients with facial paralysis, especially on one side of the face. This robot has a mechanism for supporting eyelid movements, made from a soft material, which is called the eyelid gating mechanism (ELGM). The ELGM deforms by simple rotational actuation inputs and its deformation is customized to the eyelid movements. Therefore, this robot can provide non-invasive and gentle support for eyelid movements. We herein describe the design rule of the ELGM, and based on this, we conducted a deformation analysis with a non-linear finite element method. We verified the deformation trend from the results, and developed three prototypes based on this trend. Using these prototypes, we conducted a clinical study with facial paralysis patients to evaluate if the ELGM is capable of assisting in closing the eyelid.

Keywords: Cardiovascular regulation - Autonomic nervous system, Cardiovascular regulation - Baroreflex
Abstract: Baroreceptors respond to fluctuations in blood pressure (BP) by modifying physiology in order to maintain a homeostatic set point. Baroreflex sensitivity (BRS) is used to quantify baroreceptor function and is a useful metric for tracking cardiovascular disease state and treatment effects. Pathological conditions such as hypertension (HTN) alter baroreflex function and reduce BRS. Traditionally, the sequence method is used to measure BRS, in which the linear slope of concomitant changes in BP and RR intervals are assessed. However, in rats, a high respiratory rate reduces the reliability of the sequence method. Here, we present a modified sequence method that captures BRS at lower frequencies and decreases the variability of the BRS estimate. This method was demonstrated using ECG and BP data from two groups of HTN rats: Sham rats and rats treated with vagus nerve stimulation. The modified sequence method resulted in lower BRS estimates than the traditional sequence technique when applied to the same data sets. Additionally, the modified sequence method resulted in lower BRS estimate variability.

Keywords: Cardiovascular regulation - Heart rate variability, Cardiovascular regulation - Baroreflex, Cardiovascular regulation - Blood pressure variability
Abstract: Previous research has supported the use of virtual reality (VR) to decrease stress, anxiety, perceptions of pain, and increase positive affect. However, VR’s blood pressure (BP) and autonomic effects in healthy populations have not been explored. This study quantifies the effect of instructed meditation augmented by a virtual environment (VE) on BP, heart rate variability (HRV) and baroreceptor sensitivity (BRS) during rest and following physical (isometric handgrip) or mental (serial sevens subtraction) stress. Sixteen healthy participants underwent all conditions, and those that responded to the stress tests were included in the analysis of stress recovery. Results showed that under resting conditions, VE had no significant effect on BP, HRV or BRS when compared to seated rest and the video on a 2D screen. Following handgrip, VE maintained a higher BRS (82±13 ms/mmHg, mean±standard error), with seated rest showing recovery (56±12 ms/mmHg, p=0.044). Following serial sevens, VE maintained the increased low frequency (LF) power of HRV (66±4 normalized units (n.u.)) compared to seated rest (55±5 n.u., p=0.0060); VE maintained the decreased high frequency (HF) power of HRV (34±4 n.u.) compared to seated rest (44±5 n.u., p=0.014); and VE maintained the increased LF/HF ratio (2.4±0.5) compared to seated rest (1.6±0.3, p=0.012). Hence, after mental stress, VE sustains the increased sympathetic drive and reduced parasympathetic drive. VE may act as a stimulatory driver for autonomic activity and BP. Further studies are required to investigate the effect of different types of VE on BP and autonomic function.

Keywords: Vascular mechanics and hemodynamics - Pulse wave velocity, Cardiovascular and respiratory system modeling - Blood flow models
Abstract: Many studies dealing with blood pressure modeling are evaluated based on a single type of provocation. This paper investigates widely used provocations such as controlled breathing, mental arithmetic and Stroop tests, Valsalva maneuver, cold pressor and muscle tension and combines them in a versatile laboratory protocol. The protocol was tested in an experiment where pulse arrival time (PAT) and heart rate (HR) were measured with chest ECG and finger PPG sensors and blood pressure (BP) with continuous finger-cuff monitor. The experiment results show that mental tasks provoked HR, BP and PAT very little while cold pressor and muscle tension had strong impact in all parameters. Valsalva maneuver had strongest impact in HR and PAT but the effect was transient like. We also predicted systolic BP based on the PAT values. We selected nine points in the protocol to calculate linear prediction model for each subject and then fitted data points to the models. If only the calibration points are taken into account, the correlation between the predicted and measured systolic BP was 0.91. When all the data points are fed into model, correlation was 0.75.

Keywords: Cardiovascular regulation - Baroreflex, Cardiovascular regulation - Heart rate variability
Abstract: Listening to music has been known to affect autonomic function of cardiovascular regulation. Baroreflex is a feedback control loop that uses rate changes of the heart in order to regulate beat by beat changes in blood pressure (BP). In this study, we used two approaches to compute measures of sensitivity of the baroreflex (BRS), a time domain sequence approach and frequency domain transfer functions. Subjects listened to slow and fast tempo songs during the study. Electrocardiogram (ECG) and non-invasive continuous BP were recorded in 14 subjects (7 males and females). From these signals, either beat by beat or equi-sampled in time RR intervals and systolic BP (SBP) were computed. BRS was then estimated using RR and SBP. Our results show that the sequence method consistently provided higher values of BRS than the transfer function method (up to two fold). The two measures were reasonably well correlated (R>0.84) during control and the slow song, but not during the fast song. The BRS was lower (~20%) than control when listening to fast songs (p< 0.005). These results show the effects of listening to songs on BRS changes, but also show that the two methods to estimate BRS, although reasonably correlated, do not always provide similar estimates of BRS.

Keywords: Cardiovascular regulation - Baroreflex, Cardiovascular and respiratory signal processing - Heart Rate and Blood Pressure Variability, Cardiovascular regulation - Blood pressure variability
Abstract: Baroreflex sensitivity (BRS) can be noninvasively assessed from heart period (HP) and arterial pressure (AP) variability series via the estimation of the gain of the transfer function (TF) in the low frequency (LF, 0.04-0.15 Hz) band. However, different strategies can be adopted to pick the value of the TF gain and different fiducial AP values can be considered. In this study we compared different strategies to reduce the TF gain into a unique maker: i) sampling the TF gain in correspondence of the maximum of the HP-AP squared coherence; ii) sampling the TF gain at the weighted average of the central frequencies of AP spectral components; iii) calculating the average of the TF gain in the LF band. Indexes were computed using alternatively systolic AP (SAP) or diastolic AP (DAP) series in combination with HP. Results were obtained in 129 patients undergoing coronary artery bypass graft surgery before (PRE) and after (POST) the induction of general anesthesia with propofol and remifentanil. The reduction of BRS during general anesthesia is expected as a result of overall depression of the cardiovascular control even in this group of pathological subjects already featuring a low BRS before general anesthesia induction. We found that the expected decrease of BRS was observed regardless of the strategy using DAP. Moreover, regardless of series (i.e. SAP or DAP), the sampling of TF gain at the weighted average of the central frequencies of the AP spectral components has the greatest statistical power in distinguishing the two experimental conditions. We recommend the use of this strategy in assessing BRS via TF analysis and a more frequent exploitation of the DAP series.

Keywords: Cardiovascular and respiratory system modeling - Cardiovascular control models, Cardiovascular and respiratory signal processing - Heart Rate and Blood Pressure Variability
Abstract: A life threatening condition in Intensive Care Unit (ICU) is the Acute Hypotensive Episode (AHE). Patients experiencing an AHE may suffer from irreversible organ damage associated with increased mortality. Predicting the onset of AHE could be of pivotal importance to establish appropriate and timely interventions. We propose a method that, using waveforms widely acquired in ICU, like Arterial Blood Pressure (ABP) and Electrocardiogram (ECG), will extract features relative to the cardiac system to predict whether or not a patient will experience a hypotensive episode. Specifically, we want to assess if there are hidden patterns in the dynamics of baroreflex able to improve the prediction of AHEs. We will investigate the predictive power of features related to the baroreflex by performing classifications with and without them. Results are obtained using 17 classifiers belonging to different model families: classification trees, Support Vector Machines (SVMs), K-Nearest Neighbors (KNNs) replicated with different set of hyper-parameters and logistic regression. On average, the use of baroreflex features in the AHE prediction process increases the Area Under the Curve (AUC) by 10%

Keywords: Rehabilitation robotics and biomechanics - Exoskeleton robotics, Therapeutic robotics in rehabilitation, Hardware and control developments in rehabilitation robotics
Abstract: This paper describes the effects of a novel functional electrical stimulation (FES) system which has been integrated in a powered exoskeleton to provide up to 10 channels of stimulation to users with paraplegia via surface electrodes. Experimental data collected from three users with spinal cord injury (SCI) indicate the system reduced the exoskeleton motor torques necessary to perform sit-to-stand transitions in the exoskeleton. All subjects exhibited reduced muscle spasticity immediately after walking in the exoskeleton with FES. Additionally, one subject with stretch-reflex spasms exhibited increased joint excursion and reduced exoskeleton motor torques required to achieve over-ground gait when FES was incorporated.

Keywords: Rehabilitation robotics and biomechanics - Exoskeleton robotics, Therapeutic robotics in rehabilitation, Biomechanics and robotics - Clinical evaluation in rehabilitation and orthopedics
Abstract: The goal of this study was to understand the rehabilitative effects of longitudinal overground exoskeleton training (>100 hours) on gait mechanics, especially foot loading, for gains in walking speed in an individual with chronic motor-incomplete SCI. Biomechanical measures included: normalized plantar loading forces, walking speed and bilateral weight transfer ratio during walking in the EksoGT™ exoskeleton. Longitudinal training with a robotic exoskeleton yielded improvements in clinical outcomes (AIS classification, ISNCSCI motor scores and 10MWT) and provided functional gains in terms of biomechanical outcomes (plantar forces, weight transfer point) to increase overall walking speed.

Keywords: Wearable robotic systems - Orthotics, Robotics - Orthotics, Rehabilitation robotics and biomechanics - Exoskeleton robotics
Abstract: This paper describes a controller for guiding and assisting leg movement during walking with a lower limb exoskeleton with actuated hip and knee joints. The primary novel aspect of the controller is that it employs a virtual flow field to influence movement during swing, rather than a more typical potential-energy-based field. The controller was tested on a single stroke subject. The stroke subject’s leg kinematics demonstrate that the controller is capable of appropriately influencing leg kinematics during overground walking.

Keywords: Rehabilitation robotics and biomechanics - Exoskeleton robotics, Wearable robotic systems - Orthotics, Biomechanics and robotics - Clinical evaluation in rehabilitation and orthopedics
Abstract: Effective solutions for gait rehabilitation in children with cerebral palsy (CP) remain elusive. Wearable robotic exoskeletons offer the potential to greatly increase the dosage and intensity of gait training in this population, which may improve outcomes. We recently reported that a robotic exoskeleton significantly improved knee extension in children with crouch gait from CP. Longitudinal studies are necessary to fully understand long term biomechanical effects of exoskeleton gait training. Given that children's gait can change both as they develop and throughout their therapy, advanced control strategies which can adapt assistance over time may be beneficial. But, stride-to-stride variability makes it difficult to ascertain the effects of exoskeleton assistance and therefore complicates implementation of adaptable control algorithms. Here, we examine the use of the variance ratio (VR), a previously published measure, to assess the effect of exoskeleton assistance on knee extensor and flexor EMG variability in children with CP. Our results show that VR was significantly increased (p < 0.001) compared to baseline during walking with exoskeleton assistance. After five practice sessions, we found that VR was reduced though still greater than baseline levels. Given its sensitivity to exoskeleton assistance and ease of computation, VR may be a useful measure in the future for evaluating stride-to-stride variability in real time to inform algorithmic decision making for autonomous adaptable control.

Keywords: Rehabilitation robotics and biomechanics - Exoskeleton robotics, Modeling and simulation in biomechanics - Orthotics, Wearable robotic systems - Orthotics
Abstract: The goal of this study was to establish stride-parameter gait models correlated to speed on individuals with chronic SCI and able-bodied controls walking with a powered robotic exoskeleton (EksoGT™). Longitudinal exoskeleton training (>100 hours) across eight individuals with SCI resulted in a 30% increase in walking speed. A simple linear regression between step length, stride length for given speed were very tightly correlated along a line of best fit (p < .001). The temporal parameters of stride time, stance time and double support time depicted a non-linear exponentially decaying relationship for given walking speed. The research findings indicate that although longitudinal exoskeleton training reduces the temporal parameters, increases in spatial parameters are only marginal.

Keywords: Rehabilitation robotics and biomechanics - Exoskeleton robotics, Therapeutic robotics in rehabilitation, Joint biomechanics
Abstract: Severe and moderate traumatic brain injury (TBI) causes motor deficits leading to impairments in functional ambulation. Motor recovery involves intensive rehabilitation through physical therapy. Current practices in rehabilitation results in variable recovery of motor function and may result in residual gait deviations. Wearable robotic exoskeletons can provide the user with intensive, goal-directed repetition of movement as well as provide the user with stability and balance during gait, compared to conventional physical therapy. During the acute stage of recovery, the brain is healing and relearning and increased intensive motor rehabilitation throughout this stage could result in improved functional ambulation, especially in individuals with severe impairments who are not independent ambulators. This pilot study evaluates the effect of early intervention robotic exoskeleton gait training on lower extremity biomechanics on a 21 year old young adult with TBI.

Keywords: Ambulatory diagnostic and therapeutic devices - Ambulatory and ADL technologies, Diagnostic devices - Physiological monitoring, Clinical engineering
Abstract: Pathologic movement patterns are characterized by abnormal kinematics that alter how muscles support the body during walking. Individual muscles are often the target of interventions with physical therapy and surgery alike, yet the tools to assess individual muscles clinically remain limited. The aim of this study is to assess OpenSim as a clinical tool for individualized rehabilitative evaluation of children using orthotics. This anatomic and kinematic modeling study was focused on pre- and post-treatment assessment of gait characteristics in fourteen children using orthotic devices. A range of four to twelve acceptable gait capture trials was collected for each child before therapy began and again after four weeks of treatment. The effects of therapy were significant in four of the lower extremity muscle analyses, three of the temporal parameters, and eighteen of the spatial parameters. All muscle lengths showed less deviation from normal values after physical therapy across all subjects. Results of this study support the further evaluation of OpenSim as a tool to improve quantitative assessment of musculoskeletal dynamics during the course of rehabilitative therapy in children using orthotics.

Keywords: Ambulatory Diagnostic devices - Point of care technologies, Diagnostic devices - Physiological monitoring, Health technology - Verification and validation
Abstract: In the last decade, the use of electronic olfaction systems for the early diagnosis of several pathologies by breath analysis has been investigated. In this study, an electronic nose including seven polyaniline sensors has been developed. An impedance measurement circuit and a micro-computer to process the sensor responses were studied to give a pre-diagnosis conclusion. The measurement accuracy is 97% when it is exposed to a simulated human breath and different concentration of ammonia, from 500 ppb to 2.8 ppm. The described prototype weights about 300 g and can be used for 14 hours with a smartphone battery.

Keywords: Clinical laboratory, assay and pathology technologies
Abstract: Epigenetics is a chemical modification to DNA without changes in the base sequence. While it is known that epigenetic modifications have far reaching implications on how genes are expressed, it is difficult to identify what the modification is or where it can be found. A next-generation method of sequencing called nanopore sequencing may be the solution. Nanopore sequencing runs a voltage bias across the DNA sequence and outputs a unique electric response to each genetic unit. Epigenetic modifications may then be identified by their distinct electric response. In this paper we provide preliminary results of applying dynamic time warp Barycenter Averaging (DBA) to multiple noisy nanopore streams to generate a consensus signal that can be used to identify genetic sequences and their modifications. (DBA) convergence rates, time complexity, together with qualitative and quantitative metrics to compare the consensus signal with gold standards are evaluated.

Keywords: Ambulatory diagnostic and therapeutic devices - Ambulatory and ADL technologies, Health technology management and assessment, Ambulatory diagnostic and therapeutic devices - Wireless telemetric systems
Abstract: Stroke is a leading cause of long-term disability that may lead to significant functional motor impairments in the upper limb (UL). Wrist-worn inertial sensors have emerged as an objective, minimally-obtrusive tool to monitor UL motor function in the real-world setting, such that rehabilitation interventions can be individually tailored to maximize functional performance. However, current wearable solutions focus on capturing the quantity of movement without considering the quality of movement. This paper introduces a novel approach to unobtrusively estimate the quality of UL movements in stroke survivors using a single wrist-worn inertial sensor during any type of voluntary UL movements. The proposed method exploits kinematic characteristics of voluntary limb movements that are optimized by the central nervous system during motor control. This work demonstrates that the proposed method could extract clinically important information during random UL movements in 16 stroke survivors, showing a statistically significant correlation to the Functional Ability Scale -- a clinically validated score for movement quality.

Keywords: Ambulatory diagnostic and therapeutic devices - Wireless telemetric systems, Neuromodulation devices, Pacemakers (implantable or external)
Abstract: The purpose of this paper is to develop and validate a configurable system that can wirelessly acquire gastric electrical activity called slow waves, and deliver high energy electrical pulses to modulate its activity. The system is composed of a front-end unit, and an external stationary back-end unit that is connected to a computer. The front-end unit contains a recording module with four channels, and a stimulating module with two channels. Commercial off-the-shelf components were used to develop front- and back-end units. A graphical user interface (GUI) was designed in LabVIEW to process and display the recorded data in real-time, and store the data for off-line analysis. Besides, the gain of the analog conditioning circuit as well as the stimulation pulse configuration is programmable directly through the GUI. The system was successfully validated on bench top. The bench-top studies showed an appropriate frequency response for analog conditioning and digitization resolution to acquire gastric slow waves. Moreover, the system was able to deliver electrical pulses at amplitudes up to ±24 mA and ±12 mA to a load of up to 0.5 kΩ and 1 kΩ, respectively. This study reports the first high-energy stimulator that can be controlled wirelessly and integrated into a gastric bioelectrical activity monitoring system. The system can be used for treating functional gastrointestinal disorders.

Keywords: Ambulatory Diagnostic devices - Point of care technologies, Clinical laboratory, assay and pathology technologies, Diagnostic devices - Physiological monitoring
Abstract: Nanopore-based diagnostic systems are a promising tool for counting viruses in a specimen one by one. However, despite intensive R&D efforts, it remains difficult to recognize virus subtypes by nanopore devices. We thus propose a novel diagnostic system that combines a specialized virus recognition procedure with a nanopore detection procedure. This recognition procedure consists of three steps: 1) capture target viruses using specific probes for recognition; 2) release captured targets; and 3) detect released targets by nanopore. Proof-of-concept tests are conducted using avidin-modified fluorescent particles (as a model for viruses) and biotin-modified alkane thiol (as a model for probes). The avidin-modified particles are confirmed to be captured on electrode by biotin-modified probes and then, the particles are electrochemically released from the electrode. Consequently, the released particles are successfully detected by nanopore devices. Furthermore, the concept is also proved by using human influenza viruses (H1N1, A/PR/8/34) and sugar chain (6’-sialyllactose)-modified probes. This suggests that our concept is applicable to various infectious diseases by changing probes (ligands).

Keywords: Optical imaging and microscopy - Multi photon imaging, Optical imaging and microscopy - Microscopy, Optical imaging and microscopy - Fluorescence microscopy
Abstract: We provide a practical demonstration that non-degenerate 2-photon excitation offers up to 2-fold increase in excitation efficiency compared to the degenerate case and an additional gain in fluorescence due to the nature of multicolor absorption.

Keywords: Optical imaging and microscopy - Neuroimaging, Optical imaging and microscopy - Fluorescence microscopy, Optical imaging and microscopy - Multi photon imaging
Abstract: Swept confocally aligned planar excitation (SCAPE) microscopy is a high-speed 3D imaging method that combines light sheet excitation with a scanning-descanning strategy that enables imaging through a single, stationary objective lens. SCAPE permits high-speed 3D imaging of neural activity in a wide range of behaving organisms from flies to mice.

Keywords: Optical imaging and microscopy - Optical vascular imaging, Optical imaging and microscopy - Multi photon imaging, Optical imaging and microscopy - Microscopy
Abstract: Many optical techniques have been developed for high resolution, three-dimensional imaging of cerebral hemodynamics and neural activity. Laser speckle contrast imaging has become one of the most widely used techniques due to its simple instrumentation and its ability to visualize blood flow over a wide range of spatial scales. However, obtaining quantitative blood flow information remains a challenge for laser speckle imaging. Recently, an extension to laser speckle imaging, called Multi-Exposure Speckle Imaging (MESI), was introduced that increases the quantitative accuracy of CBF images. This talk will describe technical developments in laser speckle imaging as well as new methods for three-dimensional visualization of blood vessels that can be used to improve our understanding of blood flow measures inferred from speckle images.

Keywords: Optical imaging, Multimodal imaging, Brain imaging and image analysis
Abstract: Clinicians encounter significant challenges when treating acute brain injury, including focal or global stroke, hemorrhagic stroke, trauma, and many other disorders. Advancements in treatments have been partly limited by lack of proper monitoring of the brain that can offer accurate feedback to clinicians on the state of a brain. Optical approaches now offer immense hope that can transform how clinicians can diagnose and treat brain injured patients. Using optical methods in a preclinical model of acute brain injury, we are able to capture diagnostic and prognostic information.

Keywords: Health Informatics - Mobile health, General and theoretical informatics - Algorithms
Abstract: This paper presents the integration of three major modules of the signal processing pipeline that go into a typical digital hearing aid as a real-time smartphone app. These modules include voice activity detection, noise reduction, and compression. The steps taken to allow the real-time implementation of this integration or signal processing pipeline are discussed. These steps can be utilized to create similar signal processing pipelines or integrated apps to evaluate hearing improvement algorithms. The real-time characteristics of the developed integrated app are reported as well as an objective evaluation of its noise reduction.

Keywords: General and theoretical informatics - Algorithms, General and theoretical informatics - Machine learning, Health Informatics - Mobile health
Abstract: In this paper we show early evidence of the feasibility of detecting labour during pregnancy, non-invasively and in free-living. In particular, we present machine learning models aiming at dealing with the challenges of unsupervised, free-living data collection, such as identifying periods of high quality data and detecting physiological changes as labour approaches. During a first phase, physiological data including electrohysterography (EHG, the electrical activity of the uterus), heart rate (HR) and gestational age (GA) were collected in laboratory conditions for model development. In particular, data were collected 1) during simulated activities of daily living, aiming at eliciting artifacts and developing diagnostic models for free-living data 2) during pregnancy, including labour, aiming at developing labour probability models from clean, supervised physiological recordings. Machine learning models using datasets 1) and 2) were deployed in free-living, longitudinally, in 142 pregnant women, between week 22 of pregnancy and delivery. A total of 1014 hours of data and an average of 7 hours per person were collected. Output of the developed models was analyzed to determine the feasibility of detecting labour non-invasively using physiological data, acquired with a single sensor placed on the abdomen. Results showed that the probability of being in labour for recordings collected during the last 24 hours of pregnancy was consistently higher than the probability during any other pregnancy week. Thus, non-invasive labour detection from physiological data seems promising.

Keywords: General and theoretical informatics - Algorithms, Health Informatics - Mobile health, General and theoretical informatics - Machine learning
Abstract: In this work, we use data acquired longitudinally, in free-living, to provide accurate estimates of running performance. In particular, we used the HRV4Training app and integrated APIs (e.g. Strava and TrainingPeaks) to acquire different sets of parameters, either via user input, morning measurements of resting physiology, or running workouts to estimate running 10 km running time. Our unique dataset comprises data on 2113 individuals, from world class triathletes to individuals just getting started with running, and it spans over 2 years. Analyzed predictors of running performance include anthropometrics, resting heart rate (HR) and heart rate variability (HRV), training physiology (heart rate during exercise), training volume, training patterns (training intensity distribution over multiple workouts, or training polarization) and previous performance. We build multiple linear regression models and highlight the relative impact of different predictors as well as trade-offs between the amount of data required for features extraction and the models accuracy in estimating running performance (10 km time). Cross-validated root mean square error (RMSE) for 10 km running time estimation was 2:6 minutes (4% mean average error, MAE, 0.87 R2), an improvement of 58% with respect to estimation models using anthropometrics data only as predictors. Finally, we provide insights on the relationship between training and performance, including further evidence of the importance of training volume and a polarized training approach to improve performance.

Keywords: Bioinformatics - Bioinformatics for health monitoring, General and theoretical informatics - Algorithms
Abstract: Many algorithms have been used to estimate respiratory rate (RR) from Photoplethysmography (PPG) recently. However, the accuracy and time consumption are still a challenging issue. In this paper, we propose a novel algorithm for RR estimation using Joint Sparse Signal Reconstruction (JSSR) based on Regularized Sparsity Adaptive Matching Pursuit (RSAMP) in a real-time fashion. The algorithm has been tested on Capnobase dataset and the results showed that the mean absolute error (MAE) and root mean squared error between estimates and references are 1.09 breaths per minute (bpm) and 2.44 bpm, respectively.

Keywords: General and theoretical informatics - Algorithms, General and theoretical informatics - Statistical data analysis, General and theoretical informatics - Computational disease profiling
Abstract: Functional near-infrared spectroscopy (fNIRS) is an emerging non-invasive functional brain imaging technique, through detecting the changes of hemoglobin concentrations to investigate brain activities in various tasks. The aim of this study is to investigate the complexity of near-infrared spectroscopy signals during resting state and upper limb movements. Experimental study was designed by applying NIRS to collect the data especially for both healthy subjects and traumatic brain injury (TBI) patients. The modified multiscale entropy (MMSE) algorithm was employed to assess the complexity of fNIRS signals which may reflect the changes of brain activity when people underwent brain injury. The results that the mean MMSE of oxyhemoglobin values was lower in TBI patients compared to healthy subjects, indicated that MMSE was feasible to measure complexity of cerebral near-infrared spectroscopy signals in TBI patients, and that brain injury was associated with the decreased complexity of cerebrovascular reactivity. Moreover, measurement of complexity of brain signals has potential to provide significant guidance for rehabilitation.

Keywords: General and theoretical informatics - Algorithms, General and theoretical informatics - Machine learning, Sensor Informatics - Wearable systems and sensors
Abstract: Several studies have been proposed to estimate blood pressure (BP) with cuffless devices using only a Photoplethysmograph (PPG) sensor on the basis of the physiological knowledge that the PPG changes depend on the state of the cardiovascular system. In these studies, machine learning algorithms were used to extract various features from the wave height and the elapsed time from the rising point of the pulse wave to feature points have been used using to estimate the BP. However, the accuracy is still not adequate to be used as medical equipment because their features cannot express fully information of the pulse waveform which changes according to the BP. And, no other effective knowledge about the pulse waveform for estimating BP has been found yet. Therefore, in this study, we focus on the autoencoder which can extract complex features and can add new features of the pulse waveform for estimating the BP. By using autoencoder, we extracted 100 features from the coupling signal of the pulse wave and from its first-order differentiation and second-order differentiation. The result of examination with 1363 test subjects show that the correlation coefficients and the standard deviation of the difference between the measured BP and the estimated BP got improved from R = 0.67, SD = 13.97 without autoencoder to R = 0.78, SD = 11.86 with autoencoder.

Keywords: Wearable sensor systems - User centered design and applications, Optical and photonic sensors and systems, Integrated sensor systems
Abstract: Sleep monitoring provides valuable insights into the general health of an individual and helps in the diagnostic of sleep-derived illnesses. Polysomnography, is considered the gold standard for such task. However, it is very unwieldy and therefore not suitable for long-term analysis. Here, we present a non-intrusive wearable system that, by using photoplethysmography, it can estimate beat-to-beat intervals, pulse rate, and breathing rate reliably during the night. The performance of the proposed approach was evaluated empirically in the Department of Psychology at the University of Fribourg. Each participant was wearing two smart-bracelets from Ava as well as a complete polysomnographic setup as reference. The resulting mean absolute errors are 17.4ms (MAPE 1.8%) for the beat-to-beat intervals, 0.13 beats-per-minute (MAPE 0.20%) for the pulse rate, and 0.9 breaths-per-minute (MAPE 6.7%) for the breath rate.

Keywords: Smart textiles and clothings, Textile-electronic integration, Sensor systems and Instrumentation
Abstract: Wearable systems are gaining broad acceptance for monitoring physiological parameters in several medical applications. Among a number of approaches, smart textiles have attracted interest because they are comfortable and do not impair patients’ movements. In this article, we aim at developing a smart textile for respiratory monitoring based on a piezoresistive sensing element. Firstly, the calibration curve of the system and its hysteresis have been investigated. Then, the proposed system has been assessed on 6 healthy subjects. The volunteers were invited to wear the system to monitor their breathing rate. The results of the calibration show a good mean sensitivity (i.e., approximately 0.11V·% ^-1 ); although the hysteresis is not negligible, the system is able to follow the cycles also at high rates (up to 36 cycle·min ^-1 ). The feasibility assessment on 6 volunteers (two trials for each one) shows that the proposed system can estimate with good accuracy the breathing rate. Indeed, the results obtained by the proposed system were compared with the ones collected with a spirometer, used as reference. Considering all the experiments, a mean percentage error was approximately 2%. In conclusion, the proposed system has several valuable features (e.g., the sensing element is lightweight, the sensitivity is high, and it is possible to develop comfortable smart textile); in addition, the promising performances considering both metrological properties and assessment on volunteers foster future tests focused on: i) the possibility of developing and system embedding several sensing elements, and ii) to develop a wireless acquisition system, to allow comfortable and long-term acquisition in both patients and during sport activities.

Keywords: Smart textiles and clothings, Wearable wireless sensors, motes and systems, Modeling and analysis
Abstract: — The aim of this present research was to determine whether the cardiovascular function has been affected by wearing compression garments during the recovery phase. Fourteen subjects (men, n=7; women, n=7; 24.7±4.5 years, 166.0±7.6 cm; 60.9±12.0 kg) completed a running protocol on a treadmill. Each subject participated in two running experiments, using either compression garments (CGs) or non- compression garments (NCGs) during exercise and 2 hours recovering time. Electrocardiogram (ECG) signals were collected during 2 hours recovery using wearable sensors. The present work indicated a statistically significant difference between CGs and NCGs from 90 minutes onwards (p<0.05). ECG parameters showed some significant difference in heart rate (HR), ST and corrected QT (QTc) (p<0.05). Therefore, the cardiovascular function was positively influenced by the application of CGs during the recovery phase.

Keywords: Smart textiles and clothings, Portable miniaturized systems, Integrated sensor systems
Abstract: In surgery, orthopedic cast is often implemented to stabilize and fix anatomical structures like broken bones. Plaster could harden after mixed with water, thus it is commonly utilized with cotton bandage to form a solid structure to encase a limb or other body parts. As plaster is heavy and impervious, cast could easily result in itching, rashes, allergic contact dermatitis or other cutaneous complications. In this paper we present a novel implementation for surgical fixation with low melting point alloy (LMPA) stuffed in silicone tubes, which is dubbed “LMPA enhanced bandage”. The alloy is heated by an enameled copper wire to alter the stiffness. When the alloy is in solid state, the bandage could withstand high load without significant deformation, while if heated to its melting point, the entire bandage would soften. We present several conceptual experiments to evaluate the mechanical performance and body fixation of the proposed bandage. Phase change process and temperature variation were recorded by an infrared camera. Preliminary results showed that the present fixation bandage design owns sufficient mechanical strength and necessary thermal response performance to meet the requirement of clinical applications.

Keywords: Portable miniaturized systems, Wearable low power, wireless sensing methods, Integrated sensor systems
Abstract: Down syndrome is one of the health disorders that interferes with regular and healthy sleep. Most children with Down syndrome are referred to a sleep clinic for the assessment of the severity of their apnea. Regular polysomnography based assessment of apnea has been challenging with this sensitive patient population. We present our efforts towards developing a flexible adhesive bandage sized near infrared spectroscopy system (pediBand) for home-assessment of apnea in children with Down syndrome. Combined with inertial measurement units, pediBand record heart rate, heart rate variability, respiratory rate, arterial oxygen saturation and cerebral oxygen saturation. These are the essential parameters to assess sleep apnea and could also potentially be used in the assessment of sleep performance in general. A modified version of pediBand system was evaluated on adult patients and successfully demonstrated the changes in hemodynamic system triggered by sleep apnea.

Keywords: Sensor systems and Instrumentation, Portable miniaturized systems, Wearable low power, wireless sensing methods
Abstract: Habitual snoring has been known to increase the risk for serious health problems in addition to affecting the quality of others’ sleep. Several recent consumer products aim to automatically detect snoring events and wake the snorer to elicit a posture change. In this paper, we present a study comparing two of the methods, electromyography vs. accelerometry, proposed for automated snoring detection and incorporation of these into a wearable system. The study includes (a) the testing of various sensor configurations and placements to obtain optimal electromyography and accelerometry signals, (b) a review of the accuracy of a variety of snore detection algorithms from previously attained biological signals, and (3) design of an embedded device with integrated sensors and haptic feedback capability. Our preliminary results indicate superiority of accelerometry over electromyography. Further research opportunities to prove the concept and improve the design are then detailed for future work.

Keywords: Chemo/bio-sensing - Chemical sensors and systems, Bio-electric sensors - Sensing methods, Chemo/bio-sensing - Biological sensors and systems
Abstract: Hydrogen peroxide (H2O2) plays a critical role in the regulation of multifarious physiological processes. We developed a sensor containing a mercaptopropionic acid (MPA) monolayer covalently immobilized with Horseradish peroxidase (HRP) enzyme for the electrochemical detection of hydrogen peroxide (H2O2). A gold foil substrate was chemically treated with nitric acid and were used as working electrode. Platinum wire and Ag-Ag/Cl were used as counter and reference electrodes, respectively. The acid treated gold electrode with the immobilized enzyme shown to have improved catalytic activity in the reduction of H2O2. The steady-state current response increases linearly with H2O2 concentration from 10 μM to 9 mM with a low detection limit of 60 μM and showed a sensitivity of 0.4 mA/ mM cm2. This electrochemical sensor is demonstrated to be highly selective and sensitive in the presence of interfering analytes. The improved activity and simple preparation method of the electrode makes the MPA-HRP modified gold electrode promising for being developed as an attractive robust material for electrochemical H2O2 sensing.

Keywords: Optical and photonic sensors and systems, Portable miniaturized systems, Integrated sensor systems
Abstract: A platform coupled with an integrating sphere for portable fluorescence detection was presented in this paper. The detector under testing has a demonstrated lower limit of detection 0.4 nM for detecting fluorescein solutions and 0.00128 ng/L for detecting SYBR-Green stained dsDNA in this preliminary work. The signal-to-noise ratio analysis suggests that the limit of detection could be even lower than presented herein.

Keywords: Chemo/bio-sensing - Biological sensors and systems, Chemo/bio-sensing - Techniques, New sensing techniques
Abstract: Escherichia coli detects and follows chemical gradients in its environment in a process known as chemotaxis. The performance of chemotaxis approaches fundamental biosensor speed and sensitivity limits, but there have been relatively few attempts to incorporate the response into a functional biosensor. Toward that end, we have developed software to process microscopy of a large number of tethered E. coli responding to different chemical perturbations. Upwards of fifty cells can be recorded in one experiment, allowing for rapid labeling of responses. After we collected hundreds of wild-type chemotactic E. coli motor responses to dilutions of aspartate and leucine, we trained a support vector classifier (SVC) to estimate the order of magnitude of aspartate concentration between 0 molar, 100 nanomolar, and 1 micromolar, with a single cell classification subset accuracy of 69.2%. We trained another SVC to differentiate between aspartate and leucine with a single cell classification subset accuracy of 83.3%. Using a majority-vote method on a bacterial population of size N, estimates have 95% confidence for N = 27 bacteria for concentration detection and N = 9 bacteria for chemical differentiation. These methods are a step towards adaptable chemotaxis-based biosensing.

Keywords: Chemo/bio-sensing - Biological sensors and systems, Mechanical sensors and systems
Abstract: Accurate detection of neuropeptides in cerebrospinal fluid (CSF) plays an important role in both in-depth studies and early diagnosis of neurological diseases. Here, we report a biosensor based on Capacitive Micromachined Ultrasonic Transducer (CMUT) which is capable of detecting low concentrations (pg ~ ng/ml) of a neuropeptide involved with the progression of Alzheimer’s diseases, somatostatin (SST). A 10-MHz CMUT was fabricated and utilized as a physical resonant sensor which detects the change in the concentration of analyte through the mass-loading mechanism. The resonant plate was sequentially coated with protein G and antibodies to provide specificity to SST; Cysteine-tagged protein G layer enables controlled immobilization of antibodies in a well-oriented manner. The change in the resonant frequency of the CMUT sensor was measured after incubating the sensor in various concentrations of SST. The significant shifts in the resonant frequency were observed for SST concentrations in the range of 10 pg/ml ~ 1 ng/ml. Compared to the previously reported biosensors developed for SST detection, our sensor shows discernable responses for SST that are ~6 orders of magnitude lower in concentration. Thus, this work demonstrates the potential of the CMUT resonant sensor as a promising biosensor platform for detection of neuropeptides involved with neurodegenerative diseases that often exist in low concentrations in CSF.

Keywords: Chemo/bio-sensing - Micrototal analysis and lab-on-chip systems, Bio-electric sensors - Sensing methods
Abstract: Follicle-stimulating hormone (FSH) is an important indicator of ovarian reserve function in women in clinical testing. In this work, a label-free paper-based immunosensor was developed for electrochemical rapid detection of FSH. A hydrophilic channel surrounded with hydrophobic barriers was firstly fabricated on the chromatography paper by wax printing technology. Then three electrodes were screen-printed on the circle zones of the channel, in which one carbon electrode further modified by reduced graphene-oxide /thionine /gold nanoparticles nanocomposites and FSH monoclonal antibody was used as the working electrode to provide sensitivity of the immunosensor. The detection of FSH is based on the decreased electrochemical current of Thi produced from the specific binding of the FSH and anti-FSH, and the decrease of the current is proportional to the concentration of the FSH. The experimental results exhibited that the immunosensor could be used to detect the standard FSH in range of 1-100 mIU/mL with the detection limit of 1 mIU/mL. And the proposed immunosensor had been successfully applied to detect FSH in serum samples.

Keywords: Modeling and analysis, Novel methods
Abstract: Biomarkers in urine samples are widely used in clinical diagnosis. Involving image processing and data analysis, urinalysis is very popular in hospitals because of its convenience and speediness; and the most important reason is its high accuracy rating. This paper presents colorimetric recognition for urine test device with different algorithms aiming to find a good-performance classifier. Those algorithms can train a set of data and get a model to discriminate the test data. Almost the accuracy of each classifier is beyond 92%, even 99%. Although the classifier that has highest average accurate rate of recognition is K-Nearest Neighbor, we cannot overlook the performance of Support Vector Machine, which perform best in protein test. In order to compare these eight algorithms, we use Python simulation to validate the results and show the accuracy of each classifier.

Keywords: Health Informatics - Disease profiling and personalized treatment, Health Informatics - eHealth, Health Informatics - Mobile health
Abstract: The remote monitoring of patients is based on digital systems that enable the remote collection, usually at home, of health data and its transmission to health centers. The telemedicine paradigm is of particular interest in chronic diseases, fragile population and elderly monitoring. Parkinson’s disease (PD) is a neurodegenerative disorder having high impact on the lives of patients and their families. Such a disease impacts on the physical and psychological abilities of the patient and may have an effect on the relationship among family members. The strict monitoring of PD patients and their caregivers is of paramount importance in the implementation of prompt actions counteracting the worsening of the disease or that of the caring process. In this paper we present a mobile App developed for PD patients and their caregiver. The App aims at improving the communication among the patient/caregiver and the specialists, covering aspects related to both the disease symptoms and the caring process. In the paper we describe the App along with results collected during a one year experimentation on a cohort of 10 patients and 7 caregivers. The results show that the approach is accepted by patients and caregivers. Furthermore, obtained results demonstrate that the monitoring system is effective in the identification of dangerous conditions for the patient and useful in the implementation of reactive health management strategies.

Keywords: Health Informatics - Mobile and wearable technologies for elderly, Health Informatics - Informatics for chronic disease management, Health Informatics - Mobile health
Abstract: Parkinson’s disease (PD) is a degenerative and long-term disorder of the central nervous system, which often causes motor symptoms, e.g. tremor, rigidity, and slowness. Currently, the diagnosis of PD is based on patient history and clinical examination. Technology-derived decision support systems utilizing, for example, sensor-rich smartphones can facilitate more accurate PD diagnosis. These technologies could provide less obtrusive and more comfortable remote symptom monitoring. The recent studies showed that motor symptoms of PD can reliably be detected from data gathered via smartphones. The current study utilized an open-access dataset named ”mPower” to assess the feasibility of discriminating PD from non-PD by analyzing a single self-administered 20-step walking test. From this dataset, 1237 subjects (616 had PD) who were age and gender matched were selected and classified into PD and non-PD categories. Linear acceleration (ACC) and gyroscope (GYRO) were recorded by built-in sensors of smartphones. Walking bouts were extracted by thresholding signal magnitude area of the ACC signals. Features were computed from both ACC and GYRO signals and fed into a random forest classifier of size 128 trees. The classifier was evaluated deploying 100-fold cross-validation and provided an accumulated accuracy rate of 0.7 after 10k validations. The results show that PD and non-PD patients can be separated based on a single short-lasting self-administered walking test gathered by smartphones’ built-in inertial measurement units.

Keywords: Health Informatics - Behavioral health informatics, Health Informatics - Mobile health, Sensor Informatics - Behavioral informatics
Abstract: Task and activity tracking has been an effective industrial management and research technique for generations. It is applied to workflow optimization, group coordination, task sequencing, individual time management and environmental exposures. Appropriately, task tracking technologies are migrating to personal mobile devices. At the same time, individual survey approaches have been advanced tremendously as mobile apps. We report on a method of dynamic task registration with momentary assessment systems in natural environments that apply knowledge of context. We describe how the app was refined by a user acceptance study and its deployment in studies on agricultural exposure and industrial operations.

Keywords: Health Informatics - Mobile health, Health Informatics - Information technologies for healthcare delivery and management, Health Informatics - Health data acquisition, transmission, management and visualization
Abstract: This research develops a novel dynamic mobile health (mHealth) application (app), called the Clinical Event Annotator (CEA). The CEA comprises of a native Android tablet app and an administrative web app. The native app is used at the patient bedside to manually annotate clinical events in real-time. Event types include patient monitor alarms, routine care, clinical interventions, and patient movements. The app can be dynamically updated with user-defined customized events. The web app generates reports of the annotation sessions. The CEA app is developed to support a clinical study that explores the use of pressure-sensitive mats (PSM) in the neonatal intensive care unit (NICU) to detect the respiratory rate (RR), heart rate (HR), and movement of critically ill neonatal patients. High-fidelity CEA app annotations are synced with a backend database that enables integration and synchronization with independently acquired patient monitoring data, such as RR, HR, and contact pressure data from the PSM. The gold standard CEA annotations serve the purpose of retrospectively training machine learning algorithms for clinical event detection. Preliminary test results from use of the app in the clinical study are presented. Development of the CEA app is a unique and novel contribution that addresses the well-known problem of manually annotating physiologic data streams to support clinical data mining applications.

Keywords: Health Informatics - Personal health systems, Health Informatics - Mobile health, Health Informatics - Decision support methods and systems
Abstract: Congestive heart failure (CHF) occurs when the heart cannot provide the necessary cardiac output for the metabolic needs of the human body. The most prominent symptoms are increased venous pressure, abnormal heart and breathing rate, tiredness and leg swelling. Most important pathogenesis influence are: age, gender, high blood pressure, alcohol and smoking, sedentary lifestyle and diet, genetic predisposition and family history, diabetes, and atherosclerosis. Common causes are considered to be valvular heart disease, coronary heart disease and hypertension. CHF diagnosis can be achieved through physical examination (i.e. blood pressure, body mass index, blood tests) and echocardiography. In this work, we present a smart mobile application and internet of things capable of the early detection and real-time monitoring of CHF exacerbations, enforcing prevention on a daily basis. We refer to the architectural elements of our approach accounting for the integration of a secure access scheme, following GDPR regulation, as a novel biometric solution to increase security and access. A first validation of the system is also presented.

Keywords: Health Informatics - Healthcare modeling and simulation, Health Informatics - Informatics for chronic disease management, Health Informatics - Mobile health
Abstract: Asthma and Chronic Obstructive Pulmonary Disease are chronic and long-term lung diseases. Disease monitoring with minimal sensors with high efficacy can make the disease control simple and practical for patients. We propose a model for the severity assessment of the diseases through wearables and compatible with mobile health applications, using only heart rate and SpO2 (from pulse oximeter sensor). Patient data were obtained from the MIMIC-III Waveform Database Matched Subset. The dataset consists of 158 subjects. Both heart rate and SpO2 signal of patients are analyzed via the proposed algorithm to classify the severity of the diseases. Strategically, a rule-based threshold approach in real time evaluation is considered for the categorization scheme. Furthermore, a method is proposed to assess severity as an Event of Interest (EOI) from the computed metrics in retrospective. This type of autonomous system for real-time evaluation of patient’s condition has the potential to improve individual health through continual monitoring and self-management, as well as improve the health status of the overall Smart and Connected Community (SCC).

Keywords: Nonlinear dynamic analysis - Phase locking estimation, Physiological systems modeling - Closed loop systems, Time-frequency and time-scale analysis - Nonstationary processing
Abstract: Neural oscillations enable communication between brain regions. Closed-loop brain stimulation attempts to modify this activity by stimulation locked to the phase of concurrent neural oscillations. If successful, this may be a major step forward for clinical brain stimulation therapies. The challenge for effective phase-locked systems is accurately calculating the phase of a source oscillation in real time. The basic operations of filtering the source signal to a frequency band of interest and extracting its phase cannot be performed in real time without distortion. We present a method for continuously estimating phase that reduces this distortion by using an autoregressive model to predict the future of a filtered signal before passing it though the Hilbert transform. This method outperforms published approaches on real data and is available as a reusable open-source module. We also examine the challenge of compensating for the filter phase response and outline promising directions of future study.

Keywords: Neural stimulation (including deep brain stimulation)
Abstract: Objective: To investigate mechanical and functional failure of diffuse axonal injury in nerve bundles following frontal head impacts, by finite element simulations. Methods: Simulating frontal head impact by using a Head Model. Then, investigation of the changes in induced-electro-mechanical responses at the cellular level is carried out in two scaled nerve bundle models, made of myelinated or unmyelinated nerve fibres. The study is carried in finite element analysis with Abaqus CAE 6.13-3. Conclusion: The myelin layer protects the fibre from mechanical damage, preserving its functionalities.

Keywords: Neural stimulation (including deep brain stimulation)
Abstract: Abstract— Objective: Brain plasticity changes in space because of human adaptation to space environment. A multi-scale analysis approach is used to simulate the structural and functional cellular changes, following frontal head impact at Earth-gravity level, hyper-gravity and microgravity. Methods: The study is carried in finite element analysis with Abaqus CAE 6.13-3 by using a Head Model and a Nerve Bundle Model. Conclusion: Severe level of axonal injury is reached during high-speed impact only. Myelinated fibre show higher functional and structural resistance for the loading conditions considered.

Keywords: Computer modeling for treatment planning
Abstract: Electrical properties tomography (EPT) has been studied in order to non-invasively map the conductivity and permittivity of brain tissues. Various approaches have been investigated for predicting the EP at either very low or high Lamour frequencies. An example of the latter is water-based EPT (wEPT) which derives the EP maps from the image ratio of two T1-weighted images with different repetition times. The objective of this study is to examine if wEPT can be adapted to accurately create conductivity maps of the brain between 100 and 1000 kHz, and if EP of pathological tissues can be estimated with the same approach. Experimental measurements and wEPT imaging estimations were performed in animal brain samples and tumor-bearing rats.

Keywords: Medical devices interfacing with the brain or nerves, Computer modeling for treatment planning, Diagnostic devices - Physiological monitoring
Abstract: Invasive electrophysiological measurement of brain activity is commonly employed during epilepsy surgery to provide final validation of required resection regions. These data are critical to clinical decision making, but manual expert analysis of these data can be complicated; The epileptic onset region must be inferred from data collected at multiple electrodes. Source analysis can improve the analysis of these data, but requires accurate models of bioelectric signal conduction. Given the proximity of the measurement locations to the generating cortical sources, modeling of electrode-tissue interactions is particularly important for invasive measurements. Here, we evaluate the effect of a finite difference complete electrode model on the accuracy of leadfield computations for invasive electrocorticography. Our results show that in the vicinity of electrode locations, use of the simpler point electrode model produces large (> 10%) topographic and magnitude differences. Additionally, while differences reduce with increasing distance from electrode locations, they remain large enough to likely have a measurable effect on localization accuracy.

Keywords: Medical devices interfacing with the brain or nerves, Diagnostic devices - Physiological monitoring
Abstract: Machine Learning is becoming increasingly important in interpreting biological signals. In this work, we examine the potential for classification in brain haemorrhage detection. Numerical head and brain models with and without haemorrhagic lesions are designed. Impedance measurements from an electrode array positioned on the exterior of the head are used to train and test linear support vector machine (SVM) classifiers. The results show that this emerging measurement technique may have promise for detection and diagnosis of brain haemorrhage when coupled with such classifiers.

Keywords: Neural interfaces - Implantable systems, Neural interfaces - Body interfaces, Brain-computer/machine interface
Abstract: Implantable electronic packages for neural implants utilize reliable electrical feedthroughs that connect the inside of a sealed capsule to the components that are exposed to the surrounding body tissue. With the ongoing miniaturization of implants requiring ever higher integration densities of such feedthroughs new technologies have to be investigated. The presented work investigates the sealing of vertical feedthroughs in aluminum-oxide-substrates with gold stud-bumps. The technology enables integration densities of up to 1600/cm² while delivering suitable water leak rates for realistic implantation durations of miniaturized packages (feedthrough-count > 50, package-volume < 2 cm³) of more than 50 years. All manufacturing steps require temperatures below 420 °C and are suitable for maskless rapid prototyping.

Keywords: Neural interfaces - Implantable systems, Motor neuroprostheses - Epidural stimulation, Neural stimulation
Abstract: This paper presents the design and fabrication of an implantable control unit intended for epidural spinal cord stimulation (ESCS) in rats. The device offers full programmability over stimulation parameters and delivers a constant current to an electrode array to be located within the spinal canal. It implements an adaptive voltage compliance in order to reduce the unnecessary power dissipation often experienced in current-controlled stimulation (CCS) devices. The compliance is provided by an adjustable boost converter that offers a voltage output in the range of 6.24 V to 28 V, allowing the device to deliver currents up to 1 mA through loads up to 25 kΩ. The system has been fabricated using discrete components, paving the way to an inexpensive product that can easily be manufactured and batch produced. The control unit occupies a total volume of ~13.5 cm^3 and therefore fulfills the size restrictions of a system to be implanted in a rat. Results indicate that by adjusting the voltage compliance a total power efficiency up to 35.5% can be achieved, saving around 60 mW when using lower stimulation currents or operating on smaller impedances. The achieved efficiency is the highest compared to similar state-of-the-art systems.

Keywords: Neural interfaces - Implantable systems, Neural interfaces - Microelectrode technology, Brain-computer/machine interface
Abstract: This work presents reliability investigations of silicone gasket as solid underfill for interconnection interfaces in hybrid implant systems with high channel count flexible electrode arrays and hermetically packed electronics. The gasket is fabricated by laser structuring thin sheet of silicone rubber. The surface activation of silicone sheet ensures mechanical bonds with the mating surfaces thereby improving the mechanical stability of the assembly and the insulation of the interconnects. The gasket samples with 10 × 10 openings for interconnect pads, each with diameter of 270 µm and a center to center pitch size of 490 µm, were sandwiched between a polyimide array and a metallized ceramic substrate. The gasket maintained high insulation impedance of 15 ± 0.30 MΩ between the adjacent interconnects with markedly capacitive behavior (phase angle, -89 °) after 17 weeks in soaked conditions under accelerated aging at 60 °C. The gasket also survived electrical stresses and sustained high impedance (10.93 MΩ with phase angle of -88 °) when subjected to constant 3 VDC for 100 days.

Keywords: Neural interfaces - Implantable systems, Neural signal processing, Neural stimulation
Abstract: Neural interfaces that can both stimulate and record from the peripheral nervous system are an important component of future bioelectronic devices. However, despite a long history of neurostimulation, there has been relatively little success in the design of a chronically implantable device for recording from peripheral nerves. This fundamental road block must be overcome if the design of advanced implantable devices is to continue. In this paper, we demonstrate the effectiveness of one method: velocity selective recording, a method that has been proposed as a tool for online neural recording that does not require training. We present results and analysis from in-vivo recordings made on the right vagus nerve of pig using a multiple-electrode cuff as a chronically implantable recording array.

Keywords: Neural interfaces - Implantable systems
Abstract: Impedance measurement using Electrochemical Impedance Spectroscopy is a widely utilized technique in neural electrodes. Research and clinical devices that incorporate stimulating and recording microelectrodes routinely characterize the material’s integrity and its functionality through impedance measurement. Nominal impedance values ensure a stable neural electrode-tissue contact capable of passing through power efficient electric signals with desired signal-to-noise ratio or effective volume coverage. However, the complexity of the in vivo environment limits the usage of the three-electrode setup, which has been accepted as the ideal method in providing a stable impedance measurement. Impedance data measured from microelectrodes in three-electrode and two-electrode setups show that the two setups have similar outcomes in terms of the impedance modulus over a 0.5 Hz-100 kHz frequency range. Usage of a platinum counter electrode lowered the overall variance in impedance readings compared to the stainless steel counter electrode. However, correlation coefficient values (>0.97) between three-electrode and two-electrode setups show that impedance values seldom deviate due to changes in electrode setup. Based on the results of this study, the usage of the two-electrode setup in vivo allowed acceptable electrochemical impedance spectroscopy accuracy, and the utilization of a platinum counter electrode is recommended to reduce measurement variance.

Keywords: Neural interfaces - Implantable systems, Brain-computer/machine interface, Neural stimulation
Abstract: The number of implantable bidirectional neural interfaces available for neuroscientific research applications is still limited, despite the rapidly increasing number of customized components. We previously reported on how to translate available components into "ready-to-use" wireless implantable systems utilizing components off-the-shelf (COTS). The aim of the present study was to verify the viability of a micro-electrocorticographic (muECoG) device built by this approach. Functionality for both neural recording and stimulation was evaluated in an ovine animal model using acoustic stimuli and cortical electrical stimulation, respectively. We show that auditory evoked responses were reliably recorded in both time and frequency domain and present data that demonstrates the cortical electrical stimulation functionality. The successful recording of neuronal activity suggests that the device can compete with existing implantable systems as a neurotechnological research tool.

Keywords: Nonlinear dynamic analysis - Biomedical signals, Physiological systems modeling - Signal processing in physiological systems, Physiological systems modeling - Signals and systems
Abstract: Abstract—The aim of this study was to investigate the effect of head-up tilt (HUT) test on male and female young patients, diagnosed with orthostatic intolerance (OI), in comparison to male and female healthy subjects. Twenty seven OI patients (21 women, 6 men) and 26 age-matched healthy subjects (13 women, 13 men) were enrolled in a 70° HUT test. In addition to hemodynamic variables, cardiovascular and respiratory parameters were determined using linear and nonlinear methods to analyze heart rate (HRV) and blood pressure variability (BPV). During the complete test, HRV was lower in healthy men than in female controls. Decreased HRV and increased BPV were observed in female patients compared to healthy women. Furthermore, systolic BPV was increased in male and female patients. However, linear (rmssd) and nonlinear (plvar2) parameters indicated that diastolic BPV decreased in male patients during orthostatic phase, but remained unchanged in female patients. Findings indicated gender dependent mechanisms for the regulation of diastolic blood pressure during orthostatic stress in patients.

Keywords: Nonlinear dynamic analysis - Biomedical signals, Coupling and synchronization - Nonlinear coupling, Physiological systems modeling - Signal processing in physiological systems
Abstract: Recent studies have shown that occasional short term coupling between fetal and maternal cardiac systems occurs. Fetal magnetocardiography (fMCG) is a non-invasive technique that records the magnetic fields associated with the electrical activity of the fetal heart through sensors placed over the maternal abdomen. The fMCG allows accurate estimation of fetal heart rates (fHR) due to its high signal-to-noise ratio (SNR) and temporal resolution. In this study, we analyzed coupling between fHR and maternal heart rates (mHR) using Transfer Entropy (TE). TE determines coupling between two variables by quantifying the information transferred between them in both directions. In this work, we used 74 fMCG recordings to compute TE in both directions over 1-minute disjoint time windows (TW). We examined the effect of fetal movement (FM) as a factor of influence on the TE analysis. We identified 21 subjects with FM during the recording and separated them into two gestational age (GA) groups (GA1<32 and GA2≥32 weeks). Next, TE values were compared between TWs containing non- FM with TWs containing FM using Wilcoxon Signed-Rank test. In addition, we compared TE calculations for non-FM segments obtained from the 74 subjects using Rank-Sum test in the two GA groups. Our results showed that TE values from TWs containing FM are not significantly different than those computed for TWs of non-FM. In both directions, we found that TE values obtained from the 74 subjects did not show any significant difference between GA1 and GA2 which is consistent with previous studies. Our study suggests that FM does not affect the TE computations.

Keywords: Nonlinear dynamic analysis - Biomedical signals, Adaptive filtering, Time-frequency and time-scale analysis - Time-frequency analysis
Abstract: This study proposes a new method for motion artifact (MA) removal in Photoplethysmography (PPG) signal that combines accurate heart rate (HR) frequency estimation and notch filtering. The method applies LMS-Newton adaptive filtering to reduce motion artifact noise and uses a novel HR correction stage for accurate HR frequency estimation. Notch filters are used to recover clean PPG signal from HR frequency and second harmonic frequency of PPG. On a widely used dataset of 12 recordings, our method achieves an averaged HR error of 0.92bpm and a Pearson correlation of 0.997. Experimental results further show that our method can recover the PPG waveform with clear dicrotic waves even for strongly MA-corrupted PPG signal.

Keywords: Nonlinear dynamic analysis - Biomedical signals, Nonlinear dynamic analysis - Deterministic chaos
Abstract: Human photoplethysmogram (PPG) is one of the signals widely applied for health monitoring. Development of the new techniques made possible evolution of traditional contact PPG which was measured at red and near-infrared light (NIR) to the contactless, imaging PPG (iPPG) that can be recorded at various light wavelengths, including ambient visible light. However, despite the numerous advantages of iPPG its applications demonstrated so far are quite limited. The NIR PPG was previously found to be useful for various applications in the area of physiological and mental health monitoring by utilizing advanced methods of nonlinear time-series analysis applied on its reconstructed dynamics. The main purpose of this study is to demonstrate data-driven approach with time-delay-reconstructed attractor obtained from the iPPG. The results of this study demonstrated that the iPPG dynamics can be reconstructed with fine data resolution, and its time-delay-reconstructed trajectory is almost deterministic, though contains noise. The obtained results might be useful for further applied studies on the iPPG.

Keywords: Nonlinear dynamic analysis - Biomedical signals, Physiological systems modeling - Signal processing in physiological systems, Physiological systems modeling - Signals and systems
Abstract: Dynamic changes in autonomic stress responses may provide details on autonomic nervous system functions. Time-varying evaluation can be achieved with a sliding window, however, in order to learn dynamic changes, an evaluation method needs to not only conduct calculation with a short sliding step but also derive evaluation indices with a narrow window. Stress analysis using HRV data shorter than one minute is still a challenge in this field. This paper investigates a Poincare plot analysis method for stress evaluation based on short term heart rate variability (HRV) data. First a sliding window, with no overlap, is used to segment data in order to form Poincare plots. Then a simple index, which corresponds to mean distance between two adjacent points in the plot, is calculated on each evaluation window. The window length is defined with time duration and four lengths are examined in this paper, namely, 15, 30, 45, and 60 s. Two mental stress induction experiments, mental arithmetic and Stroop color-word tests, are utilized to validate the proposed method.

Keywords: Image analysis and classification - Machine learning / Deep learning approaches, Image classification, Image feature extraction
Abstract: Chest x-ray (CXR) analysis is a common part of the protocol for confirming active pulmonary Tuberculosis (TB). However, many TB endemic regions are severely resource constrained in radiological services impairing timely detection and treatment. Computer-aided diagnosis (CADx) tools can supplement decision-making while simultaneously addressing the gap in expert radiological interpretation during mobile field screening. These tools use hand-engineered and/or convolutional neural networks (CNN) computed image features. CNN, a class of deep learning (DL) models, has gained research prominence in visual recognition. It has been shown that Ensemble learning has an inherent advantage of constructing non-linear decision making functions and improve visual recognition. We create a stacking of classifiers with hand-engineered and CNN features toward improving TB detection in CXRs. The results obtained are highly promising and superior to the state-of-the-art.

Keywords: Magnetic resonance imaging - Other organs, Image feature extraction, Image classification
Abstract: Quantitative models are essential in precision medicine that can be used to predict health status and prevent disease and disability. Current radiomics models for clinical outcome prediction often depend on huge amount of image features and may include redundant information and ignore individual feature importance. In this work, we propose a prognostic discrimination ranking strategy to select the most relevant image features for image assisted clinical outcome prediction. Firstly, a redundancy and prognostic discrimination evaluation method is proposed to evaluate and rank a large number of features extracted from images. Secondly, forward sequential feature selection is performed to select the top ranked relevant features in each discriminate quantization. Finally, representative vectors are generated by the fusion of pivotal clinical parameters and selected image signatures to be fed into a classification model. The proposed model was trained and tested over 70 patient studies with six MR sequences and four clinical parameters from ISLES challenges. The evaluations using ROC curves demonstrated the improved performance over five other feature selection models where the proposed model achieved AUCs of 0.821, 0.968, 0.983, 0.896 and 1 when predicting five clinical outcome scores respectively.

Keywords: CT imaging, Image analysis and classification - Machine learning / Deep learning approaches, Image registration, segmentation, compression and visualization - Machine learning / Deep learning approaches
Abstract: Early detection of lung nodules is of great importance in lung cancer screening. Existing research recognizes the critical role played by CAD systems in early detection and diagnosis of lung nodules. However, many CAD systems, which are used as cancer detection tools, produce a lot of false positives (FP) and require a further FP reduction step. Furthermore, guidelines for early diagnosis and treatment of lung cancer are consist of different shape and volume measurements of abnormalities. Segmentation is at the heart of our understanding of nodules morphology making it a major area of interest within the field of computer aided diagnosis systems. This study set out to test the hypothesis that joint learning of false positive (FP) nodule reduction and nodule segmentation can improve the computer aided diagnosis (CAD) systems’ performance on both tasks. To support this hypothesis we propose a 3D deep multi-task CNN to tackle these two problems jointly. We tested our system on LUNA16 dataset and achieved an average dice similarity coefficient (DSC) of 91% as segmentation accuracy and a score of nearly 92% for FP reduction. As a proof of our hypothesis, we showed improvements of segmentation and FP reduction tasks over two baselines. Our results support that joint training of these two tasks through a multi-task learning approach improves system performance on both. We also showed that a semi-supervised approach can be used to overcome the limitation of lack of labeled data for the 3D segmentation task.

Keywords: Image feature extraction, Image classification
Abstract: The precise identification of loss of consciousness (LOC) is key to studying the effects of anesthetic drugs in neural systems. The standard behavioral assay for identifying LOC in rodents is the Loss of Righting Reflex (LORR), assessed by placing the animal in the supine position every minute until it fails to right itself. However, this assay cannot be used when the rodents are head-fixed, which limits the use of powerful techniques such as multi-electrode recordings, textit{in-vivo} patch clamp, and neuronal imaging. In these situations, an alternative way to assess LOC is needed. We propose that loss of movement (LOM) in whiskers and paws of head-fixed animals can be used as an alternative behavioral assay in head-fixed animals. Unlike LORR, LOM in whiskers and paws is much harder to detect by visual inspection. Therefore, we developed a method to automatically assess for LOM of whiskers and paws in head fixed rodents during textit{in vivo} patch clamp recordings. Our method uses an algorithm based on optical flow and point-process filtering which can be run on images acquired on regular cameras at low frame-rates. We show that the algorithm can achieve at least comparable accuracy in detecting LOC when compared with consensus among human observers, as well as improved precision when compared with individual observers. In the future, we aim to to expand the method to detect more behavioral end-points during anesthesia such as paradoxical excitation. Eventually, we hope to enable multi-modal anesthesia studies, which incorporates behavioral and neurophysiological data.

Keywords: Image analysis and classification - Machine learning / Deep learning approaches, Image feature extraction, Image classification
Abstract: Non-invasive optical imaging techniques have been recently proposed for distinguishing between different types of tissue in burns generated in porcine models. These techniques are designed to assist surgeons during the process of burn debridement, to identify regions requiring excision and their appropriate excision depth. This paper presents a machine learning tool for discriminating between Viable and Non-Viable tissues in human injuries. This tool merges a supervised (QDA) with an unsupervised (k-means clustering) classification algorithms. This combination improves the Non-Viable tissue detection in 23.7% with respect to a simple QDA classifier.

Keywords: Implantable systems
Abstract: A 3.3 V CMOS bandgap reference (BGR) was presented in this study that utilizes MOS transistors operating in the sub- threshold region. The complexity of the circuit and the dependency of the voltage reference on power supply variations are simultaneously decreased through the use of a new compensation circuit technique. The proposed BGR is simulated using a 0.35 µm CMOS standard process. Consequently, a 5.53 ppm/°C temperature coefficient is obtained in the -40~+125 °C temperature range, the maximum power supply rejection ratio is -62 dB, and a 2.033 mV/V voltage line regulation is achieved for the 2.3~4.3 V supply voltage. The proposed circuit dissipates a supply current of 8.89 µA at a 3.3 V supply voltage, and the active area is 112 µm × 60 µm.

Keywords: Implantable technologies, Implantable sensors - biocompatibility, Portable miniaturized systems
Abstract: Bioelectronic medicine requires miniaturized implants to selectively interface small target structures in the autonomous nervous system. Long-term stable non-hermetic packaging techniques have to be developed for smallest implantable electronics and interfaces. A process for the fabrication of chip-in-foil implants is proposed that combines a flip-chip approach for bare die embedding with a silicone rubber backbone. The conducting tracks are structured on polyimide (PI), enabling the use of microsystems fabrication technologies. The long-term stability of the interface between PI and silicone rubber is investigated by peel tests in phosphate buffered saline after prolonged soaking at 37 °C. With a peel force of 721 mN after 14 days of soaking, the combination of 10-nm-thick titanium oxide and the adhesion promoter Dow Corning 1200 OS leads to the highest interface stability of the tested methods. This conforms to the results of atomic force microscopy measurements, where this treatment increased the surface roughness from 0.44 nm to 46.45 nm. The devised interface enables the construction of a chip-in-foil system with silicone rubber for height levelling in combination with polyimide-based micro structuring.

Keywords: Implantable technologies, Implantable sensors - biocompatibility
Abstract: Neuromodulation devices have been approved for the treatment of epilepsy and seizures, with many other applications currently under research investigations. These devices rely on implanted battery powered pulse generators, that require replacement over time. Miniaturized ultrasound powered implantable devices have the potential can potentially eliminate the need for battery in neuromodulation devices. While these devices have been assessed in vitro, long-term in vivo assessment is required to determine device safety and performance. In this study, we developed a multi-stage long-term test platform to assess the performance of miniaturized ultrasound powered implantable devices.

Keywords: Physiological monitoring - Novel methods, Implantable sensors, New sensing techniques
Abstract: Real-time monitoring of arteriovenous graft blood flow would provide early warning of graft failure to permit interventions such as angioplasty or graft replacement to avoid catastrophic failure. We have developed a new type of flexible pulsation sensor (FPS) consisting of a 3D printed elastic cuff wrapped around a graft and thus not in contact with blood. The FPS uses multi-walled carbon nanotubes (MWCNTs) dispersed in polydimethylsiloxane (PDMS) as a piezoresistive sensor layer, which is embedded within structural thixotropic PDMS. These materials were specifically developed to enable sensor additive manufacturing via 3D Bio-plotting, and the resulting strain sensor is more compliant and has a wider maximum strain range than graft materials. Here, we analyze the strain transduction mechanics on a vascular graft and describe the memristive properties of MWCNT-PDMS composites, which may be mitigated using AC biasing. In vitro testing of the FPS on a vascular graft phantom showed a robust, linear sensor output to pulsatile flows (170-650 mL/min) and pressures (62-175 mmHg). The FPS showed an RMS error when measuring pressure and flow of 7.7 mmHg and 29.3 mL/min, with a mean measurement error of 6.5% (pressure) and 8.0% (flow).

Keywords: Implantable systems, Integrated sensor systems, Wearable body sensor networks and telemetric systems
Abstract: This paper presents the design of an 800 MHz VCO for both free-running and injection locked operation in a novel low power transmitter with wireless frequency reference, operating in the MICS band (402-405 MHz). The transmitter employs simultaneous tuning and locking, to set the desired channel with a minimal injected power. The VCO is designed and fabricated in a 0.13 µm SiGe BiCMOS process and has a core area of 0.5 mm². The measurement of the free-running VCO shows -107 dBc/Hz phase noise at 300 kHz frequency offset. If locked to an external frequency reference the VCO shows -118 dBc/Hz phase noise at 300 kHz offset, while consuming 3 mA from a 1.2 V supply (3.6 mW). When the VCO is tuned during the locking, -20 dBm of reference power is required to enable operation in the whole MICS band. The measured phase noise of the free-running VCO ensures reliable calibration of the proposed transmitter and the locked VCO satisfies all requirements of an implantable device using MICS band data transmission. Therefore, this VCO presents a key building block of an injection locked, frequency agile, implantable transmitter for the MICS band.

Keywords: Bio-electric sensors - Sensor systems, Implantable sensors, Novel methods
Abstract: New research and diagnosis tools are needed to continuously measure bowel state and activity. We investigated functionality of several sensors in vivo and in vitro. Five sensor types, including pressure, infrared, color, conductivity and capacitance, were tested to validate functionality inside the colon. Initial wired prototypes were tested and calibrated in benchtop testing and then inserted intraluminally into pig colon and rectum in three acute surgical procedures. The results from both benchtop and in-vivo testing correlate and indicate that pressure, conductivity, and capacitance measurements could provide information on the state of the bowel and its activity.

Keywords: Optical imaging, Optical imaging and microscopy - Optical vascular imaging, Functional image analysis
Abstract: Induction of thermal damage to tissue through delivery of microwave energy is frequently applied in surgery to destroy diseased tissue such as cancer cells. Minimization of unwanted harm to healthy tissue is still achieved subjectively, and the surgeon has few tools at their disposal to monitor the spread of the induced damage. This work describes the use of optical methods to monitor the time course of changes to the tissue during delivery of microwave energy in the porcine liver. Multispectral imaging and diffuse reflectance spectroscopy are used to monitor temporal changes in optical properties in parallel with thermal imaging. The results demonstrate the ability to monitor the spatial extent of thermal damage on a whole organ, including possible secondary effects due to vascular damage. Future applications of this type of imaging may see the multispectral data used as a feedback mechanism to avoid collateral damage to critical healthy structures and to potentially verify sufficient application of energy to the diseased tissue.

Keywords: Optical imaging and microscopy - Multi photon imaging, Optical imaging and microscopy - Fluorescence microscopy, Optical imaging
Abstract: Optimal cancer therapy requires targeted and individualized treatment of all tumor cells, including both gross and microscopic disease. Intraoperatively hard to visualize and often left behind, microscopic foci of residual cancer cells significantly increase the risk of cancer recurrence and treatment failure rates. Fluorescently-tagged targeted molecular labels are employed to guide surgery, but conventional fluorescent intraoperative imagers suffer from lack of sensitivity and maneuverability, limiting practicality in small tumor cavities owing to their cumbersome sizes driven by optics. This work does away with conventional lenses and filters and introduces an optics-free molecular imaging “skin” consisting of only a 25μm thin CMOS contact imager that synergistically integrates the long emission lifetimes of upconverting nanoparticles(UCNP) combined with upconversion to use a time domain approach to acquire the image coupled with infrared illumination allowing deep tissue penetration and elimination of autofluorescence. Using this strategy, we are able to visualize UCNPs at fluences (W/cm²) compatible with intraoperative use, opening the door to visualize targeted areas with microscopic sensitivity and facilitate residual microscopic disease detection during surgery, and laying the groundwork for precision post-operative radiation.

Keywords: Optical imaging and microscopy - Fluorescence microscopy, Image analysis and classification - Machine learning / Deep learning approaches, Optical imaging
Abstract: We have performed an in vivo pilot study, in which multispectral endogenous or autofluorescence lifetime imaging (FLIM) was performed on clinically suspicious oral lesions of 73 patients undergoing tissue biopsy for oral dysplasia and cancer diagnosis. The results from this pilot study indicated that mild-dysplasia and early stage oral cancer could be detected from benign lesions using a computed aided diagnosis system developed based on biochemical and metabolic biomarkers derived from the endogenous FLIM images. The diagnostic performance of this novel FLIM clinical tool was estimated using a leave-one-patient-out cross-validation approach, showing levels of sensitivity >90%, specificity >85%, and Area Under the Receiving Operating Curve (ROC- AUC) >0.9.

Keywords: Optical imaging and microscopy - Multi photon imaging
Abstract: The size and curvature of the macaque brain present challenges for two photon laser scanning microscopy (2P-LSM). General access to the cortex requires 5-axis positioning over a range of motion wider than existing designs offer. In addition, movement artifacts due to physiological pulsations and bodily movement present particular challenges. We present a microscope and implant platform that allows for repeatable, motorized positioning and stable imaging at any point on the dorsal convexity of macaque cortex. While testing the system to image neurons expressing GCaMP6f in an awake macaque, motion artifacts were limited to several microns.

Keywords: Optical imaging and microscopy - Multi photon imaging, Brain imaging and image analysis, Novel imaging modalities
Abstract: Determining how a neuron computes requires an understanding of the complex spatiotemporal relationship between its input (e.g. synaptic input as a result of external stimuli) and action potential output. Recent advances in in vivo, laser-scanning multiphoton technology, known as random-access microscopy (RAM), can capture this relationship by imaging fluorescent light, emitted from calcium-sensitive biosensors responding to synaptic and action potential firing in a neuron’s full dendritic arbor and cell body. Ideally, a continuous output of fluorescent intensities from the neuron would be converted to a binary output (‘event’, ‘or no-event’). These binary events can be used to correlate temporal and spatial associations between the input and output. However, neurons contain hundreds-to-thousands of synapses on the dendritic arbors generating an enormous quantity of data composed of physiological signals, which vary greatly in shape and size. Thus, automating data-processing tasks is essential to support high-throughput analysis for real-time/post-processing operations and to improve operators’ comprehension of the data used to decipher neuron computations. Here, we describe an automated software algorithm to detect brain neuron events in real-time using an acousto-optic, multiphoton, laser scanning RAM developed in our laboratory. The fluorescent light intensities, from a genetically encoded, calcium biosensor (GCAMP6m), are measured by our RAM system and are input to our ‘event-detector’, which converts them to a binary output meant for real-time applications. We evaluate three algorithms for this purpose: exponentially weighted moving average, cumulative sum, and template matching; present each algorithm’s performance; and discuss user-feasibility of each. We validated our system in vivo, using the visual circuit of the Xenopus laevis.

Keywords: Optical imaging
Abstract: Remote heart rate (HR) estimation from videos is useful because it facilitates monitoring ongoing health conditions without sensors that are often uncomfortable to wear. In the HR estimation from videos, choice of the image region, at which the HR is calculated, is critically important as it greatly affects the estimation accuracy. In this paper, a novel algorithm for HR estimation that uses dynamic region selection is proposed. The image regions that clearly contain pulse waveforms are quickly found by a region selector using a machine learning technique. In addition, the proposed method enhances the robustness of tracking the temporal change of the HR by using a particle filter. The experimental results show that the proposed method achieves the absolute average error less than 1.1BPM (Beats Per Minute) with the processing time less than 0.6 s for a single HR estimation.

Keywords: Human performance - Ergonomics and human factors, Neuromuscular systems - Learning and adaption, Human performance - Activities of daily living
Abstract: Upper-limb rehabilitation training for hemiplegic patients has been primarily conducted by human therapists, and, hence, their use of training methods and conditions strongly depends on their expertise. The force control and motion sensing functions of rehabilitation robots are expected to be used for the qualitative training/assessment in the next-generation computerized rehabilitation. In this study, we developed a desktop rehabilitation robot for upper limbs (D-SEMUL). In addition, we also assessed the usability of its user interface and the affinity (acceptance) of the training program with a questionnaire for elderly hemiplegic/non-hemiplegic participants (nine hemiplegic, five males and four females and seven non-hemiplegic, two males and five females). The results indicated that the touchscreen is acceptable for the user interface, and the background music used significantly affects the affinity of the program.

Keywords: Human performance - Fatigue, Human performance - Ergonomics and human factors, Human performance - Modelling and prediction
Abstract: — The specific purpose of this present paper was to investigate whether the EEG activity has been affected by wearing whole body compression garments during a running test. Ten subjects (men, n=5; women, n=5; age: 24.11 ± 4.48 years; height: 163.56 ± 7.70 cm; chest: 87.78 ± 6.92 cm; weight: 58.67 ± 10.96 kg; BMI: 21.77 ± 2.63 kg.m-2) completed a running protocol on a treadmill. Each subject participated in two running trials, wearing either a compression garment (CG) or a non-compression garment (NCG) during exercise and 2 hours recovering time. Electroencephalogram (EEG) signals were collected during exercise and 2 hours recovery using wearable sensors. The present study revealed a statistically significant difference between CGs and NCGs in alpha, beta and theta power spectral density (p<0.05). EEG parameters showed some significant difference in alpha and beta (p<0.05). Therefore, the EEG activity was influenced by the application of CGs during the running test.

Keywords: Human performance - Drowsiness and microsleeps, Brain functional imaging - EEG, Human performance - Driving
Abstract: Microsleeps are brief and involuntary instances of complete loss of sleep-related consciousness. We present a novel approach of creating overlapping clusters of subjects and training of an ensemble classifier to enhance the prediction of microsleep states from EEG. Overlapping clusters are created using Kullback-Leibler divergence between responsive state features of each pair of training subjects. Highly correlated features within each overlapping cluster are discarded. The remaining features are selected via Fisher score based ranking followed by an average of 5-fold cross-validation areas under the curves of receiver operating characteristics (AUCROC) of a linear discriminant analysis (LDA) classifier. The decisions of LDA classifiers on overlapping clusters are fused using weighted average. We evaluated this new approach on 16- channel EEG data from 8 subjects who had performed a 1-D visuomotor task for two 1-h sessions. Joint entropy features were extracted from a 5-s window of EEG with steps of 0.25 s. Test performances were evaluated using leave-one-subject-out cross-validation. Our ensemble of overlapping clusters of subjects achieved a mean prediction performance, phi, of 0.42 compared with 0.39 for a single LDA classifier and 0.37 for generalized stacking.

Keywords: Human performance - Activities of daily living, Neurological disorders, Neurorehabilitation
Abstract: This pilot study investigated the impact of freezing of gait, objectively measured with three inertial sensors, on mobility function during seven days of community-living monitoring in people with Parkinson’s disease. Twenty-four subjects with PD, of which 14 experiencing freezing of gait, were recruited in this study. Subjects wore three inertial sensors (Opals, APDM) attached to both feet and the lumbar region for a week of continuous monitoring. Walking bouts, of at least 10s, were first identified, and then features of freezing, quantity and quality of mobility were extracted and averaged across the seven days. Results showed significant impairments in freezing and quality of mobility in the freezers group compared to the non-freezers. Our measures of average and variability of time spent freezing was associated to the subjects’ perception of freezing, assessed with the New Freezing of Gait Questionnaire. These preliminary results are introducing promising measures of mobility impairments measured during community-living in PD.

Keywords: Human performance, Human performance - Sensory-motor, Motor learning, neural control, and neuromuscular systems
Abstract: During locomotion, energy flow through the legs is governed by the mechanical impedance of each joint. These mechanical properties, including stiffness and damping, have recently been quantified at the ankle joint. However, the relevance of these properties in human sensorimotor control is unclear. An important aspect of sensorimotor control is the ability to sense small changes in stimuli. Thus, we investigated the human ability to detect small changes in the stiffness and damping components of leg joint impedance when interacting with a mechanical system coupled to the ankle or knee. The perception threshold was determined via a psychophysical paradigm that required subjects to compare the mechanical impedance of virtual spring-mass-damper systems. Subjects reliably detected impedance changes of 11% and 12% at the ankle and knee, respectively. Additionally, the perception of stiffness and damping were comparable, indicating that the biomechanical relevance of the stiffness and damping components of impedance may be similar. Finally, these results offer novel insight into the design and control of impedance-based technologies, such as prostheses and exoskeletons.

Keywords: Physiological systems modeling - Signal processing in physiological systems, Data mining and processing in biosignals
Abstract: Neural processing of sensory stimuli can be studied using EEG by estimation of the evoked potential using the averages of large sets of trials. However, it is not always possible to include all stimulus parameters in a conventional analysis, since this would lead to an insufficient amount of trials to obtain the evoked potential by averaging. Linear mixed models use dependencies within the data to combine information from all data for the estimation of the evoked potential. In this work, it is shown that in multi-stimulus EEG data the quality of an evoked potential estimate can be improved by using a linear mixed model. Furthermore, the linear mixed model effectively deals with correlation between parameters in the data and reveals the influence of individual stimulus parameters.

Keywords: Signal pattern classification, Time-frequency and time-scale analysis - Empirical mode decomposition in biosignal analysis
Abstract: Electrocorticogram (ECoG) has been used as a reliable modality to control a brain machine interface (BMI). Recently, promising results of high-density ECoG have shown that non redundant information can be recorded with finer spatial resolution from the cortical surface. In this study, high-density ECoG was recorded intraoperatively from two patients during awake brain surgery while performing instructed hand flexion and extension. Event related desynchronization (ERD) were found in the low frequency band (LFB: 8-32 Hz) band while event related synchronization (ERS) were found in the high frequency band (HFB: 60-200 Hz). The classification between hand flexion and extension was performed by using common spatial pattern (CSP) as a feature extraction technique and linear discriminant analysis (LDA) as a classifier. In order to compare the high-density ECoG and normal ECoG in terms of classifying between hand flexion and extension, we simulated a typical clinical ECoG (8 mm spacing) by averaging the neural activity of nearest four channels. The same classification methods were applied on the averaged recordings. In HFB, the classification error rate using simulated ECoG greatly increased and lagged the movement onset compared to the original high- density ECoG. In LFB, the differences between them were not prominent. These results indicated that high-density ECoG is able to capture non-redundant task-related information from the motor cortex and potentially serves as a better modality to drive a neural prosthetic compared to typical clinical electrodes.

Keywords: Physiological systems modeling - Signal processing in simulation, Physiological systems modeling - Signals and systems, Nonlinear dynamic analysis - Biomedical signals
Abstract: Retinal prostheses work by delivering electrical pulses to the surviving retinal neurons. A pattern of electrical stimulation can generate a perception of vision in blind patients. To improve efficacy of retinal implants, it is important to understand how different classes of retinal neurons respond to electrical stimulation and if a classification can be made based on the electrophysiological properties of neurons. We use previously recorded patch clamp data from retinal ganglion cells classified into morphological classes (A, B, C, D) and functional types (ON, OFF, ON-OFF). We use a machine learning technique to separate data based on the recorded electrophysiological parameters. Results show that the clusters discovered using the machine learning technique do not correspond to the morphological or functional classes used by neuroscientists.

Keywords: Physiological systems modeling - Signal processing in physiological systems, Nonlinear dynamic analysis - Biomedical signals
Abstract: The Electroencephalogram (EEG) can be considered as the output of a nonlinear system whose dynamics is significantly affected by motor tasks. Nevertheless, computational approaches derived from the complex system theory has not been fully exploited for characterising motor imagery tasks. To this extent, in this study we investigated EEG complexity changes throughout the following categories of imaginary motor tasks of the upper limb: transitive (actions involving an object), intransitive (meaningful gestures that do not include the use of objects), and tool-mediated (actions using an object to interact with another one). EEG irregularity was quantified following the definition of Fuzzy Entropy, which has been demonstrated to be a reliable quantifier of system complexity with low dependence on data length. Experimental results from paired statistical analyses revealed minor topographical changes between EEG complexity associated with transitive and tool-mediated tasks, whereas major significant differences were shown between the intransitive actions vs. the others. Our results suggest that EEG complexity level during motor imagery tasks of the upper limb are strongly biased by the presence of an object.

Keywords: Nonlinear dynamic analysis - Biomedical signals, Nonlinear dynamic analysis - Deterministic chaos
Abstract: In this paper, we investigate the hippocampal and cortical sleep EEG of adult rats at different sleep stages by employing Lyapunov exponent and third-order cumulant measures to quantify and compare the chaotic and nonlinear behavior of EEG obtained during vigilance states of quiet-waking, slow-wave sleep, and rapid eye movement (REM) sleep. Lyapunov exponent was computed to characterize the EEG for chaos and third-order cumulant was used to measure the deviations from Gaussianity of the signal. Our results show positive Lyapunov exponents for all EEG states indicating a low-dimensional chaos for both REM and non-REM system. Furthermore, REM sleep EEG exhibits the largest Lyapunov exponent in both hippocampal and cortical EEG amongst other vigilance states. We also identified non-zero third-order cumulant for all the vigilance states which suggests their non-Gaussian behavior.

Keywords: Physiological systems modeling - Signals and systems, Physiological systems modeling - Signal processing in physiological systems
Abstract: Sympathovagal balance, an autonomic index resulting from the sympathetic and parasympathetic influences on cardiovascular control, has been extensively used in the research practice. The current assessment is based on analyzing Heart Rate Variability (HRV) series in the frequency domain by regarding the ratio between the low and high frequency components (LF/HF). Nevertheless, LF and HF powers are known to be both influenced by vagal activity which strongly bias the accuracy of this method. To this extent, in this study we combine time-varying estimates from electrodermal activity (EDA) and HRV to propose a novel index of sympathovagal balance. Particularly, sympathetic activity is estimated from the EDA power calculated within the 0.045-0.25Hz bandwidth (EDASymp), whereas parasympathetic dynamics is measured instantaneously through a point-process modeling framework devised for heartbeat dynamics (HFpp). We test our new index SV = EDASymp/HFpp on data gathered from 22 healthy subjects (7 females and 15 males) undergoing a 3 minutes gold-standard protocol for sympathetic elicitation as the cold-pressor test (CPT). Results show that the activation of the proposed sympathovagal tone is consistent with CPT elicitation and is associated with a significantly higher statistical discriminant power than the standard LF/HF ratio, also revealing different dynamics between female and male subjects.

Keywords: Neural stimulation, Brain functional imaging - Segmentation
Abstract: Research in the area of transcranial electrical stimulation (TES) often relies on computational models of current flow in the brain. Models are built on magnetic resonance images (MRI) of the human head to capture detailed individual anatomy. To simulate current flow, MRIs have to be segmented, virtual electrodes have to be placed on the scalp, the volume is tessellated into a mesh, and the finite element model is solved numerically to estimate the current flow. Various software tools are available for each step, as well as processing pipelines that connect these tools for automated or semi-automated processing. The goal of the present tool -- ROAST -- is to provide an end-to-end pipeline that can automatically process individual heads with realistic volumetric anatomy leveraging open-source software (SPM8, iso2mesh and getDP) and custom scripts to improve segmentation and execute electrode placement. When we compare the results on a standard head with other major commercial software for finite element modeling (ScanIP, Abaqus), ROAST only leads to a small difference of 9% in the estimated electric field in the brain. We obtain a larger difference of 47% when comparing results with SimNIBS, an automated pipeline that is based on surface segmentation of the head. We release ROAST as an open-source, fully-automated pipeline at https://www.parralab.org/roast/.

Keywords: Neural stimulation, Neurological disorders - Stroke, Neural signals - Coding
Abstract: Intracortical microstimulation can be successfully used to manipulate neuronal activity and connectivity, thus representing a potentially powerful tool to steer neuroplasticity occurring after brain injury. Activity-dependent stimulation (ADS), in which the spikes recorded from a single neuron are used to trigger stimulation at another cortical location, is able to potentiate cortical connections between distant brain areas. Here, we developed an experimental procedure and a computational pipeline aimed at investigating the ability of ADS to induce changes in intra-cortical activity of healthy anesthetized rats.

Keywords: Neural stimulation, Neurorehabilitation, Neurological disorders - Stroke
Abstract: Paired associative stimulation (PAS) has been shown to increase corticospinal excitability (CSE) providing a promising adjuvant therapeutic approach for stroke. Combining PAS with movement of the stimulated limb may further increase enhancement of CSE, however, individuals with moderate to severe stroke may not be able to engage in the necessary repetitive voluntary movements of the paretic limb. The objective of this study was to investigate the feasibility of contralaterally coordinated PAS (ccPAS) applied to the resting hand extensors during fast extension of the contralateral hand. A potential dependency of CSE modulation on the phase of the movement of the opposite hand was evaluated. Eleven participants each completed three session: PAS applied to the resting right hand during the preparation phase of the extension of the contralateral (left) hand; PAS applied during the execution phase of the left hand extension; and PAS applied with both hands at rest. Motor evoked potentials (MEPs) were evoked from the right extensor digitorum communis (EDC) and flexor digitorum superficialis (FDS) muscles prior and immediately after each session. PAS delivered during the muscle contraction of the left hand and PAS delivered at rest both increased the MEP amplitude in the right EDC. PAS delivered before the left hand movement onset led to a decrease in the MEP amplitude measured in the right EDC muscle. We conclude that PAS induced bidirectional changes in the amplitude of MEPs that were dependent on the phase of the movement of the opposite hand.

Keywords: Neural stimulation
Abstract: While intracranial volume is thought to be fixed throughout the lifespan, there is little doubt that the brain shrinks with age and it is most precipitous after about the age of 50. This as a consequence reflects an increase in cranial cerebrospinal fluid (CSF) with age. Of the myriad factors influencing brain current flow, these changes in CSF volume are expected to play a profound role given its high electrical conductivity. The aim of this study is to investigate the effects of age related morphological changes on brain current flow patterns due to transcranial Direct Current Stimulation (tDCS). Anatomical MRI data were collected for 5 healthy subjects spanning 5 decades of life (ages: 43 to 85). Finite element models derived from the MRI were used to calculate cortical electrical field values during tDCS. The widely used C3-Fp2 (M1-SO) and the F3-F4 montage along with two High Definition-tDCS electrode montages 4X1 (C3-centered) and 4X1 (F3-centered) were simulated. Peak induced electrical field at the intended brain target (assumed to be directly underneath the electrode) and at non-intended brain regions was compared with the individual brain atrophy coefficients. Findings across 4 subjects (ages: 43 to 75) indicate reduced peak electrical field with increasing age. However, this trend reverses for the oldest subject. While age-related morphological changes lead to significant changes in current flow distribution, they are not substantially different than younger adults. The predictions of this study are the first step to assess safety of tDCS in elderly subjects and provide a rational path in customizing stimulation dose for trials involving them.

Keywords: Neural stimulation, Neural interfaces - Implantable systems, Brain-computer/machine interface
Abstract: Electrical stimulation of neural tissue and recording of neural activity are the bases of emerging prostheses and treatments for spinal cord injury, stroke, sensory deficits, and drug-resistant neurological disorders. Safety and efficacy are key aspects for the clinical acceptance of therapeutic neural stimulators. The cortical vasculature has been shown to be a safe site for implantation of electrodes for chronically recording neural activity, requiring no craniotomy to access high-bandwidth, intracranial EEG. This work presents the first characterization of endovascular cortical stimulation measured using cortical subdural surface recordings. Visual stimulation was used to verify electrode viability and cortical activation was compared with electrically evoked activity. Due to direct activation of the neural tissue, the latency of responses to electrical stimulation was shorter than for that of visual stimulation. We also found that the center of neural activation was different for visual and electrical stimulation indicating an ability of the stentrode to provide localized activation of neural tissue.

Keywords: Neural stimulation, Neurorehabilitation
Abstract: Abstract—Transcranial direct current stimulation (tDCS) is an emerging non-invasive neuromodulation method that is convenient and popular in clinical use. However, there is a practical issue in applying tDCS; it is difficult to optimize the montage for each individual because of inherent inter-subject variability. Thus, the stimulation effect of such individual anatomical head variation has been investigated using anatomically realistic models. In this work, we developed a multi-scale computational model, which combined head models based on magnetic resonance imaging (MRI) and multi-compartmental neuronal models of pyramidal neurons (PNs), to investigate both the macroscopic and microscopic effects of tDCS. We constructed three different head models and compared the stimulation effects of tDCS in the primary cortex area (Brodmann area 4) with respect to the electric fields induced and steady-state membrane polarizations. We observed that the electric field behavior and somatic polarizations induced varied across subjects depending on the thicknesses of cerebrospinal fluid (CSF) and skull. Thus, we concluded that variations in the CSF and skull might be correlated with the effects of tDCS.

Keywords: Time-frequency and time-scale analysis - Empirical mode decomposition in biosignal analysis
Abstract: In the present study we propose a data-driven, fully unsupervised denoising approach for multi-trial univariate signals. The proposed methodology is based on Empirical Mode Decomposition (EMD) and hence also applicable for transient or non-stationary signals.

The rationale of the presented method is that different realizations (multiple trials) of the same underlying process have also similar intrinsic signal components. These components may be extracted by EMD for each single realization and finally, the entirety of all signal components forms clusters of corresponding components with similar spectral characteristics. A denoising is then tantamount to identifying the cluster(s) containing high-frequency noise components.

The effectiveness of the proposed methodology is demonstrated on the basis of visual event-related potentials (ERPs) of dyslexic and normal control children. We could show that the novel method allows for a reliable ERP estimation and that it provides a tool for an objective extraction of ERPs on both a single-subject as well as on a single-trial basis.

Keywords: Time-frequency and time-scale analysis - Nonstationary processing, Nonlinear dynamic analysis - Biomedical signals, Connectivity measurements
Abstract: This study was aimed at characterizing spontaneous electroencephalography (EEG) activity in Alzheimer’s disease (AD) using a novel approach named weighted visibility graph (WVG). More than 10 minutes of spontaneous EEG were recorded from 15 AD patients and 15 age-matched normal controls. Two graph metrics, clustering coefficient and average weighted degree, are extracted in different frequency bands for each EEG channel based on the WVG methodology. Furthermore, statistical analysis was performed in different bands and channels for both groups. It is demonstrated that AD patients are characterized with a significant increase of clustering coefficient and degree in theta band, which can be observed in most brain regions. Our results suggest that the WVG method can be are effective to distinguish different brain states (AD and normal) and may provide further insights into the underlying brain dynamics in AD.

Keywords: Nonlinear dynamic analysis - Biomedical signals, Time-frequency and time-scale analysis - Time-frequency analysis, Data mining and processing - Pattern recognition
Abstract: The biomedical signal classification accuracy on motor imagery is not always satisfactory, partially because not all the important features have been effectively extracted. This paper proposes an improved dynamic feature extraction approach based on a time-frequency representation and an optimal sequence similarity measurement. Since the wavelet packet decomposition (WPD) generates more detailed signal variation information and the dynamic time warping (DTW) helps optimally measure the sequence similarity, more important features are kept for classification. We apply the extracted features from our proposed method to Electroencephalogram (EEG) based motor imagery through the OpenBCI device and obtain higher classification accuracy. Compared with traditional feature extraction methods, there is a significant classification accuracy improvement from 83.53% to 90.89%. Our work demonstrates the importance of the advanced feature extraction in time series data analysis, e.g. biomedical signal.

Keywords: Time-frequency and time-scale analysis - Time-frequency analysis, Time-frequency and time-scale analysis - Nonstationary processing, Time-frequency and time-scale analysis - Wavelets
Abstract: Electroencephalogram (EEG) signals have been widely used to analyze brain activities so as to diagnose certain brain-related diseases. They are usually recorded for a fairly long interval with adequate resolution, consequently requiring a considerable amount of memory space for storage and transmission. Recently compressed sensing (CS) has been proposed in order to effectively compress EEG signals. However, its performance is closely dependent on how a compression dictionary is built. Through our study, we notice that building the best fit over-complete Gabor dictionary plays an important role in this task. In this paper, we evaluate the effect of different time and frequency step sizes in building Gabor atoms on the performance of EEG signal compression using CS with three common EEG databases used by the research community. Taking the Normalized Mean Square Error (NMSE) as a performance metric, we present a quantitative study with an attempt to provide more insight on how to adopt CS in EEG signal compression.

Keywords: Time-frequency and time-scale analysis - Nonstationary processing, Nonlinear dynamic analysis - Biomedical signals, Data mining and processing in biosignals
Abstract: We introduce a novel technique to analyze non-stationarity in single-channel Electroencephalogram (EEG) traces: the Embedding Transform. The approach is based on Walter J. Freeman's studies concerning active and rest stages and deviations from Gaussianity. Specifically, we generalize his idea in order to include cases where the neuromodulations are sparse in time. Specifically, the transform maps the temporal sequences to a set of ell^2-norms where modulated patters are emphasized. In this way, the background, chaotic activity can be modeled as the main lobe of the distribution, while the relevant synchronizations (or desynchronizations) fall on the right (or left) tail of the density of such norms. We test the algorithm on two different datasets: alpha bursts on synthetic data simulated in the BESA software and low-gamma oscillations in the motor cortex from the Brain-Computer Interface (BCI) Competition 4 Dataset 4. The results are promising and place the Embedding Transform as a quick, single-parameter tool to effectively assess which channels (or regions) are actively engaged in particular behaviors and which are in a more silent type of stage.

Keywords: Time-frequency and time-scale analysis - Time-frequency analysis, Signal pattern classification, Data mining and processing in biosignals
Abstract: High frequency oscillations (HFOs) > 80 Hz are a promising biomarker of epileptic tissue. Recent evidence has shown that spontaneous HFOs can be recorded from the scalp, but detection of these electrographic events remains a challenge. Here, we modified a simple automatic detector, used originally for intracranial EEG (iEEG) recordings, to detect ripples and fast ripples in scalp EEG. We analyzed scalp EEG recordings of seven subjects and validated our detector and artifact rejection algorithm via visual review. Of the candidate events marked by the detector, 40% and 60% were confirmed to be ripples and fast ripples, respectively, by human visual review, making this algorithm suitable for supervised detection. Detected HFOs occurred at a rate of < 1/min in most channels, and the average duration was 47 and 24 ms for ripples and fast ripples, respectively. The simplicity of the algorithm, with only a single parameter, enables the consistent application of automatic detection across recording modalities, and it could therefore be a tool for the assessment and localization of epileptic activity.

Keywords: Ultrasound imaging - Cardiac, Image segmentation, Cardiac imaging and image analysis
Abstract: Rheumatic heart disease is the responsible of heart valves damage, caused by the repeated episodes of rheumatic fever. The disease commonly inflamed and scarred the mitral valve of the heart, resulting in the interruption the normal flow of blood from the left atrium to the left ventricle. It is important to measure and quantify the early manifestation of the disease like, thickness, shape and mobility to avoid permanent valve damage. These changes are visible in the echocardiographic screening. The first step towards the defined objective is to segment the anterior mitral leaflet throughout the cardiac cycle. In this work, a new algorithm for the segmentation of the whole region of the anterior mitral leaflet in the virtual M-mode space is proposed. The algorithm requires a single point initialization on the posterior wall of the aorta, in the very first frame of the video. A junction point is computed, showing the location where two leaflets connect. The junction point helps to redefine the range of virtual M-mode images requires to completely segment the region of the anterior mitral leaflet. The segmented anterior mitral leaflet region in the virtual M-mode space is transferred back to Cartesian coordinates and is refined by the localized active contours. Results suggest the suitability of the proposed algorithm for the segmentation of anterior mitral leaflet with the median similarity score of 62%, with median precision and sensitivity of 57% and 73% respectively.

Keywords: Image classification, Multimodal imaging, Ultrasound imaging - Other organs
Abstract: Contrast-enhanced ultrasound (CEUS) is a valuable imaging modality for diagnosis of liver cancers. However, the complexity of CEUS-based diagnosis limits its wide application, and the B-mode ultrasound (BUS) is still the most popular diagnosis modality in clinical practice. In order to promote BUS-based computer-aided diagnosis (CAD) for liver cancers, we propose a learning using privileged information (LUPI) based CAD with BUS as the diagnosis modality and CEUS as PI. Particularly, the multimodal restricted Boltzmann machine (MRBM) works as a LUPI paradigm. That is, one BUS image and three CEUS images from the arterial phase, portal venous phase and delayed phase, respectively, are used to train three multimodal restricted Boltzmann machine (MRBM) models during training stage, but only the BUS data will be fed to MRBM to generate new feature representation at testing phase. A multiple empirical kernel learning machine (MEKLM) classifier is then performed on three new feature vectors from three MRBM models for classification of liver cancers. The experimental results show that the proposed MRBM-MEKLM algorithm outperforms all the compared algorithms, suggesting the effectiveness of the proposed LUPI-based CAD for liver cancer.

Keywords: Ultrasound imaging - Other organs
Abstract: This paper introduces a non-invasive, contrast-enhanced ultrasound (CEUS) infusion method to quantify the health of viable donor livers. The method uses the infusion of microbubbles and their destruction and subsequent replenishment to measure the perfusion rate in the liver micro-vasculature. The proposed method improves on the previous parameter extraction approaches applied to the flash-replenishment technique by addressing the effects of the microbubble mixing within the perfusate bath and destruction rate. By doing so, the tissue perfusion rate can be extracted from the data even though the microbubble concentration is not constant throughout image acquisition. The measured changes in the tissue perfusion rate showed that CEUS infusion is a viable biomarker for assessing liver health.

Keywords: Image classification, Brain imaging and image analysis, Ultrasound imaging - Other organs
Abstract: Transcranial sonography (TCS) has become more popular for diagnosis of Parkinson’s disease (PD), and the TCS-based computer-aided diagnosis (CAD) for PD also attracts considerable attention, in which classifier is a critical component. Broad learning system (BLS) is a newly proposed single layer feedforward neural network for classification. In BLS, the original input features are mapped to several new feature representations to form the feature nodes, and then these mapped features are expanded to enhancement nodes by random mapping in a wide sense. However, random mapping performed for enhancement nodes is too simple and the generated features lack interpretability together with relative low representation. In this work, we propose a multiple empirical kernel mapping (MEKM) based BLS (MEKM-BLS) algorithm, which adopts MEKM to map the data of feature nodes to enhancement nodes. MEKM-BLS then has more meaningful enhancement layer in feedforward neural network. Moreover, the experiment for PD diagnosis with TCS shows that MEKM-BLS achieves superior performance to the original BLS algorithm.

Keywords: Fetal and Pediatric Imaging, Brain imaging and image analysis, Image segmentation
Abstract: Intraventricular hemorrhage (IVH) followed by post hemorrhagic hydrocephalus (PHH) in premature neonates is one of the recognized reasons of brain injury in newborns. Cranial ultrasound (CUS) is a noninvasive imaging tool that has been used widely to diagnose and monitor neonates with IVH. In our previous work, we showed the potential of quantitative morphological analysis of lateral ventricles from early CUS to predict the PHH outcome in neonates with IVH. In this paper, we first present a new automatic method for ventricle segmentation in 2D CUS images. We detect the brain bounding box and brain mid-line to estimate the anatomical positions of ventricles and correct the brain rotation. The ventricles are segmented using a combination of fuzzy c-means, phase congruency, and active contour algorithms. Finally, we compare this fully automated approach with our previous work for the prediction of the outcome of PHH on a set of 2D CUS images taken from 60 premature neonates with different IVH grades. Experimental results showed that our method could segment ventricles with an average Dice similarity coefficient of 0.8 ± 0.12. In addition, our fully automated method could predict the outcome of PHH based on the extracted ventricle regions with similar accuracy to our previous semi-automated approach (83% vs. 84%, respectively, p-value = 0.8). This method has the potential to standardize the evaluation of CUS images and can be a helpful clinical tool for early monitoring and treatment of IVH and PHH.

Keywords: Ultrasound imaging - Cardiac, Cardiac imaging and image analysis, Image analysis and classification - Machine learning / Deep learning approaches
Abstract: Compared to other modalities such as computed tomography or magnetic resonance imaging, the appearance of ultrasound images is highly dependent on the expertise of the sonographer or clinician making the image acquisition, as well as the machine used, making it a challenge to analyze due to the frequent presence of artefacts, missing boundaries, attenuation, shadows, and speckle. In addition, manual contouring of the epicardial and endocardial walls exhibits large inconsistencies and variations as it is strongly dependent on the sonographer’s training and expertise. Hence, in this paper we propose a fully automated image analysis framework to ultimately perform wall motion abnormality classification in 2D+T images. We explore both traditional Random Forests classification with handcrafted features and spatio-temporal hierarchical aggregation of information with a deep learning CNN-based approach. Regarding the later classifier, we also investigate the effect of local phase information retrieval through the use of Feature Asymmetry (FA), and demonstrate that pre-processing videos with FA enables the spatio-temporal CNN to better discover relevant left ventricle endocardial abstractions from low-level features to high-level representations automatically.

Keywords: Neurological disorders - Diagnostic and evaluation techniques, Human performance - Modelling and prediction, Human performance - Sensory-motor
Abstract: The assessment of Parkinson's disease (PD) using wearable sensors in non-clinical environments presents an opportunity for objective disease classification and severity prediction on a high-frequency and longitudinal basis. However, many challenges exist in analysing remotely collected data due to many sources of data corruption.

Using a cohort of 1,815 participants (866 controls and 949 with PD) we implement a range of classification algorithms on Alternate Finger Tapping test data collected on smart-phones in remote environments. We compare the disease classification ability of two traditional machine learning methods against two state-of-the-art deep learning approaches, determining if the latter is successful without the definition of an explicit feature set. We find the deep learning approaches capable of disease classification, often outperforming traditional methods. We show similarities between the participants successfully classified through use of a manually extracted feature set, and the features learnt by a convolutional neural network. Finally, we discuss the broader challenges of analysing remotely collected datasets whilst highlighting the suitability of deep learning for assessing PD when large participant numbers are available.

Keywords: Neurological disorders - Epilepsy, Neurological disorders - Diagnostic and evaluation techniques, Neural signal processing
Abstract: Diagnosis of epilepsy based on visual inspection of electroencephalogram (EEG) abnormalities is an inefficient, time-consuming, and expert-centered process. Moreover, the diagnosis based on ictal epileptiform events is challenging as the ictal patterns are infrequent. Consequently, the development of an automated, fast, and reliable epileptic EEG diagnostic system is essential. The interictal epileptiform discharges (IEDs) are recurring patterns that are highly suggestive of epilepsy. In this paper, we propose an epileptic EEG classification system based on IED detection. The proposed system comprises of three modules: pre-processing, waveform-level classification, and EEG-level classification. We employ a Convolutional Neural Network (CNN) for waveform-level classification and a Support Vector Machine (SVM) for EEG-level classification. We evaluated the proposed system on a dataset of 156 EEGs recorded at Massachusetts General Hospital (MGH), Boston. The system achieved a mean 4-fold classification accuracy of 83.86% for classifying EEGs with and without IEDs.

Keywords: Neurological disorders - Epilepsy, Motor learning, neural control, and neuromuscular systems, Neuromuscular systems - Learning and adaption
Abstract: This study presents a model predictive control approach for seizure reduction in a computational model of epilepsy. The differential dynamic programming (DDP) algorithm is implemented in a model predictive fashion to optimize a controller for suppressing seizures in a chaotic oscillator model. The chaotic oscillator model uses proportional-integral (PI) controller to represent the internal control mechanism that maintains stable neural activity in a healthy brain. In the pathological case, the gains of this PI controller are reduced, preventing sufficient feedback to suppress correlation increase between normal and pathological brain dynamics. This increase in correlation leads to synchronization of oscillator dynamics leading to the destabilization of neural activity and epileptic behavior. The pathological case of the chaotic oscillator model is formulated as an optimal control problem, which we solve using the dynamic programming principle. We propose using model predictive control with differential dynamic programming optimization as a possible method for controlling epileptic seizures in known models of epilepsy.

Keywords: Neurological disorders, Human performance - Gait, Brain functional imaging - NIR
Abstract: Multiple sclerosis (MS) is one of the neurodegenerative diseases that damages the nervous system and inflicts cognitive and motor deficits. In motor domain, MS mainly causes slower gait that challenges daily life activities. Premotor cortices are affected by MS, where several imaging studies have reported re-organization in the activity and connectivity of these regions. Recent advancements in mobile imaging technologies and signal processing techniques have made it possible to study supraspinal modulation of walking in able-bodied individuals and persons with injuries or neurological disorders. Functional near-infrared spectroscopy (fNIRS), for example, was used in studying dual-tasking in MS population. In the current study, we used fNIRS to record activities of premotor and supplementary motor areas in MS and healthy populations during standing and walking. Fourteen healthy controls and 14 persons with MS were tested during overground walking. Results show higher right premotor cortex activities compared with left premotor and bilateral supplementary motor areas in the MS group. In the healthy control group, activity was higher during walking in all the 4 studied brain regions. These results confirm the role of the premotor cortices in movement planning and modulating walking activities; they also confirm that mild MS has a similar premotor control strategy for a same physical task as healthy controls.

Keywords: Neurological disorders - Stroke, Human performance - Activities of daily living
Abstract: The development of motor impairment after the onset of an injury such as stroke may result in long–term compensatory behaviors. Because compensation often evolves in ambient settings (outside the purview of monitoring clinicians), there is a need for quantitative tools capable of accurately detecting the subtleties of compensation and related reduction in interlimb coordination. Improvement in interlimb coordination may serve as a marker of recovery from stroke, and rehabilitation progress. The current study investigates measures of upper extremity interlimb coordination in persons post–stroke and healthy controls. It introduces a novel algorithm for objective characterization of interlimb coordination during the performance of real–world tasks.

Keywords: Neurological disorders - Traumatic brain injury, Brain functional imaging - MEG, Brain functional imaging - Source localization
Abstract: In this study we investigated whether source connectivity analysis of resting-state Magnetoencephalographic (MEG) activity could be used to separate patients with mild traumatic brain injury (mTBI) from age- and sex-matched controls. For each subject, we used artifact-free data recorded on three sessions to estimate intracranial sources which were then projected onto a standardized brain atlas for common reference. The statistical and topological properties of functional brain networks, estimated using Granger causality, were analyzed using MANOVA, with group and recording session as the independent variables and number of in-going and out-going connections in each atlas region as dependent variables. Overall, mTBI subjects showed a larger number of stronger connections compared to controls. The number and topology of in-going and out-going connections were significantly different across the two groups in areas involved with spatial memory, perception of visual space, emotion formation and processing, learning, and memory. Additionally, the difference between patients and controls was decreasing across the three sessions indicating patient improvement. Our results suggest that connectivity analysis may be used as a reliable biomarker of mTBI and can also help with the diagnosis and assessment of patient recovery.

Keywords: Cardiovascular and respiratory signal processing - Cardiovascular signal processing, Cardiovascular and respiratory system modeling - Cerebrovascular models, Cardiovascular and respiratory system modeling - Vascular mechanics and hemodynamics
Abstract: The compliance of the intracranial space is such that the intracranial pressure (ICP) waveform is approximately the aortic blood pressure (BP) waveform minus the dominant harmonic (heart rate (HR)) of that waveform. This has been tested across species of different resting HR. Whether this filter characteristic holds true for large changes in HR, or under changed ICP, has not been examined. This study tested these changes in 11 anesthetized rats instrumented to measure and change ICP, HR, aortic BP, and carotid blood flow. The aortic BP to ICP and carotid blood flow to ICP transfer function was determined for normal ICP at paced HRs of 300, 400 and 500 bpm, and at raised ICP at HRs of 400 and 500 bpm. The aortic BP to ICP transfer function magnitude showed a dependency on HR (-0.15±0.04 ×10-3 bpm-1, p<0.001) and mean ICP (4.4±0.6 ×10-3 mmHg-1). The carotid blood flow to ICP transfer function magnitude showed a dependency on mean ICP (11.1±1.8 ×10-3 mmHg/ml/min/mmHg, p<0.001) but not on HR. The changes with different HRs indicates a degree of non-linearity in the system. Though small, this may need to be accounted for in understanding the relationship between systemic blood pressure and flow and ICP. This data is useful in understanding the relationship between cardiovascular signals and ICP, valuable in advancing the ability to estimate ICP non-invasively.

Keywords: Vascular mechanics and hemodynamics - Vascular Hemodynamics, Vascular mechanics and hemodynamics - Vascular mechanics, Cardiovascular and respiratory system modeling - Blood flow models
Abstract: Atherosclerosis plays the major role in myocardial infarction and stroke and its pathophysiology is closely related to blood flow. Among clinical imaging modalities, ultrasound has the highest temporal resolution. Doppler ultrasound has been clinically applied for blood flow measurement and several parameters obtained with Doppler have been considered as essential for diagnosis. However, conventional Doppler method merely measures one-dimensional component of the blood flow along the ultrasonic beam. Based on previous approaches with multi-angle Doppler measurement for two-dimensional (2D) blood flow, this study aims to expand 2D flow measurement into three-dimensional (3D) flow estimation by applying continuity equation on multiplane 2D velocity mapping. The algorithm was validated by numerical simulation based on computational fluid dynamics and comparison with particle image velocimetry of carotid artery model. The method visualized 3D spiral flow in the carotid artery bifurcation model where 2D blood flow showed laminar flow. Clinical application of 3D blood flow visualization will provide important information on pathophysiology in common sites of atherosclerosis.

Keywords: Cardiac mechanics, structure & function - Heart failure, Cardiac mechanics, structure & function - Cardiac structure from imaging, Cardiac mechanics, structure & function - Ventricular mechanics
Abstract: Surgical intervention for aortic valve stenosis (AS) has been established; however its diagnosis based on echocardiographic assessment is still limited by aortic valvular velocity, aortic valvular pressure gradients, and color Doppler imaging. Echo-dynamography (EDG) is a method to determine intracardiac flow dynamics, such as two-dimensional blood flow velocity, vortex, and dynamic pressure. These flow dynamics may be influenced by left ventricular (LV) wall motion and the resistance in LV outflow caused by AS. The objective of the present study was to assess the changes and differences in LV vortices and vorticity before and after AS surgery. Five patients who underwent aortic valve replacement surgery for AS and five control patients were included. Besides routine echocardiographic measurement, EDG was applied to determine the two-dimensional blood flow vector and vorticity. The LV vortex flow in the isovolumetric contraction phase had multiple formations in preoperative cases. The clockwise vortex was found in all cases postoperatively; the vortex formation showed no significant difference between postoperative and normal control groups. EDG may serve important information on LV flow dynamics, non-invasively.

Keywords: Coronary blood flow, Vascular mechanics and hemodynamics - Vascular Hemodynamics
Abstract: We investigated the influence of stent implanted in left main coronary artery trifurcation on blood flow by means of CFD. We simulated various stent positions and arrangement patterns considering KBT. The velocity and WSS (wall shear stress) distribution were found to depend on the stent arrangements. In addition, a strut position inhibiting the inflow velocity peaks into the branched (LCX) vessel exhibited a strong impact, which provided suppression of WSS on the high-lateral-side surface of the LCX entrance. By KBT, such an impact of stent implantation can be avoided.

Keywords: Vascular mechanics and hemodynamics - Vascular Hemodynamics, Cardiovascular and respiratory system modeling - Blood flow models, Cardiovascular and respiratory system modeling - Cardiac models
Abstract: Coronary arteries are medium-small caliber vessels, in which low shear rate values are encountered, where non-Newtonian blood effects cannot be neglected. This work aims to study a comparison between Newtonian and non-Newtonian blood behaviors in a cohort of forty-eight 3D patient-specific stenotic vessels (right (RCA), left (LAD) and circumflex (LCX) coronary artery) at different grades of stenosis. Numerical simulation was carried out by means of Computational Fluid Dynamics (CFD) Analysis to investigate the blood velocity and distribution of the shear stress indices at different times of the cardiac cycle. A statistical analysis was performed to have a prediction of increment or decrement of the various hemodynamic parameters. The results show that the non-Newtonian effects are mostly important in shear stress indices distributions.