Keywords: Neural networks and support vector machines in biosignal processing and classification
Abstract: Abstract— Annotation of sleep disordered breathing, including Cheyne-Stokes Breathing (CSB), is an expensive and time- consuming process for the clinician. To solve the problem, this paper presents a deep learning-based algorithm for automatic sample-wise detection of CSB in nocturnal polysomnographic (PSG) recordings. 523 PSG recordings were retrieved from four different sleep cohorts and subsequently scored for CSB by three certified sleep technicians. The data was pre-processed and 16 time domain features were extracted and passed into a neural network inspired by the transformer unit. Finally, the network output was post-processed to achieve physiologically meaningful predictions. The algorithm reached a F1-score of 0.76, close to the certified sleep technicians showing that it is possible to automatically detect CSB with the proposed model.The algorithm had difficulties distinguishing between severe obstructive sleep apnea and CSB but this was not dissimilar to technician performance. In conclusion, the proposed algorithm showed promising results and a confirmation of the performance could make it relevant as a screening tool in a clinical setting.

Keywords: Neural networks and support vector machines in biosignal processing and classification, Data mining and big data methods - Machine learning and deep learning methods, Data mining and big data methods - Biosignal classification
Abstract: This paper presents a new medical severity scoring system, used to assess the risk of hemodynamic and pulmonary decompensation for patients being treated in intensive care units. The score presented here includes drug circulatory support and ventilation mode data for the evaluation of the patient’s biosignals and laboratory values. It is shown that Gated Recurrent Unit-based neural networks are able to predict the maximal severity class within a 24 hour prediction time-frame (hemodynamic: 0.85 AUROC / pulmonary: 0.9 AUROC), and can estimate the underlying decompensation score for prediction times of up to 24 hours with mean errors of 6.3% of the maximal possible pulmonary, and 9.6% of the hemodynamic score. These results are based on 60h observation period.

Keywords: Neural networks and support vector machines in biosignal processing and classification, Physiological systems modeling - Multivariate signal processing
Abstract: Prenatal fetal monitoring, which can monitor the growth and health of the fetus, is vital for pregnant women before delivery. During pregnancy, it is essential to classify whether the fetus is abnormal, which helps physicians carry out early intervention to avoid fetal heart hypoxia and even death. Fetal heart rate and uterine contraction signals obtained by fetal heart monitoring equipment are essential to estimate fetal health status. In this paper, we pre-process the obtained data set and enhance them using Hermite interpolation on the abnormal classification in the samples. We use the 1D-CNN and GRU hybrid models to extract the abstract features of fetal heart rate and uterine contraction signals. Several evaluation metrics are used for evaluation, and the accuracy is 96%, while the sensitivity is 95%, and the specificity is 96%. The experiments show the effectiveness of the proposed method, which can provide physicians and users with more stable, efficient, and convenient diagnosis and decision support.

Keywords: Signal pattern classification, Neural networks and support vector machines in biosignal processing and classification
Abstract: This paper evaluates a range of deep learning frameworks for detecting respiratory anomalies from input audio. Audio recordings of respiratory cycles collected from patients are transformed into time-frequency spectrograms to serve as front-end two-dimensional features. Cropped spectrogram segments are then used to train a range of back-end deep learning networks to classify respiratory cycles into predefined medically-relevant categories. A set of those trained high-performing deep learning frameworks are then fused to obtain the best score. Our experiments on the ICBHI benchmark dataset achieve the highest ICBHI score to date of 57.3%. This is derived from a late fusion of inception based and transfer learning based deep learning frameworks, easily outperforming other state-of-the-art systems.

Keywords: Neural networks and support vector machines in biosignal processing and classification, Time-frequency and time-scale analysis - Time-frequency analysis, Data mining and big data methods - Pattern recognition
Abstract: The COVID-19 pandemic has fueled exponential growth in the adoption of remote delivery of primary, specialty, and urgent health care services. One major challenge is the lack of access to physical exam including accurate and inexpensive measurement of remote vital signs. Here we present a novel method for machine learning-based estimation of patient respiratory rate from audio. There exist non-learning methods but their accuracy is limited and work using machine learning known to us is either not directly useful or uses non-public datasets. We are aware of only one publicly available dataset which is small and which we use to evaluate our algorithm. However, to avoid the overfitting problem, we expand its effective size by proposing a new data augmentation method. Our algorithm uses the spectrogram representation and requires labels for breathing cycles, which are used to train a recurrent neural network for recognizing the cycles. Our augmentation method exploits the independence property of the most periodic frequency components of the spectrogram and permutes their order to create multiple signal representations. Our experiments show that our method almost halves the errors obtained by the existing (non-learning) methods.

Keywords: Physiological monitoring - Instrumentation, Optical and photonic sensors and systems, Health monitoring applications
Abstract: Monitoring of heart rate in patients in the general ward is necessary to assess the clinical situation of the patient. Currently, this is done via spot-checks on pulse rate manually or on heart rate using Electrocardiogram (ECG) by nurses. More frequent measurements would allow early detection of adverse cardiac events. In this work, we investigate a contactless measurement setup combined with a signal processing pipeline, which is based on speckle vibrometry (SV), to perform contactless heart rate monitoring of human subjects in a supine position, mimicking a resting scenario in the general ward. Our results demonstrate the feasibility of extracting heart rate with SV through varying textile thicknesses (i.e., 8 mm, 32mm and 64 mm), with an error smaller than 3 beats per minute on average compared to the ground-truth heart rate derived from ECG.

Keywords: Acoustic sensors and systems, New sensing techniques, Novel methods
Abstract: The design and performance of a transducer for low frequency ultrasound tomography is presented, motivated by recent research demonstrating that acoustic waves transmitting at frequencies between 10 kHz and 750 kHz penetrate the lungs and may be useful for thoracic imaging. An adaptation of the traditional Tonpilz design was developed, vibrational amplitude and electrical impedance were measured, and an optimal frequency was determined. The design is found to meet the desired mechanical, electrical, and safety specifications. Thus, it was considered a promising option for the target application of pulmonary imaging with ultrasound computed tomography between 50 and 200 kHz; highest efficiency achieved around 125 kHz and 156 kHz, and beam divergence of 40°.

Keywords: Wearable body-compliant, flexible and printed electronics, Wearable sensor systems - User centered design and applications, Wearable low power, wireless sensing methods
Abstract: a button sensor antenna for wireless medical body area networks (WMBAN) is presented, which works through the IEEE 802.11b/g/n standard. Due to strong interaction between the sensor antenna and body, an innovative robust system is designed with a small footprint that can serve on- and off-body healthcare applications. The measured and simulated results are in good agreement. The design offers a wide range of omnidirectional radiation patterns in free space, with a reflection coefficient (S11) of -29.30 (-30.97) dB in the lower (upper) bands. S11 reaches up to -23.07 (-27.07) dB and -30.76 (-31.12) dB, respectively, on the human body chest and arm. The Specific Absorption Rate (SAR) values are below the regulatory limitations for both 1-gram (1.6 W/Kg) and 10-gram tissues (2.0 W/Kg). Experimental tests of the read range validate the results of a maximum coverage range of 40 meters.

Keywords: Integrated sensor systems, Chemo/bio-sensing - Biological sensors and systems
Abstract: A capacitor transimpedance amplifier (CTIA) for wireless-powered wearable electrochemical sweat sensors is designed and tested. Correlative double sampling (CDS) technology is utilized to suppress offset voltage and noise. Dedicated low-power and low-voltage designs meet the requirements of wireless-powered wearable applications where the power supply is limited, and voltae is unstable. The proposed CDS-CTIA is fabricated in 0.18µm complementary metal oxide semiconductor (CMOS) process, occupying an active area of 0.0285mm². The measured low-frequency gain is 33.2MΩ/150.4dBΩ under a 1.8V supply voltage, with a total power consumption of 17.694µW (including 14.882µW static power and 2.812µW dynamic power). The input current ranges from -24nA to +19nA, and the input referred noise current is 7.76pArms in the 0.1-100 Hz frequency band. The proposed CDS-CTIA is capable of operating under a wide power supply ranging from 0.9V to 2.0V. In addition, its practicality is verified by measuring glucose concentration with an enzyme electrode.

Keywords: Integrated sensor systems, Chemo/bio-sensing - Biological sensors and systems, Implantable systems
Abstract: Diabetes has become a leading cause of death and disability in the past decades. Continuous glucose monitoring (CGM) is a prevailing technique to determine the glucose level and provide in-time treatment. However, conventional CGM systems combine an electrochemical sensor with a CMOS chip, suffering from bulky size and interface issues. Integrating the CGM sensor on silicon is potential to miniaturize the CGM system and reduce the cost, while the recent silicon-based sensors show limited detection range and sensitivity. In this work, we present a silicon-based CGM chiplet with wireless power transfer (WPT) and real-time wireless telemetry. Fabricated on a single silicon substrate, the chiplet consists of a silicon-based CGM sensor, a power-harvesting wirelesstelemetry chip, and a silicon-based antenna. Measured results show that the chiplet achieves a sensitivity of 4 µA·mM·cm−2 and a linear detection range of 0−10 mM. Based on WPT and backscattering communication, the chiplet consumes 18.8 µW power in glucose telemetry.

Keywords: Connectivity, Time-frequency and time-scale analysis - Time-frequency analysis, Independent component analysis
Abstract: Functional connectivity is a widely used measure for finding the relationships between functional entities of the brain. Recently, more focus has been put on the methods that aim to estimate these relationships in a time-resolved fashion. However, the similarities and differences between these methods are not always clear and can result in unfair and incorrect comparisons. Here, we present a framework that provides a unified, systematic view for some of the more well-known methods. Using the proposed unified framework, we explain different methodologies using a unified language and show how they are similar and different conceptually. We give examples of how this framework exposes important assumptions made by various methods, which can help clarify differences in results and facilitate reproducibility. We also show how such a framework will enable us to develop methods that improve upon previous methods.

Keywords: Time-frequency and time-scale analysis - Wavelets, Data mining and big data methods - Biosignal classification
Abstract: In this paper, we explore the use of the 4D discrete wavelet transform (DWT) on fMRI data. The data set comes from a study on schizophrenia. The compact support of the wavelet transform means that it keeps phenomenon localized in all 4 dimensions. We examine 16 sub-signals (frequency banded components of the signal) resulting from the 4D DWT, representing the sub-signals as energy fluctuations over time. Next, we correlate these, and examine the variance. We find that grouping the variance data shows a small but clear trend of the control group versus those with a schizophrenia diagnosis. The DWT analyzes a signal into approximations and details, which we expect reflect the similarities and differences between brain activity, respectively. We want to know how the regions of brain activity vary with time, and the high-pass data encodes these changes over time.

Keywords: Data mining and big data methods - Machine learning and deep learning methods, Connectivity, Coupling and synchronization - Nonlinear coupling
Abstract: Many studies in neuroscience have focused on interpreting brain activity using functional connectivity (FC). The most widely used approach for measuring FC is based on linear correlation (e.g., the Pearson correlation), where the temporal cofluctuations between functional brain regions are computed. However, such approaches ignore nonlinear dependencies among regions that might carry distinctive information across groups of subjects. In this study, we offer a deep learning-based approach that also captures nonlinear temporal relationships between brain networks. Our approach consists of two main parts: a decoder that learns domain-specific embeddings of time courses estimated from independent component analysis (ICA) and a similarity metric that measures the similarities between the embeddings. We call such similarities as nonlinear functional relationships between networks. Our findings on a large dataset (including above 11k normal control subjects) suggest that male subjects exhibit stronger nonlinear network-network relationships than female subjects in most cases. Furthermore, we observe that, unlike FC, our approach could capture some intra-network relationships, especially between cognitive control and visual networks, which are significantly different between males and females, suggesting that our approach can provide a complementary

Keywords: Data mining and big data methods - Machine learning and deep learning methods, Connectivity, Time-frequency and time-scale analysis - Wavelets
Abstract: The high dimensionality and complexity of time-varying measures of functional brain connectivity have created an environment in which a very rich transformation of the data remains difficult to map into disease states without some form of reduction (averaging, clustering, statistical blindness to the multivariate interactions between features that modulate their contributions). In this work, employing a recently developed architecture for long short-term memory classifiers that supports use of gradient-based model interpretability techniques, we predict progression or recovery from mild cognitive impairment (MCI) from an instantaneous (windowless) wavelet-based measure of dynamic functional network connectivity. This time-attention LSTM (TA-LSTM) model achieves 0.79 AUC on the task of predicting which MCI patients who will recover (RMCI) vs. those who will progress (PMCI) to AZD within a three-year timeframe. Using a common gradient-based model interpretation technique, saliency analysis, on this TA-LSTM points to potentially important predictive dynamic biomarkers, including the duration of the highly salient time intervals and the average connectivity patterns within these highly salient intervals.

Keywords: Independent component analysis, Physiological systems modeling - Closed loop systems, Signal pattern classification
Abstract: Functional near-infrared spectroscopy (fNIRS), a non-invasive optical neuroimaging technique, has demonstrated its great potential in monitoring cerebral activity as an alternative to functional magnetic resonance imaging (fMRI) in research and clinical usage. fNIRS has seen increasing applications in studying the auditory cortex in healthy subjects and cochlear implant users. However, fNIRS is susceptible to motion artifacts, especially those related to jaw movement, which can affect fNIRS signals in speech and auditory tasks. This study aimed to investigate the motion artifacts related to jaw movements including clenching, speaking, swallowing, and sniffing in a group of human subjects, and test whether our previously established denoising algorithm namely PCA-GLM can reduce the motion artifacts. Our results have shown that the jaw movements introduced artifacts that resemble task-evoked activations and that the PCA-GLM method effectively reduced the motion artifacts due to the clenching movements. The preliminary results of the present study underline the importance of the removal of the jaw-movement-related artifacts in fNIRS signals and suggest the efficacy of our PCA-GLM method in reducing the motion artifacts.

Keywords: Sensor systems and Instrumentation, Modeling and analysis, Wearable sensor systems - User centered design and applications
Abstract: A cognitive and physical stress co-classification effort started with acquisition of a training dataset and generation of machine learning models from 17 heart rate variability parameters. Accuracy was improved with multilayer perceptron models and tested on 85 firefighters in a cage maze. A specific platform acquired a dataset with better label accuracy providing a second model. Feature importance and model performance were assessed using the cage maze data. A SHAP analysis provided the basis for the model comparison and feature important assessment. Conclusions were drawn on best time windows, feature selection, and model hyperparameters.

Keywords: Health monitoring applications, Physiological monitoring - Novel methods, Modeling and analysis
Abstract: Stress detection and monitoring is an active area of research with important implications for an individual's personal, professional, and social health. Current approaches for stress classification use traditional machine learning algorithms trained on features computed from multiple sensor modalities. These methods are data and computation-intensive, rely on hand-crafted features, and lack reproducibility. These limitations impede the practical use of stress detection and classification systems in the real world. To overcome these shortcomings, we propose Stressalyzer, a novel stress classification and personalization framework from single-modality sensor data without feature computation and selection. Stressalyzer uses only Electrodermal activity (EDA) sensor data while providing competitive results compared to the state-of-the-art techniques that use traditional machine learning models. Our single-channel neural network-based model achieves a classification accuracy of 92.9% and an f1 score of 0.89 for binary stress classification. Our leave-one-subject-out analysis establishes the subjective nature of stress and shows that personalizing stress models using Stressalyzer significantly improves the model performance. Without model personalization, we found a performance decline in 40% of the subjects, suggesting the need for model personalization.

Keywords: Wearable sensor systems - User centered design and applications, Physiological monitoring - Modeling and analysis
Abstract: The aim of this study was to create a robust generalizable model to classify a person's affective state based on physiological signals obtained using wearable sensor devices. Traditional machine learning methods require manual feature extraction from time sequences. Deep learning methods, such as Convolutional Neural Networks (CNN), can automatically extract features from time sequences. However, CNN models can be prone to overfitting, especially when the dataset is small. We apply a novel idea of using unsupervised convolutional autoencoders to automatically extract features from time-series data that are then fed to supervised classifier to classify people's affective state. We achieve almost 3% accuracy increase over traditional CNN model using all physio data from WESAD dataset, 2% increase using chest only physio data, and 8% increase using wrist only physio data while classifying neutral, stress, and amusement states. Code to reproduce the results can be found at https://github.com/srovins/wesad

Keywords: Physiological monitoring - Novel methods
Abstract: Detecting the correct emotion is crucial for improved mental health outcomes. While the existing works on emotion recognition focus on detecting the common or primary emotions only, there exists some uncommon or secondary emotions of similar kind (e.g., contempt vs anger) that makes accurate emotion detection challenging. Moreover, there exists limited labeled data on such secondary emotions. We present the first work to accurately discriminate among such similar emotions by generating distinguishable data for the secondary emotions using convenient multimodal wrist sensors. Extensive evaluations show that our novel solution provides around 7-36% F1-score improvement to existing solutions for similar emotions, and also significantly reduces the burden on providing labeled emotion data.

Keywords: Physiological monitoring - Modeling and analysis, Wearable sensor systems - User centered design and applications, Modeling and analysis
Abstract: Flow is a mental state experienced during holistic involvement in a certain task, and it is a factor that promotes motivation, development, and performance. A reliable and objective estimation of the flow is essential for moving away from the traditional self-reporting subjective questionnaires, and for developing closed-loop human-computer interfaces. In this study, we recorded EEG and pupil dilation in a cohort of participants solving arithmetic problems. In particular, the EEG activity was acquired with a prototype of a commercial headset from Logitech with nine dry electrodes incorporated in a pair of over-ear headphones. The difficulty of the tasks was adapted to induce mental Boredom, Flow and Overload, corresponding to too easy, optimal and too challenging tasks, respectively. Results indicated statistically significant differences between all pairs of conditions for the pupil dilation, as well as for the EEG activity for the electrodes in the ear-pads. Furthermore, we built a predictive model that estimated the mental state of the user from their EEG data with 65% accuracy.

Keywords: Health monitoring applications, Wearable sensor systems - User centered design and applications, IoT sensors for health monitoring
Abstract: Previous studies have shown the correlation between sensor data collected from mobile phones and human depression states. Compared to the traditional self-assessment questionnaires, the passive data collected from mobile phones is easier to access and less time-consuming. In particular, passive mobile phone data can be collected on a flexible time interval, thus detecting moment-by-moment psychological changes and helping achieve earlier interventions. Moreover, while previous studies mainly focused on depression diagnosis using mobile phone data, depression forecasting has not received sufficient attention. In this work, we extract four types of passive features from mobile phone data, including phone call, phone usage, user activity, and GPS features. We implement a long short-term memory (LSTM) network in a subject-independent 10-fold cross-validation setup to model both a diagnostic and a forecasting tasks. Experimental results show that the forecasting task achieves comparable results with the diagnostic task, which indicates the possibility of forecasting depression from mobile phone sensor data. Our model achieves an accuracy of 77.0% for major depression forecasting (binary), an accuracy of 53.7% for depression severity forecasting (5 classes), and a best RMSE score of 4.094 (PHQ-9, range from 0 to 27).

Keywords: Health Informatics - Health data acquisition, transmission, management and visualization, Health Informatics - Computer-aided decision making, Health Informatics - Decision support methods and systems
Abstract: Falls associated injuries often result not only increasing the medical-, social- and care-cost but also loss of mobility, impair chronic health and even potential risk of fatality. Because of elderly population growth, it is one of the major global public health problems. To address such issue, we present a Deep Learning enabled Fall Detection (DLFD) method exploiting Gait Analysis. More in details, firstly, we propose a framework for fall detection system. Secondly, we discussed the proposed DLFD method which exploits fall and non-fall RGB video to extract gait features using MediaPipe framework, applies normalization algorithm and classifies using bi-directional Long Short-Term Memory (bi-LSTM) model. Finally, the model is tested on collected three public datasets of 434x2 videos (more than 1 million frames) which consists of different activities and varieties of falls. The experimental results show that the model can achieve the accuracy of 96.35% and reveals the effectiveness of the proposal. This could play a significant role to alleviate falls problem by immediate alerting to emergency and relevant teams for taking necessary actions. This will speed up the assistance proceedings, reduce the risk of prolonged injury and save lives.

Keywords: Health Informatics - Health data acquisition, transmission, management and visualization, Health Informatics - Health information system interoperability, Health Informatics - Health information systems
Abstract: Shriners Children's (SHC) is a hospital system whose mission is to advance the treatment and research of pediatric diseases. SHC success has generated a wealth of clinical data. Unfortunately, barriers to healthcare data access often limit data-driven clinical research. We decreased this burden by allowing access to clinical data via the standardized data access standard called FHIR (Fast Healthcare Interoperability Resources). Specifically, we converted existing data in the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) standard into FHIR data elements using a technology called OMOP-on-FHIR. In addition, we developed two applications leveraging the FHIR data elements to facilitate patient cohort curation to advance research into pediatric musculoskeletal diseases. Our work enables clinicians and clinical researchers to use hundreds of currently available open-sourced FHIR applications. Our successful implementation of OMOP-on-FHIR within a large hospital system will accelerate advancements in pediatric disease treatment and research.

Keywords: Public Health Informatics - Health risk evaluation and modeling, General and theoretical informatics - Machine learning, General and theoretical informatics - Natural language processing
Abstract: Depression is among the most prevalent mental health disorders with increasing prevalence worldwide. While early detection is critical for the prognosis of depression treatment, detecting depression is challenging. Previous deep learning research has thus begun to detect depression with the transcripts of clinical interview questions. Since approaches using Bidirectional Encoder Representations from Transformers (BERT) have demonstrated particular promise, we hypothesize that ensembles of BERT variants will improve depression detection. Thus, in this research, we compare the depression classification abilities of three BERT variants and four ensembles of BERT variants on the transcripts of responses to 12 clinical interview questions. Specifically, we implement the ensembles with different ensemble strategies, number of model components, and architectural layer combinations. Our results demonstrate that ensembles increase mean F1 scores and robustness across clinical interview data. This research highlights the potential of ensembles to detect depression with text which is important to guide future development of healthcare application ecosystems.

Keywords: Public Health Informatics - Infectious disease outbreak modeling, Public Health Informatics - Epidemiological modeling, Public Health Informatics - Non-medical data analytics in public health
Abstract: A novel compartmental model that includes vaccination strategy, permanence in hospital wards and tracing of infected individuals has been implemented to forecast hospital overload caused by COVID-19 pandemics in Italy. The model parameters were calibrated according to available data on cases, hospital admissions, and number of deaths in Italy during the second wave, and were validated in the timeframe corresponding to the first successive wave where vaccination campaign was fully operational. This model allowed quantifying the decrease of hospital demand in Italy associated with the vaccination campaign. Clinical relevance: this study provides evidence for the ability of deterministic SIR-based models to accurately forecast hospital demand dynamics, and support informed decisions regarding dimensioning of hospital personnel and technologies to respond to large-scale epidemics, even when vaccination campaigns are available.

Keywords: Point of care - Biomarkers, Point of care - Detection and monitoring, Point of care - Diagnostics
Abstract: In the past half-century, the advent of solid-state electronics, i.e., microcontrollers, transistors, photodiodes, light-emitting diodes and more, has led to the improvement of the tools we, as a human race, need and use in our daily lives. Solid-state electronics has specifically contributed significantly to the field of biomedical engineering and has allowed various round-the-clock point-of-care testing applications. These include handheld, wearable, and implantable sensors and devices for accelerated interventions. Furthermore, miniaturization has accelerated the implementation of low-cost and energy-efficient systems with increased performance. In this paper, we have used optical techniques along with the benefits of solid-state electronics to measure bilirubin concentration in plasma with concentrations projected from healthy individuals to hyperbilirubinemia (0 – 30 mg/dL). Traditionally, full-range spectrophotometry is the gold standard optical method and provides the most accurate results but suffers from instrument complexity. Thus, this paper proposes and investigates the measurement of bilirubin by using a dual-wavelength approach combined with photodegradation kinetics. By tracking the changes in the spectral characteristics of bilirubin for 10 minutes (~3 J/cm2), a new model was built to measure bilirubin concentrations and distinguish between low vs high and risky vs non-risky levels. Results show a high positive correlation between the optical responses and concentration (R-square > 0.93) with an average accuracy of ~1.4 mg/dL. On top of that, the technique's viability for point-of-care testing of bilirubin levels was studied using a system-on-chip optical module. Thus, this could help suggest neonatal therapeutic interventions, including enteral feeding, phototherapy, and blood transfusion.

Keywords: Modeling of cell, tissue, and regenerative medicine - PK/PD, Model building - Algorithms and techniques for systems modeling
Abstract: The gold-standard approach to quantifying dynamic PET images relies on using invasive measures of the arterial plasma tracer concentration. An attractive alternative is to employ an image-derived input function (IDIF), corrected for spillover effects and rescaled with venous plasma samples. However, venous samples are not always available for every participant. In this work, we used the nonlinear mixed-effects modeling approach to develop a model which infers venous tracer kinetics by using venous samples obtained from a population of healthy individuals and integrating subject-specific covariates. Population parameters (fixed effects), their between-subject variability (random effects), and the effects of covariates were estimated. The selected model will allow to reliably infer venous tracer kinetics in subjects with missing measurements from the same population.

Keywords: Data-driven modeling, Model building - Parameter estimation, Systems biology and systems medicine - Modeling of biomolecular system dynamics
Abstract: The societal and economic burden of unassessed and unmodeled postoperative pain is high and predicted to rise over the next decade, leading to over-dosing as a result of subjective (NRS-based) over-estimation by the patient. This study identifies how post-surgical trauma alters the parameters of impedance models, to detect and examine acute pain variability. Model identification is performed on clinical data captured from post-anesthetized patients, using Anspec-PRO prototype apriori validated for clinical pain assessment. The multisine excitation of this in-house developed device enables utilizing the complex skin impedance frequency response in data-driven electrical models. The single-dispersion Cole model is proposed to fit the clinical curve in the given frequency range. Changes in identified parameters are analyzed for correlation with the patient’s reported pain for the same time moment. The results suggest a significant correlation for the capacitor component. Clinical relevance— Individual model parameters validated on patients in the post-anesthesia care unit extend the knowledge for objective pain detection to positively influence the outcome of clinical analgesia management.

Keywords: Model building - Algorithms and techniques for systems modeling, Model building - Parameter estimation
Abstract: Among the non-invasive methods employed for brain stimulation, transcranial Focused Ultrasound Stimulation (tFUS) is the technique with the best penetration into the tissues and spatial resolution. The development of computational models of ultrasound propagation in brain tissue can be useful for estimating the behaviour of neural cells subjected to mechanical stimulus due to ultrasound. This paper aims at studying the neural cell response of a cortical Regular Spiking point neuron model, for different values of stimulus Duty Cycle (DC). The main goal is to use a multiscale approach to couple the results obtained from a macroscale simulation on wave propagation in tissue, with neuron model described by Hodgkin-Huxley equations to study latency and firing rate of the RS model. The obtained results showed that latency and firing rate have slight variations along the propagation direction of the ultrasound beam, in the focal region under the skull model, for different stimulus DC.

Keywords: Computational modeling - Analysis of high-throughput systems biology data, Translational biomedical informatics - Knowledge modeling, Synthetic biology
Abstract: Changes in skin appearance are among the most recognizable symptoms of a number of medical conditions. The interpretation of such changes, however, may be inadvertently biased by normal physiological processes affecting skin optical properties. In this paper, we assess the impact of one of the most common of these processes, tanning, on variations in skin chromatic attributes elicited by two ubiquitous and serious medical conditions, anemia and hyperbilirubinemia. We employ a first-principles investigation approach centered on the use of predictive computer simulations of light and skin interactions, and on well-established colorimetry methods. In our in silico experiments, we considered skin chromatic attributes resulting from distinct anemia severity levels and hyperbilirubinemia toxicity stages. Our findings highlight qualitative and quantitative aspects that need to be considered in the visual screening and monitoring of these conditions, notably when they occur with the concomitant presence of tanning-induced changes in the cutaneous tissues' melanin pigmentation and thickness.

Keywords: Models of medical devices
Abstract: This study is based on the quantification of the influence of the presence of gold nanoparticles (Au NPs), of their geometry and their distance from cell membrane during time-varying electromagnetic fields cell membrane permeabilization on the pores opening dynamics. Results showed that the combined use of Au NPs and time-varying magnetic field can improve significantly the permeabilization of cell membrane. The presence of Au NPs allowed to reach transmembrane potential values enabling the cell membrane permeabilization only when placed at very short distance, equal to 20 nm. Both geometry and variability of the positioning in proximity of the cell membrane showed a strong influence on the probability of enabling pores opening.

Keywords: Synthetic biology, Model building - Sensitivity analysis, Translational biomedical informatics - Knowledge modeling
Abstract: The hyperspectral reflectance and transmittance of flowing blood samples are employed in a wide range of biomedical applied research initiatives such as the detection and monitoring of hematological abnormalities. The success of these initiatives is tied to the correct interpretation of these radiometric quantities. This, in turn, requires a comprehensive understanding about their sensitivity to variations in the experimental conditions in which they have been obtained. In this paper, we aim to contribute to these efforts by systematically examining the effects of sample thickness variations on these quantities. More specifically, we employed controlled in silico experiments to assess these effects on samples with different biophysical characteristics, notably their hematocrit, hemolysis level and orientation of their constituent cells with respect to the flow direction. To ensure a high degree of fidelity in our experiments, we used a first-principles simulation framework supported by measured data. Our findings unveil distinct spectrally-dependent trends associated with reflectance and transmittance changes elicited by sample thickness variations.

Keywords: Image segmentation, CT imaging applications, PET and SPECT Imaging applications
Abstract: The prediction of cancer characteristics, treatment planning and patient outcome from medical images generally requires tumor delineation. In Head and Neck cancer (H&N), the automatic segmentation and differentiation of primary Gross Tumor Volumes (GTVt) and malignant lymph nodes (GTVn) is a necessary step for large-scale radiomics studies to predict patient outcome such as Progression Free Survival (PFS). Detecting malignant lymph nodes is also a crucial step for Tumor-Node-Metastases (TNM) staging and to support the decision to resect the nodes. In turn, automatic TNM staging and patient outcome prediction can greatly benefit patient care by helping clinicians to find the best personalized treatment. We propose the first model to automatically individually segment GTVt and GTVn in PET/CT images. A bi-modal 3D U-Net model is trained for multi-class and multi-components segmentation on the multi-centric HECKTOR 2020 dataset containing 254 cases. The dataset has been specifically re-annotated by experts to obtain ground truth GTVn contours. The results show promising segmentation performance for the automation of radiomics pipelines and their validation on large-scale studies for which manual annotations are not available. An average test Dice Similarity Coefficients (DSC) of 0.717 is obtained for the segmentation of GTVt. The GTVn segmentation is evaluated with an aggregated DSC to account for the cases without GTVn, which is estimated at 0.729 on the test set.

Keywords: Image segmentation, Machine learning / Deep learning approaches, PET and SPECT imaging
Abstract: In metastatic breast cancer, bone metastases are prevalent and associated with multiple complications. Assessing their response to treatment is therefore crucial.

Most deep learning methods segment or detect lesions on a single acquisition while only a few focus on longitudinal studies. In this work, 45 patients with baseline (BL) and follow-up (FU) images recruited in the context of the EPICURE_seinmeta study were analyzed. The aim was to determine if a network trained for a particular timepoint can generalize well to another one, and to explore different improvement strategies. Four networks based on the same 3D U-Net framework to segment bone lesions on BL and FU images were trained with different strategies and compared. These four networks were trained 1) only with BL images 2) only with FU images 3) with both BL and FU images 4) only with FU images but with BL images and bone lesion segmentations registered as input channels. With the obtained segmentations, we computed the PET Bone Index (PBI) which assesses the bone metastases burden of patients and we analyzed its potential for treatment response evaluation.

Dice scores of 0.53, 0.55, 0.59, and 0.62 were respectively obtained on FU acquisitions. The under-performance of the first and third networks may be explained by the lower SUV uptake due to treatment response in FU images compared to BL images. The fourth network gives better results than the second network showing that the addition of BL PET images and bone lesion segmentations as prior knowledge has its importance. With an AUC of 0.86, the difference of PBI between two acquisitions could be used to assess treatment response.

Keywords: Image segmentation, Machine learning / Deep learning approaches, Image analysis and classification - Digital Pathology
Abstract: Advancements in deep learning techniques have proved useful in biomedical image segmentation. However, the large amount of unlabeled data inherent in biomedical imagery, particularly in digital pathology, creates a semi-supervised learning paradigm. Specifically, because of the time consuming nature of producing pixel-wise annotations and the high cost of having a pathologist dedicate time to labeling, there is a large amount of unlabeled data that we wish to utilize in training segmentation algorithms. Pseudo-labeling is one method to leverage the unlabeled data to increase overall model performance. We adapt a method used for image classification pseudo-labeling to select images for segmentation pseudo-labeling and apply it to 3 digital pathology datasets. To select images for pseudo-labeling, we create and explore different thresholds for confidence and uncertainty on an image level basis. Furthermore, we study the relationship between image-level uncertainty and confidence with model performance. We find that the certainty metrics do not consistently correlate with performance intuitively, and abnormal correlations serve as an indicator of a model's ability to produce pseudo-labels that are useful in training.

Keywords: Image segmentation, Brain imaging and image analysis, Image feature extraction
Abstract: Multiple Sclerosis (MS) lesions detection and disease’s progression monitoring at the same time, play an important role. The purpose of this research is to present a method for the detection of MS plaques and volume estimation from MR Images for monitoring the progression of the disease and the brain atrophy caused. The proposed study consists of a clustering-based method for the delineation of MS plaques, utilizing anatomical information, brain geometry and lesion features, while volume quantification is used for the estimation MS atrophy by determining Brain Parenchymal Fraction (BPF), also volumetric information about lesions and whole brain volume are examined. In the present study, Fluid Attenuated Inversion Recovery (FLAIR) images were utilized for the detection of MS lesions and BPF estimation, while T1-weighted MR Images utilized in volume estimation. 30 MS patients were included in a dataset consisted of 3D FLAIR and T1-weighted MR images in order to implement the proposed technique. MRI scans performed in two different clinical visits, a baseline and a visit after 6 months. The results extracted in segmentation of MS lesions in terms of sensitivity is 73.80 %. The BPF at baseline estimated to 0.82 ± 0.01, and at 1st follow up, 0.83 ± 0.01. Finally, the brain volume loss between baseline and after 6 months is 0.4%.

Keywords: Image segmentation, Machine learning / Deep learning approaches, Image analysis and classification - Machine learning / Deep learning approaches
Abstract: Precise segmentation of organs at risk (OARs) in computed tomography (CT) images is an essential step for lung cancer radiotherapy. However, the manual delineation of OARs is time-consuming and subject to inter-observer variation. Although U-like architecture has achieved great success in medical image segmentation recently, it exhibits the limitations in modeling long-range dependencies. As an alternative structure, Transformers have emerged due to the outstanding capability of capturing the global contextual information provided by Self-Attention(SA) mechanism. However, Transformers need more computational cost than CNNs for introducing the SA module. In this paper, we propose a novel module named fine-grained combination of Transformer and CNN(FcTC). FcTC module is composed of dual-path extractor and fusing unit to effectively extract local information and model long-distance dependency. Then we build FcTC-UNet to automatically segment the OARs in thoracic CT images. The experiments results demonstrate that the proposed method achieves better performance over other state-of-the-art methods.

Keywords: Image segmentation, Machine learning / Deep learning approaches
Abstract: Interactive segmentation has recently attracted attention for specialized segmentation tasks where expert input is required to further enhance the segmentation performance. In this work, we propose a novel interactive segmentation framework, where user clicks are dynamically adapted in size based on the current segmentation mask. The clicked regions form a weight map and are fed to a deep neural network as a novel weighted loss function. To evaluate our loss function, an interactive U-Net (IU-Net) model which applies both foreground and background user clicks as the main method of interaction is employed. We train and validate on the BCV dataset, while testing on both seen and unseen structures from the MSD dataset to determine the models generalization and segmentation abilities in comparison to the standard U-Net. Applying dynamic user click sizes increases the overall accuracy by 5.60% and 10.39% for seen and unseen structures respectively by utilizing only a single user interaction.

Keywords: Machine learning / Deep learning approaches, Image segmentation, Brain imaging and image analysis
Abstract: Multi-modality images have been widely used and provide comprehensive information for medical image analysis. However, acquiring all modalities among all institutes is costly and often impossible in clinical settings. To leverage more comprehensive multi-modality information, we propose privacy secured decentralized multi-modality adaptive learning architecture named ModalityBank. Our method could learn a set of effective domain-specific modulation parameters plugged into a common domain-agnostic network. We demonstrate by switching different sets of configurations, the generator could output high-quality images for a specific modality. Our method could also complete the missing modalities across all data centers, thus could be used for modality completion purposes. The downstream task trained from the synthesized multi-modality samples could achieve higher performance than learning from one real data center and achieve close-to-real performance compare with all real images.

Keywords: Machine learning / Deep learning approaches, Image segmentation, Image analysis and classification - Digital Pathology
Abstract: Accurate segmentation of nuclei is an essential step in the analysis of digital histology images for diagnostic and prognostic applications. Despite recent advances in automated frameworks for nuclei segmentation, this task is still challenging. Specifically, detecting small nuclei in large-scale histology images and delineating the border of touching nuclei accurately is a complicated task even for advanced deep neural networks. In this study, a cascaded deep learning framework is proposed to segment nuclei accurately in digitized microscopy images of histology slides. A U-Net based model with customized pixel-wised weighted loss function is adapted in the proposed framework, followed by a U-Net based model with VGG16 backbone and a soft Dice loss function. The model was pretrained on the Post-NAT-BRCA public dataset before training and independent evaluation on the MoNuSeg dataset. The cascaded model could outperform the other state-of-the-art models with an AJI of 0.72 and an F1-score of 0.83 on the MoNuSeg test set.

Keywords: Machine learning / Deep learning approaches, Image analysis and classification - Machine learning / Deep learning approaches
Abstract: Anomaly detection in a medical image is a challenging yet essential task. It relies on learning patterns/distributions from health data only, and no abnormal samples are available during training. This study proposes a novel self-supervised learning method to precisely detect and localize anomalies in MRI medical images. We synthesize abnormal images by overlaying random pseudo-outliers onto normal samples and propose a discriminative model for anomaly detection. Unlike prior arts that generate abnormalities with pre-determined regular geometric shapes, we introduce a new outlier synthesis strategy capable of generating random-shape anomalies. By learning the disentanglement of pseudo-outliers and normal regions in the synthesized images, our model can capture natural anomalies in images at both the pixel level and sample level. We present our empirical experimentation on two publicly accessible datasets and demonstrate the proposed method's superiority over SOTA solutions on MRIs.

Keywords: Machine learning / Deep learning approaches, Image feature extraction, Image analysis and classification - Machine learning / Deep learning approaches
Abstract: Running gait assessment is critical in performance optimization and injury prevention. Traditional approaches to running gait assessment are inhibited by unnatural running environments (e.g., indoor lab), varied assessor (i.e., subjective experience) and high costs with traditional reference standard equipment. Thus, development of valid, reproduceable and low-cost approaches are key. Use of wearables such as inertial measurement units have shown promise but despite their flexible use in any environment and reduced cost, they often retain complexities such as connectivity to mobile platforms and stringent attachment protocols. Here, we propose a non-wearable camera-based approach to running gait assessment, focusing on identification of initial contact events within a runner’s stride. We investigated different artificial intelligence and object tracking approaches to determine the optimal methodology. A cohort of 40 healthy runners were video recorded (240FPS, multi-angle) during 2-minute running bouts on a treadmill. Validation of the proposed approach is obtained from comparison to manually labelled videos. The computing vision approach can accurately identify initial contact events (ICC(2,1) = 0.902).

Keywords: Machine learning / Deep learning approaches, PET and SPECT Imaging applications, Image classification
Abstract: The early and asymptomatic stages of Alzheimer’s Disease (AD), such as mild cognitive impairment (MCI), are hard to classify, even by experienced physicians. Deep learning approaches, such as convolutional neural networks (CNNs), have been shown to help, achieving similar or even better results. Although these methods have the advantage that features are automatically extracted from images rather than handcrafted, they do not allow for incorporating medical knowledge. In this paper we propose curriculum learning (CL) strategies for CNNs designed to diagnose healthy subjects, MCI and AD, as a way to incorporate medical knowledge to boost the performance of the networks for early AD diagnosis. CL is a training strategy of the networks that tries to mimic the way humans, in this case doctors, learn. Several CL strategies were implemented and compared to commonly used baseline methods. The results show that they improve the performance, particularly that of MCI.

Keywords: Brain functional imaging - EEG, Brain functional imaging
Abstract: This study examines the neural activities of participants undergoing vibro-motor reprocessing therapy (VRT) while experiencing motion sickness. We evaluated the efficacy of vibro-motor reprocessing therapy, a novel therapeutic technique based on eye movement desensitization and reprocessing (EMDR), in reducing motion sickness. Based on visually induced motion sickness in two sets of performed sessions, eight participants were exposed to VRT stimulation in a VRT/non-VRT setting. Simultaneously, brain activity changes were recorded using electroencephalography (EEG) at baseline and during stimulus exposure, and comparisons made across the VRT/non-VRT conditions. A significant reduction in the alpha (8-12 Hz) spectral power was observed in the frontal and occipital locations, consistent across all participants. Furthermore, significant reductions were also found in the frontal and occipital delta (0.5-4 Hz) and theta (4-8 Hz) spectral power frequency bands between non-VRT and VRT conditions (p < 0.05). Our results offer novel insights for a potential nonpharmacological treatment and attenuation of motion sickness. Furthermore, symptoms can be observed, and alleviated, in real-time using the reported techniques.

Keywords: Neural signals - Machine learning & Classification, Brain functional imaging - Connectivity and information flow, Neurological disorders
Abstract: The goal of this paper is to use graph theory network measures derived from non-invasive electroencephalography (EEG) to develop neural decoders that can differentiate Parkinson's disease (PD) patients from healthy controls (HC). EEG signals from 27 patients and 27 demographically matched controls from New Mexico were analyzed by estimating their functional networks. Data recorded from the patients during ON and OFF levodopa sessions were included in the analysis for comparison. We used betweenness centrality of estimated functional networks to classify the HC and PD groups. The classifiers were evaluated using leave-one-out cross-validation. We observed that the PD patients (on and off medication) could be distinguished from healthy controls with 89% accuracy -- approximately 4% higher than the state-of-the-art on the same dataset. This work shows that brain network analysis using extracranial resting-state EEG can discover patterns of interactions indicative of PD. This approach can also be extended to other neurological disorders.

Keywords: Brain-computer/machine interface, Neural signal processing, Human performance - Speech
Abstract: Recent studies have shown it is possible to decode and synthesize speech directly using brain activity recorded from implanted electrodes. While this activity has been extensively examined using electrocorticographic (ECoG) recordings from cortical surface grey matter, stereotactic electroencephalography (sEEG) provides comparatively broader coverage and access to deeper brain structures including both grey and white matter. The present study examines the relative and joint contributions of grey and white matter electrodes for speech activity detection in a brain-computer interface.

Keywords: Human performance - Modelling and prediction
Abstract: Previous studies have demonstrated the existence of sex differences in emotion recognition by comparing the performance of same-sex and cross-sex training strategies. However, the EEG properties behind the sex differences have not been fully explored. To fill this research gap, we aim to investigate the sex differences in key frequency bands and channel connections of EEG signals. The single-modality attentive simple graph convolutional network (ASGC) is applied to three datasets SEED, SEED-IV and SEED-V under subject-dependence conditions. The classification rates are 90.86 ± 4.84%, 83.14 ± 8.84% and 78.33 ± 7.83%, respectively. The adjacency matrices learned by ASGC indicate that females and males have similar channel-connection patterns, but the degree of importance of channel connections varies by sex. Additionally, by comparing the classification results of 5 frequency bands, we find that males and females represent similar frequency band characteristics, i.e., high-frequency bands achieve better performance, indicating that these frequency bands are more related to emotion processing. Finally, we conduct the cross-subject experiment using ASGC and find that the same-sex strategy outperforms the cross-sex strategy, which is consistent with previous studies. The results also imply that males may be more suitable for sex generalization. However, this finding needs the support of more samples and advanced algorithms.

Keywords: Brain functional imaging - EEG, Brain functional imaging - Mapping, Neural signal processing
Abstract: Corticokinematic coherence (CKC) is computed between limb kinematics and cortical activity (e.g. MEG, EEG), and it can be used to detect, quantify and localize the cortical processing of proprioceptive afference arising from the body. EEG-based studies on CKC have been limited to lab environments due to bulky, non-portable instrumentations. We recently proposed a wireless and miniaturized EEG acquisition system aimed at enabling EEG studies outside the laboratory. The purpose of this work is to compare the EEG-based CKC values obtained with this device with a conventional wired-EEG acquisition system to validate its use in the quantification of cortical proprioceptive processing. Eleven healthy right-handed participants were recruited (six males, four females, age range: 24–40 yr). A pneumatic-movement actuator was used to evoke right index-finger flexion-extension movement at 3 Hz for 4 min. The task was repeated both with the wireless-EEG and wired-EEG devices using the same 30-channel EEG cap preparation. CKC was computed between the EEG and finger acceleration. CKC peaked at the movement frequency and its harmonics, being statistically significant (p < 0.05) in 8–10 out of 11 participants. No statistically significant differences (p < 0.05) were found in CKC strength between wireless-EEG (range 0.03–0.22) and wired-EEG (0.02–0.33) systems, that showed a good agreement between the recording systems (3 Hz: r = 0.57, p = 0.071, 6 Hz: r = 0.82, p = 0.003). As expected, CKC peaked in sensors above the left primary sensorimotor cortex contralateral to the moved right index finger. As the wired-EEG device, the tested wireless-EEG system has proven feasible to quantify CKC, and thus can be used as a tool to study proprioception in the human neocortex. Thanks to its portability, the wireless-EEG used in this study has the potential to enable the examination of cortical proprioception in more naturalistic conditions outside the laboratory environment.

Keywords: Brain functional imaging - EEG, Neurological disorders - Stroke, Neural signal processing
Abstract: Stroke is a heterogeneous condition that would benefit from valid biomarkers of recovery for research and in the clinic. We evaluated the change in resting state connectivity (RSC) via Electroencephalography (EEG) in motor areas, as well as motor recovery of the affected upper limb, in the subacute phase post-stroke. Fifteen participants who had sustained a subcortical stroke were included in this study. The group made significant gains in upper limb impairment as measured by the Upper Extremity Fugl-Meyer Assessment (UEFMA) from baseline to four months post-stroke (24.78 (SD 5.4)). During this time, there was a significant increase in RSC in the beta band from contralesional M1 to ipsilesional M1. We propose that this change in RSC may have contributed to the motor recovery seen in this group.

Clinical Relevance— This study evaluates resting state connectivity measured via EEG as a neural biomarker of recovery post-stroke. Biomarkers can help clinicians understand the potential for recovery after stroke and thus help them to establish therapy goals and determine treatment plans.

Keywords: Brain functional imaging - fMRI, Brain functional imaging - Classification, Brain functional imaging - Connectivity and information flow
Abstract: Parkinson's Disease (PD) is the second largest neurodegenerative disease. Brain functional connectivity (FC) studies for PD were useful. In this study, we employed a novel brain network construction method, edge functional connectivity (eFC), to explore FC differences between healthy control (HC) subjects and PD patients. The data used in this study included 34 HCs and 47 PDs from Huashan Hospital, Fudan University, China. Resting state functional magnetic resonance imaging (rsfMRI) and clinical information were selected. Firstly, we constructed eFC brain network and calculated network matrix for the HC and PD groups. Then, we compared brain network matrix between eFC and the traditional nodal functional connectivity (nFC) method. Receiver operating characteristic curve (ROC) analysis was applied to validate the efficiency of the eFC brain network. The results showed that both nFC and eFC brain networks could identify significantly different characteristics between the HC and PD groups. Important hubs were mainly concentrated in visual network, sensorimotor network, subcortex and cerebellum. In addition, new hubs in basal ganglia and cerebellum regions were found in eFC. Furthermore, eFC achieved better classification results (AUC=0.985) than nFC (AUC=0.861) in discriminating PD from CN subjects.

Keywords: Neurorehabilitation, Neural signal processing, Brain functional imaging - EEG
Abstract: Action Observation Therapy (AOT) is a rehabilitation method which aims at stimulating motor memory by means of the repetitive observation of motor tasks presented through video-clips. Since sleep seems to have a positive effect on learning processes, it is reasonable to hypothesize that the delivery of AOT immediately before sleep hours could enhance the effects of motor training. The objective of the present work was to test the effect of AOT delivered before the sleep hours in terms of improvements in manual dexterity and changes in cortical activity through Electroencephalography (EEG) on healthy subjects. Specifically, EEG traces acquired on a treatment and on a control group before and after three weeks of training during the execution of a NPHT were analyzed. The spectral analysis of brain signals showed an increased activation of the motor cortex on a subgroup of the treatment subjects. Moreover, a significantly higher involvement of frontal areas was observed in the treatment group.

Keywords: Sensory neuroprostheses - Somatosensory, Brain functional imaging - MEG, Smart neural implants - Neuromuscular stimulation
Abstract: Sensory feedback is a critical component in many human-machine interfaces (e.g., bionic limbs) to provide missing sensations. Specifically, electrotactile stimulation is a popular feedback modality able to evoke configurable sensations by modulating pulse amplitude, duration, and frequency of the applied stimuli. However, these sensations coded by electrotactile parameters are thus far predominantly determined by subjective user reports, which leads to heterogeneous and unstable feedback delivery. Thus, a more objective understanding of the impact that different stimulation parameters induce in the brain, is needed. Analysis of cortical responses to electrotactile afference might be an effective method in this regard. In this study, we used magnetoencephalography (MEG) to investigate the somatosensory evoked fields (SEFs) and equivalent current dipoles (ECDs) locations in nine non-invasive electrotactile stimulation conditions (1.2T, 1.5T, 1.8T) × (1 ms, 10 ms, 100 ms) with fixed 1s interval. T is the subject specific sensory threshold of the left index finger. In all conditions, we observed SEFs peaking at ~ 60 ms in the contralateral primary somatosensory cortex. While the amplitudes of the SEFs around 60 ms followed the increase in the stimulation pulse amplitude, the cortical activations were strongest when the stimulus pulse duration was set to 10 ms. These initial results indicate that the somatosensory cortical activations can provide information on the electrotactile parameters of pulse amplitude and duration, and the prosed methodology might be used for an objective interpretation of different artificial sensory feedback arrangements.

Keywords: Brain-computer/machine interface, Neural signals - Machine learning & Classification
Abstract: Detecting the salient parts of motor-imagery electroencephalogram (MI-EEG) signals can enhance the performance of the brain-computer interface (BCI) system and reduce the computational burden required for processing lengthy MI-EEG signals. In this paper, we propose an unsupervised method based on the self-attention mechanism to detect the salient intervals of MI-EEG signals automatically. Our suggested method can be used as a preprocessing step within any BCI algorithm to enhance its performance. The effectiveness of the suggested method is evaluated on the most widely used BCI algorithm, the common spatial pattern (CSP) algorithm, using dataset 2a from BCI competition IV. The results indicate that the proposed method can effectively prune MI-EEG signals and significantly enhance the performance of the CSP algorithm in terms of classification accuracy.

Keywords: Brain-computer/machine interface, Neural signal processing, Neural signals - Machine learning & Classification
Abstract: Motor Imagery-based Brain Computer Interface (MI-BCI) is a typical active BCI with a main focus on motor intention identification. Hybrid motor imagery (MI) decoding methods that based on multi-modal fusion of Electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS), especially deep learning-based methods, become popular in recent MI-BCI studies. However, the fusion strategy and network design in deep learning-based methods are complex. To solve this problem, we proposed the multi-channel fusion method (MCF) to simplify current fusion methods, and we designed a multi-channel fusion hybrid network (MCFHNet) based on MCF. MCFHNet combines depthwise convolutional layers, channel attention mechanism, and Bidirectional Long Short Term Memory (Bi-LSTM) layers, which enables strong capability of feature extraction in spatial and temporal domain. The comparison between MCFHNet and representative deep learning-based methods was performed on an open EEG-fNIRS dataset. We found the proposed method can yield superior performance (mean accuracy of 99.641% in 5-fold cross validation of an intra-subject experiment). This work provides a new option for multi-modal MI decoding, which can be applied in the rehabilitation field based on hybrid BCI systems.

Keywords: Brain functional imaging - EEG, Brain functional imaging - Source localization, Brain functional imaging - Blind source separation
Abstract: Inaccurate estimation of skull conductivity is the largest impediment to high-resolution EEG source imaging because of its strong influence and wide variability across individuals. Nonetheless, there is yet no widely applied method for noninvasively measuring individual skull conductivity. We presented a skull conductivity and source location estimation algorithm (SCALE) for simultaneously estimating skull conductivity and the cortical distributions of 18-20 effective sources derived from the EEG data by independent component analysis (ICA). SCALE combines a realistic Finite Element Method (FEM) head model built from a magnetic resonance (MR) head image with the effective source scalp maps to estimate brain-to-skull conductivity ratio (BSCR) and to map the effective sources on the cortical surface. To estimate the robustness of SCALE BSCR estimates, we applied SCALE to MR image and high-density EEG data from ten participants, five having data from 2-3 different tasks and sessions. As expected, across participants SCALE BSCR estimates differed widely (mean 32.8, range 18-78). Within-participant SCALE BSCR estimates were far more consistent than between participants. By incorporating SCALE-optimized distributed EEG source localization, stable functional imaging of cortical EEG effective sources can become routine, giving relatively low-cost EEG imaging a spatial resolution compatible with other brain imaging results and uniquely capable for studying brain dynamics supporting thought and action in laboratory, virtual, and natural environments.

Keywords: Surgical robotics, Computer-assisted surgery, New technologies and methodologies in medical robotics
Abstract: The iHandU system is a wearable device that quantitatively evaluates changes in wrist rigidity during Deep Brain Stimulation (DBS) surgery, allowing clinicians to find optimal stimulation settings that reduce patient symptoms. Robotic accuracy is also especially relevant in DBS surgery, as accurate electrode placement is required to increase effectiveness and reduce side effects. The main goal of this work is to integrate the advantages of each system in a closed-loop system between an industrial robot and the iHandU system. For this purpose, a comparative analysis of a Leksell stereotactic frame and neuro-robotic system accuracies was performed using a lab-made phantom. The neuro-robotic system reached 90% of trajectories, while the stereotactic frame reached all trajectories. There are significant differences in accuracy errors between these trajectories (p < 0.0001), which can be explained by the high correlation between the neuro-robotic system errors and the distance from the trajectory to the origin of the Leksell coordinate system (mathbit{rho}= 0.72). Overall accuracy is comparable to existing neuro-robotic systems, achieving a deviation of (1.0 ± 0.5) mm at the target point. The accuracy of DBS electrode positioning and stimulation parameters choice leads to better long-term clinical outcomes in Parkinson’s disease patients. Our neuro-robotic system combines real-time feedback assessment of the patient's symptomatic response and automatic positioning of the DBS electrode in a specific brain area.

Keywords: Haptics in robotic surgery, Tactile displays and perception, Surgical robotics
Abstract: Abstract— Haptic feedback relays important tissue mechanical properties to surgeons during open surgery. However, this information is lost during Robot-assisted Minimally Invasive Surgery (RMIS). Here we present a proofof- concept for a novel instrument-integrated sensor that uses fiber Bragg grating (FBG) arrays to identify tissues based on mechanical properties. Subjects were tasked with sorting tissue phantoms based on hardness. When using a conventional surgical robot, the average error for novices (N=5) and the expert user was 22.5% and 12.5% respectively. This reduced to 2.5% and 0% when sorting with direct palpation by hand. In contrast, the senorized instrument with automated analysis was able to perform the task without any error across all trials. Clinical Relevance— The proposed sensor has the potential of identifying different tissues based on mechanical properties and thus characterize tumors and other relevant structures. It is envisaged that this will improve decision making process during RMIS and also provide useful sensory information for autonomous surgery.

Keywords: Computer-assisted surgery, Image guided surgery
Abstract: In image-guided surgery, endoscope tracking and surgical scene reconstruction are critical, yet equally challenging tasks. We present a hybrid visual odometry and reconstruction framework for stereo endoscopy that leverages unsupervised learning-based and traditional optical flow methods to enable concurrent endoscope tracking and dense scene reconstruction. More specifically, to reconstruct texture-less tissue surfaces, we use an unsupervised learning-based optical flow method to estimate dense depth maps from stereo images. Robust 3D landmarks are selected from the dense depth maps and tracked via the Kanade-Lucas-Tomasi tracking algorithm. The hybrid visual odometry also benefits from traditional visual odometry modules, such as keyframe insertion and local bundle adjustment. We evaluate the proposed framework on endoscopic video sequences openly available via the SCARED dataset against both ground truth data, as well as two other state-of-the-art methods - ORB-SLAM2 and Endo-depth. Our proposed method achieved comparable results in terms of both RMS Absolute Trajectory Error and Cloud-to-Mesh RMS Error, suggesting its potential to enable accurate endoscope tracking and scene reconstruction.

Keywords: Surgical robotics, Human machine interfaces and robotics applications, New technologies and methodologies in medical robotics
Abstract: There is a large community of people with hand disabilities, and these disabilities can be a barrier to those looking to retain or pursue surgical careers. With the development of surgical robotics technologies, it may be possible to develop user interfaces to accommodate these individuals. This paper proposes a hand-free control method for the gripper of a patient side manipulator (PSM) in the da Vinci surgical system. Using electromyography (EMG) signals, a proportional control method was tested on its ability to grasp a pressure sensor. These preliminary results demonstrate that the user can reliably control the grasping motion of the da Vinci PSM using this system. There is a strong correlation between grasping force and normalized EMG signal (r= 0.874). Moreover, the gripper can generate a step grasping force output when feeding in a generated step signal. The results in this paper demonstrate the system integration of a research EMG system with the da Vinci surgical system and are a step towards developing accessible teleoperation systems for surgeons with disabilities. Hand-free control for remaining degrees of freedom in the PSM is under development using additional input from motion capture system.

Keywords: Surgical robotics, Humanoid robotics, Robotics - Orthotics and Exoskeletons
Abstract: In robotic urological surgery, it is common for an assistant to use laparoscopic forceps to move organs clear of the surgical working space; however, the assistance efficiency is low. In this study, we aimed to develop a three-fingered humanoid hand with multiple degrees of freedom and a folding mechanism that would allow it to be inserted through a small incision to improve the efficiency of assisting with organs. The bladder (with prostate) and kidneys were selected as the target organs. To achieve stable assistance for these organs, we analyzed three postures: "grasp," "open palm," and "pinch." We verified that the proposed hand can be inserted into the abdominal cavity through a 20 mm port and can assist in a grasping an object of the same size as the target organ with these three movements.

Keywords: Surgical robotics
Abstract: Minimally invasive instruments are inserted percutaneously and are steered toward the desired anatomy. The low stiffness of instruments is an advantage; however, once the target is reached, the instrument usually is required to transmit force to the environment. The main limitation of the constant stiffness is predetermined maneuverability and cap of force transmission. Whereas a highly flexible device can be safely steered through the body but is not suitable for payload limit, while a highly stiff device can have relatively high loads but cannot be steered in highly tortuous trajectories. To overcome this limitation, an adaptive stiffness soft robot was proposed, and the effects of the chamber pressure on the stiffness of the soft robot were investigated. To this end, a single-chamber pneumatic soft robot with one tendon was designed and fabricated. Afterward, a continuum mechanics model based on the nonlinear Cosserat rod model with hyperelastic material model and large deformation kinematics of the robot was developed. The shooting method solved the model as a boundary value problem with Dirichlet and Neumann boundary conditions. The results of the model showed stiffness adaptation feasibility with simultaneous tendon-driving and pneumatic actuation. Thus, to validate the theoretical findings, a series of experimental studies were performed with pressure in the range of 33 to 44 kPa and tendon tensions in the range of 0 to 2.7 N. The theoretical and experimental results for tip displacement and stiffness showed similar trends with a maximum error of 8.25%.

Keywords: Implantable technologies, Implantable sensors - biocompatibility, Implantable sensors
Abstract: IMDs are typically considered for chronic-use applications and a limited set of implant locations. Resorbable IMDs seek to combine advances in flexible electronics with functional soft materials to enable new applications, including acute care, aiming at temporary interfacing with soft tissues. Poly(octamethylene maleate (anhydride) citrate) (POMaC) is an elastomer with demonstrated high biocompatibility and bioresorbability, as well as tunable stiffness and surface properties. Despite its promises, POMaC has not yet been applied in engineering flexible electronics. Herein, a POMaC-based circuit board is demonstrated and characterized. The monomer composition and thermal degradation properties of the pre-polymer was characterized. POMaC-based circuit boards were constructed using traditional microfabrication methods, including spin coating and metallization. POMaC pre-polymer and films were thermally stable to 300°C, exhibit controlled degradation in simulated physiological conditions, and are cytocompatible. Deposited traces were stable during fabrication and processing, and an LED circuit was designed and fabricated using surface mount devices on a POMaC-circuit board. The results indicate the feasibility of POMaC-based circuit boards for use in resorbable IMDs. Future work will investigate more complex circuits, fully encapsulated devices, and mechanical characterization.

Keywords: Sensor systems and Instrumentation, Implantable sensors, Optical and photonic sensors and systems
Abstract: A readout device for a dual-functional neural observation system is presented. The authors separately developed the reading operation of an implantable CMOS image sensor and a setup for fast-scan cyclic voltammetry and implemented them together in a microcontroller-based device. The developed imaging readout device with a size of 3.0 × 5.5 cm2 can reach the highest reading rate of 160 fps with a 120 × 268 pixel image sensor. The voltammetry function was verified through an experiment using commercial carbon fiber electrodes in phosphate-buffered saline. When the imaging is sequentially operated with 400 V/s-scan rate voltammetry from -0.4 to 1.3 V, the system can operate at up to 60 fps. With this system, calcium imaging and dopamine recording in a freely behaving mouse can be achieved together in a simpler manner. This study aims to be the basis for the development of an implantable multi-functional sensor.

Keywords: Wearable body-compliant, flexible and printed electronics, Implantable sensors, Physiological monitoring - Instrumentation
Abstract: Bioelectrical slow waves are fundamental to maintaining the normal motility of the gastrointestinal tract. Slow wave abnormalities are associated with several major digestive disorders. High-resolution electrical mapping arrays have been used to investigate pathological slow wave abnormalities. However, conventional electrode substrate materials are opaque with high mechanical modulus, which leads to non-compliance and sub-par contact with the organ, without additional manipulations. Here we developed highly conformal and transparent conducting polymer electrode arrays using the extrusion wet-printing technique. The performance of electrodes for the electrophysiological recording of the gastric slow wave was validated using in a pig model, against a previously validated reference array over 100 s recording window. The conducting polymer electrodes registered comparable frequency to the reference array (3.31 ± 0.20 cpm vs. 3.27 ± 0.07 cpm, p = 0.067), with lower amplitude (372 ± 237 vs. 586 ± 291 μV, p < 0.001), and signal to noise ratio (10.92 ± 7.83 vs. 17.40 ± 8.01 dB, p < 0.001). Further adjustments to the deposition parameters and contact material will improve the performance of the conducting polymer array for future experimental applications.

Keywords: Magnetic sensors and systems, Implantable sensors
Abstract: This paper studies the use of electromagnetic induction in localization of wireless capsule endoscopes (WECs). There is still currently a need for an accurate localization system to enable localizing possible findings in the gastrointestinal tract, and to develop an active steering system for the capsule. Developing an optimal localization system requires the sensitivity of the system to be analyzed. In this paper, three different coil geometries are modelled with a computer simulation platform, and their sensitivities and target responses are compared. In order to do that, a formulation for the sensitivity based on the dipole model approximation is presented. The first coil array is based on literature and is used as a reference. The second array presents how having more transmit-receive channels in the array effects the sensitivity. The third coil array simulates the effect of increasing the field excitation intensity in different directions by using a three-axial Helmholtz array. In addition, both proposed coil arrays utilize larger coils than the reference. As a result, it seems that both increasing the coil size and the number of field projections interrogating the target increase the overall sensitivity in the region of interest and the target response. The findings suggest that an optimal coil array could utilize both large coils and multiple transmit-receive channels to increase the number of independent fields incident onto the target.

Keywords: Implantable systems, Implantable technologies, Implantable sensors - biocompatibility
Abstract: Cyclic Olefin Copolymer is emerging as a packaging material for implantable electrodes due to its physical properties such as low water absorption rate and low water vapor permeability. The electrode-tissue interface is often regarded as a major focus of implantable electrodes, but its packaging should also be considered thoroughly since it directly contacts the adjoining body cells. Therefore, eliminating any sharp boundaries or edges around the package would be beneficial to minimize potential inflammatory responses caused by physical/mechanical stresses. To smooth both inner/outer edges of a cyclic olefin copolymer packaging, an optimal UV laser condition was investigated by varying its marking speed and iterations.

Keywords: Implantable sensors, Bio-electric sensors - Sensing methods
Abstract: Implanted electrodes, such as those used for cochlear implants, brain-computer interfaces, and prosthetic limbs, rely on particular electrical conditions for optimal operation. Measurements of electrical impedance can be a diagnostic tool to monitor implanted electrodes for changing conditions arising from glial scarring, encapsulation, and shorted or broken wires. Such measurements provide information about the electrical impedance between a single electrode and its electrical reference, but offer no insights into the overall network of impedances between electrodes. Other solutions generally rely on geometrical assumptions of the arrangement of the electrodes and may not generalize to other electrode networks. Here, we propose a linear algebra-based approach, Cross-Channel Impedance Measurement (CCIM), for measuring a network of impedances between electrodes which all share a common electrical reference. This is accomplished by measuring the voltage response from all electrodes to a known current applied between each electrode and the shared reference, and is agnostic to the number and arrangement of electrodes. The approach is validated using a simulated 8-electrode network, demonstrating direct impedance measurements between electrodes and the reference with 96.6% ± 0.2% accuracy, and cross-channel impedance measurements with 93.3% ± 0.6% accuracy in a typical system. Subsequent analyses on randomized systems demonstrate the sensitivity of the model to impedance range and measurement noise.

Keywords: Data mining and big data methods - Biosignal classification, Data mining and big data methods - Pattern recognition, Nonlinear dynamic analysis - Biomedical signals
Abstract: This paper presents a novel approximation unit added to the conventional spike processing chain which provides an appreciable reduction of complexity of the high-hardware cost feature extractors. The use of the Taylor polynomial is proposed and modelled employing its cascaded derivatives to non-uniformly capture the essential samples in each spike for reliable feature extraction and sorting. Inclusion of the approximation unit can provide 3X compression (i.e. from 66 to 22 samples) to the spike waveforms while preserving their shapes. Detailed spike waveform sequences based on in-vivo measurements have been generated using a customized neural simulator for performance assessment of the approximation unit tested on six published feature extractors. For noise levels σ_N between 0.05 and 0.3 and groups of 3 spikes in each channel, all the feature extractors provide almost same sorting performance before and after approximation. The overall implementation cost when including the approximation unit and feature extraction shows a large reduction (i.e. up to 8.7X) in the hardware costly and more accurate feature extractors, offering a substantial improvement in feature extraction design.

Keywords: Time-frequency and time-scale analysis - Wavelets, Signal pattern classification, Time-frequency and time-scale analysis - Time-frequency analysis
Abstract: High-Frequency Oscillation (HFO) is a promising biomarker of the epileptogenic zone. However, sharp artifacts might easily pass the conventional HFO detectors as real HFOs and reduce the seizure onset zone (SOZ) localization. We hypothesize that, unlike pseudo-HFOs, which originates from artifacts with sharp changes or arbitrary waveform characteristic, real HFOs could be represented by a limited number of oscillatory waveforms. Accordingly, to distinguish true ones from pseudo-HFOs, we established a new classification method based on sparse representation of candidate events that passed an initial detector with high sensitivity but low specificity. Specifically, using the Orthogonal Matching Pursuit (OMP) and a redundant Gabor dictionary, each event was represented sparsely in an iterative fashion. The approximation error was estimated over 30 iterations which were concatenated to form a 30-dimensional feature vector and fed to a random forest classifier. Based on the selected dictionary elements, our method can further classify HFOs into Ripples (R) and Fast Ripples (FR). In this scheme, two experts visually inspected 2075 events captured in iEEG recordings from 5 different subjects and labeled them as true-HFO or Pseudo-HFO. We reached 90.22% classification accuracy in labeled events and a 21.16% SOZ localization improvement compared to the conventional amplitude-threshold-based detector. Our sparse representation framework also classified the detected HFOs into R and FR subcategories. We reached 91.24% SOZ accuracy with the detected R+FR events.

Keywords: Physiological systems modeling - Closed loop systems, Physiological systems modeling - Signal processing in physiological systems, Physiological systems modeling - Signal processing in simulation
Abstract: Cortical mapping is widely employed to define the sensorimotor area and delineate the central sulcus (CS) during awake craniotomies. The approach involves the gold standard somatosensory evoked potentials (SSEPs) recorded with electrocorticogram (ECoG) strip electrodes. However, the evoked response can be misconstrued from the manual peak interpretation due to the poor spatial resolution of the strip electrode or when the electrode does not precisely cover the desired cortical area. This can lead to unintentional damage to the eloquent cortex. We present a soft real-time computer based visualization system that uses recorded SSEPs with a subdural grid to aid in cortical mapping. The neural data during electrical stimulation of the median nerve at 0.6Hz are picked up with a bio-amplifier. The stimulation artifact recorded from the bipolar electromyogram (EMG) is used as the stimulation onset. The ECoG data are assessed online with MATLAB Simulink to process and visualize the SSEPs waveform. The visualization system is programmed to display the SSEPs peak activation as a heat map on a 2D grid and projected onto a screen, showcasing the nature of the cortical activities over the contact surface area. Since the grid occupies a large cortical surface, the heatmap is able to delineate the central sulcus. The map can be viewed at any time point along the SSEP trace without the need for peak interpretation. With the goal to provide additional information during cortical mapping and facilitate interpretation of ECoG grid data, we believe that this visualization system will aid in rapid definition of the sensorimotor area during surgical planning.

Keywords: Physiological systems modeling - Signal processing in physiological systems, Data mining and big data methods - Biosignal classification
Abstract: Abstract— Approximately 30% of patients with epilepsy do not respond to anti-epileptogenic drugs. Surgical removal of the epileptogenic zone (EZ), the brain regions where the seizures originate and spread, can be a possible therapy for these patients, but localizing the EZ is challenging due to a variety of clinical factors. High-frequency oscillations (HFOs) in intracranial electroencephalography (EEG) are a promising biomarker of the EZ, but it is currently unknown whether HFO rates and HFO morphology modulate as pathological brain networks evolve in a way that gives rise to seizures. To address this question, we assessed the temporal evolution of the duration of HFO events, amplitude of HFO events, and rates of HFOs per minute. HFO events were quantified using the 4AP in vivo rodent model of epilepsy, inducing seizures in two different brain areas. We found that the duration and amplitude of HFO events were significantly increased for the cortex model when compared to the hippocampus model. Additionally, the duration and amplitude increased significantly between baseline and pre-ictal HFOs in both models. On the other hand, the two models did not display a consistent increasing or decreasing trend in amplitude, duration or rate when comparing ictal and postictal intervals.

Clinical Relevance— We assessed the amplitude, duration, and rate of HFOs in two acute in vivo rodent models of epilepsy. The significant modulation of HFO morphology from baseline to pre-ictal periods suggests that these features may be a robust biomarker for pathological tissue involved in epileptogenesis. Moreover, the differences in HFO morphology observed between cortex and hippocampus animal models possibly indicate that different structural network characteristics of the EZ cause this modulation. In all, we found that HFO features modulate significantly with the onset of seizures, further highlighting the need to consider of HFO morphology in EZ-localizing studies.

Keywords: Signal pattern classification, Principal component analysis, Multivariate methods
Abstract: While the presence of spreading depolarization (SD) and associated spreading depression have been well studied and known to be associated with post-ischemic brain damage, the spatiotemporal spread of these events from the site of injury is not well understood. With the new development of high-density micro-electrocorticographic (ECoG) electrode arrays, monitoring the spread of the depolarizing events and associated depression is possible. The goal of this work is to define the electrocorticographic features of spreading depolarization and associated depression across the multichannel array and search for patterns in these features that emerge across both space and time. We present the spatial distribution of features found from chronic ECoG recordings acquired from awake behaving rats induced with a rodent model of stroke. SD events were detected with an unsupervised algorithm that searched for a stereotyped pattern in the first derivative of the ECoG. The algorithm yielded a 83% correct detection rate on average across 4 rats, and a 5% false positive rate. We defined key electrophysiological features onto the physical brain regions using MATLAB, such as the peak-to-peak amplitude of each SD event, the width (or duration) of the SD event, DC level, and average rate of decline in the signal baseline. We performed k-means clustering to the activity in this feature space which yielded three contiguous regions in physical space. The elbow optimization method was applied to a distortion metric and indicated that 3 clusters was optimal. These findings motivate us to conduct future studies that would verify whether these 3 clusters in electrode-space correspond to immunohistochemically defined regions of tissue health, namely, infarct, penumbra, and healthy tissue.

Keywords: Time-frequency and time-scale analysis - Time-frequency analysis
Abstract: Adaptive Deep Brain Stimulation (aDBS) has been proposed in literature to avoid the negative consequences associated with the continuous stimulation delivered through traditional deep brain stimulation. This work seeks to determine a group of neural biomarkers that a classification algorithm could use on an aDBS device using rodent animal models. The neural activities were acquired from the primary motor cortex of four Parkinsonian model rats and four healthy rats from a control group. To overcome the variability introduced from the small rat sample size, this work proposes a novel method for combining and running Genetic Feature Selection and Forward Stepwise Feature Selection in an environment where classification accuracy varies greatly based on how the folds are organized before cross-validation. Three separate classification algorithms, Logistic Regression, k-Nearest Neighbor, and Random Forest are used to verify the proposed method. For Logistic Regression, the set of Alpha Power (7-12 Hz), High Beta Power (20-30 Hz), and 55-95 Hz Gamma Power shows the best performance in classification. For k-Nearest Neighbor, the characterizing features are Low Beta Power (12-20 Hz), High Beta Power, All Beta Power (12-30 Hz), 55-95 Hz Gamma Power, and 95-105 Hz Gamma Power. For Random Forest, they are High Beta Power, All Beta Power, 55-95 Hz Gamma Power, 95-105 Hz Gamma Power, and 300-350 Hz High-Frequency Oscillations Power. With the selected feature set, experimental results show an increasing classification accuracy from 59.08% to 77.69% for Logistic Regression, from 49.53% to 73.44% for k-Nearest Neighbor, and from 54.10% to 71.15% for Random Forest.

Keywords: Wearable wireless sensors, motes and systems
Abstract: Existing approaches that assess and monitor the severity of Parkinson's Disease (PD) focus on the integration of wearable devices based on inertial sensors (accelerometers, gyroscopes) and electromyographic (EMG) transducers. Nevertheless, some of these sensors are bulky and lack comfortability. This manuscript presents triboelectric nanogenerators (TENGs) as an alternative stretchable sensor solution enabling PD monitoring systems. The prototype has been developed using a triboelectric sensor based on Ecoflex™ and PEDOT:PSS that is placed on the forearm. The movement of the skin above the forearm muscles and tendons correlates with the extension and flexion of fingers and hands. This way, the small gap of 0.5 cm between the polymer layers is displaced, generating voltage due to the triboelectric contact. Signals from preliminary experiments can discriminate different dynamics of emulated tremor and bradykinesia in hands and fingers. A modified version of the TS is integrated with a printed circuit board (PCB) in a single package with signal conditioning and wireless data transmission. The sensor platforms have demonstrated a good sensitivity to PD symptoms like bradykinesia and tremor based on the Unified Parkinson's Disease Rating Scale (MDS:UPDRS).

Keywords: Physiological monitoring - Novel methods, Physiological monitoring - Modeling and analysis, Wearable sensor systems - User centered design and applications
Abstract: The feasibility of using in-ear SpO2 to track cognitive workload induced by gaming is investigated. This is achieved by examining temporal variations in cognitive workload through the game Geometry Dash, with 250 trials across 7 subjects. The relationship between performance and cognitive load in Dark Souls III boss fights is also investigated followed by a comparison of the cognitive workload responses across three different genres of game. A robust decrease in in-ear SpO2 is observed in response to cognitive workload induced by gaming, which is consistent with existing results from memory tasks. The results tentatively suggest that in-ear SpO2 may be able to distinguish cognitive workload alone, whereas heart rate and breathing rate respond similarly to both cognitive workload and stress. This study demonstrates the feasibility of low cost wearable cognitive workload tracking in gaming with in-ear SpO2, with applications to the play testing of games and biofeedback in games of the future.

Keywords: Novel methods, Sensor systems and Instrumentation, Wearable low power, wireless sensing methods
Abstract: Cumulative screen exposure has been increased due to the explosion of digital technology ownership in the past decade for all people, including children who face exposure related risks such as obesity, eye problems and disrupted sleep. Screen exposure is linked to physical and mental health risks among both children and adults. Current methods of screen exposure assessment have their limitations, mostly in the prospective of objectiveness, robustness, and invasiveness. In this paper, we propose a novel method to measure screen exposure time using a wearable sensor and computer vision technology. We use a customized, lightweight, wearable senor to capture egocentric images and use deep learning-based object detection module to identify the existence of electronic screens. The duration of screen exposure is further estimated using post-processing technology to filter consecutive frames regarding to the screen usage. Our method is non-invasive and robust, providing an objective and accurate means to screen exposure measurement. We conduct experiments on various environments to identify the existence of three types of screens and duration of screen exposure. The experimental results demonstrate the feasibility of automatically assessing screen time exposure and great potential to be applied in large scale experiments for behavioral study.

Keywords: IoT sensors for health monitoring, Modeling and analysis, Sensor systems and Instrumentation
Abstract: This study aims to identify a set of indicators to estimate cognitive workload using a multimodal sensing approach and machine learning. A set of three cognitive tests were conducted to induce cognitive workload in twelve participants at two levels of task difficulty (Easy and Hard). Four sensors were used to measure the participants’ physiological change, including, Electrocardiogram (ECG), Electrodermal Activity (EDA), Respiration (RESP), and Blood Oxygen Saturation (SpO2). To understand the perceived cognitive workload, NASA-TLX was used after each test and analysed using Chi-Square test. Three well-know classifiers (LDA, SVM, and DT) were trained and tested independently using the physiological data. The statistical analysis showed that participants’ perceived cognitive workload was significantly different (p < 0.001) between the tests, which demonstrated the validity of the experimental conditions to induce different cognitive levels. Classification results showed that a fusion of ECG and EDA presented good discriminating power (acc = 0.74) for cognitive workload detection. This study provides preliminary results in the identification of a possible set of indicators of cognitive workload. Future work needs to be carried out to validate the indicators using more realistic scenarios and with a larger population.

Keywords: Wearable body sensor networks and telemetric systems, Wearable wireless sensors, motes and systems, Sensor systems and Instrumentation
Abstract: Cerebellar ataxia (CA) refers to the incoordination of movements of the eyes, speech, trunk, and limbs caused by cerebellar dysfunction. Conventional machine learning (ML) utilizes centralised databases to train a model of diagnosing CA. Despite the high accuracy, these approaches raise privacy concern as participants’ data revealed in the data centre. Federated learning is an effective distributed solution to exchange only the ML model weight rather than the raw data. However, FL is also vulnerable to network attacks from malicious devices. In this study, we depict the concept of blockchained FL with individual’s validators. We simulate the proposed approach with real-world dataset collected from kinematic sensors of CA individuals with four geographically separated clinics. Experimental results show the blockchained FL maintains competitive accuracy of 89.30%, while preserving both privacy and security.

Keywords: Health Informatics - Computer-aided decision making, Health Informatics - Decision support methods and systems, General and theoretical informatics - Decision support systems
Abstract: Dementia with Lewy Bodies (DLB) is the second most common form of dementia, but diagnostic markers for DLB can be expensive and inaccessible, and many cases of DLB are undiagnosed. This work applies machine learning techniques to determine the feasibility of distinguishing DLB from Alzheimer’s Disease (AD) using heterogeneous data features. The Repeated Incremental Pruning to Produce Error Reduction (RIPPER) algorithm was first applied using a Leave-One-Out Cross-Validation protocol to a dataset comprising DLB and AD cases. Then, interpretable association rule-based diagnostic classifiers were obtained for distinguishing DLB from AD. The various diagnostic classifiers generated by this process had high accuracy over the whole dataset (mean accuracy of 94%). The mean accuracy in classifying their out-of-sample case was 80.5%. Every classifier generated consisted of very simple structure, each using 1-2 classification rules and 1-3 data features. As a group, the classifiers were heterogeneous and used several different data features. In particular, some of the classifiers used very simple and inexpensive diagnostic features, yet with high diagnostic accuracy. This work suggests that opportunities may exist for incorporating accessible diagnostic assessments while improving diagnostic rate for DLB.

Keywords: Health Informatics - Computer-aided decision making, Bioinformatics - Bioinformatics for health monitoring, General and theoretical informatics - Artificial Intelligence
Abstract: Heart sound auscultation is an effective method for early-stage diagnosis of heart disease. The application of deep neural networks is gaining increasing attention in automated heart sound classification. This paper proposes deep Convolutional Neural Networks (CNNs) to classify normal/abnormal heart sound, which takes two-dimensional Mel-scale features as input, including Mel frequency cepstral coefficients (MFCCs) and the Log Mel spectrum. We employ two weighted loss functions during the training to mitigate the class imbalance issue. The model was developed on the public PhysioNet/Computing in Cardiology Challenge (CinC) 2016 heart sound database. On the considered test set, the proposed model with Log Mel spectrum as features achieves an Unweighted Average Recall (UAR) of 89.6%, with sensitivity and specificity being 89.5% and 89.7% respectively.

Keywords: Health Informatics - Computer-aided decision making, Health Informatics - Decision support methods and systems, Health Informatics - Outcome research
Abstract: Glioma, characterized by neoplastic growth in the brain, is a life-threatening condition that, in most cases, ultimately leads to death. Typical analysis of glioma development involves observation of brain tissue in the form of a histology slide under a microscope. Although brain histology images have much potential for predicting patient outcomes such as overall survival (OS), they are rarely used as the sole predictors due challenges presented by unique characteristics of brain tissue histology. However, utilizing histology in predicting overall survival can be useful for treatment and quality-of-life for patients with early-stage glioma. In this study, we investigate the use of deep learning models on histology slides combined with simple descriptor data (age and glioma subtype) as a predictor of (OS) in patients with low-grade glioma (LGG). Using novel clinical data, we show that models which are more attentive to discriminative features of the image will confer better predictions than generic models. Additionally, we show that adding age and subtype information to a histology image-based model may provide greater robustness in the model than using the image alone, while a model based on image and age but not subtype may confer the best predictive results.

Keywords: Health Informatics - Computer-aided decision making, Health Informatics - Decision support methods and systems
Abstract: This work proposes a method utilizing only the submentalis EMG channel for the classification of sleep and wake states among the healthy individuals and patients with various sleep disorders such as sleep apnea hypopnea syndrome, dyssomnia, etc. We extracted autoregressive model parameters, discrete wavelet transform coefficients, Hjorth's complexity and mobility, relative bandpowers, Poincare plot descriptors and statistical features from the EMG signal. We also used the energy of each epoch as a feature to distinguish between the sleep and wake states. Mutual information based feature selection approach was considered to obtain the top 25 features which provided maximum accuracy. For classification, we employed an ensemble of decision trees with random undersampling and boosting technique to deal with the class-imbalance problem in the sleep data. We achieved an overall accuracy of about 85% for the healthy population and about 70% on an average across different pathological groups. This work shows the potential of EMG chin activity for sleep analysis.

Keywords: Health Informatics - Decision support methods and systems, Health Informatics - Computer-aided decision making, Health Informatics - Behavioral health informatics
Abstract: As the number of elderly people suffering from depression increases today, new techniques for active monitoring of depression are in need than ever. Hence this study aimed to propose an approach of identifying depression in the elderly using gait accelerometry and a machine learning algorithm. A total of 45 community-dwelling elderly individuals participated in the study. Twenty-two out of 45 participants were patients with depression and the remaining 23 participants were individuals without depression. The participants completed a 7-meter walking twice at their preferred speeds with an accelerometer on their lower back. The anterior-posterior acceleration signals measured at the lower back while walking were segmented into acceleration falling and rising phases. Then eight descriptive statistical and six morphological parameters were extracted from each phase. The extracted parameters were ordered chronologically and used as a gait sequence feature. The 4-fold cross-validation of the bidirectional long short-term memory network-based classifiers that used the gait sequence feature as input showed an average accuracy of 0.956 in classifying the elderly with depression and those without depression. The study is expected to serve as a milestone exploring the use of gait accelerometry in assessing various health conditions in the future.

Keywords: Health Informatics - Decision support methods and systems, Health Informatics - Outcome research, General and theoretical informatics - Supervised learning method
Abstract: The state of the art is still lacking an extensive analysis of which clinical characteristics are leading to better outcomes after robot-assisted rehabilitation on post-stroke patients. Prognostic machine learning-based models could promote the identification of predictive factors and be exploited as Clinical Decision Support Systems (CDSS). For this reason, the aim of this work was to set the first steps toward the development of a CDSS, by the development of machine learning models for the functional outcome prediction of post-stroke patients after upper-limb robotic rehabilitation. Four different linear algorithms were trained and cross-validated using a nested 5x10-fold cross-validation. The performances of each model on the test set were provided through the Median Average Error (MAE) and interquartile range. Additionally, interpretability analyses were performed, to evaluate the contribution of the features to the prediction. The results on the two best performing models showed a MAE of 13.645 [13.357] and 13.343 [14.833] on the Modified Barthel Index score (MBI). The interpretability analyses highlighted the Fugl-Meyer Assessment, MBI, and age as the most relevant features for the prediction of the outcome. This work showed promising results in terms of outcome prognosis after robot-assisted treatment. Further research should be planned for the development, validation and translation into clinical practice of CDSS in rehabilitation.

Keywords: Computational modeling - Analysis of high-throughput systems biology data
Abstract: An anatomically based 3D computational model of the rat stomach was developed using experimental muscle thickness measurements and muscle fiber orientations for the longitudinal muscle (LM) and circular muscle (CM) layers. First, 15 data points corresponding to the measurements were registered on the dorsal and ventral faces of the serosal surface of an averaged 3D rat stomach model. A thickness field representing the varying wall thickness was fitted to the surface and nodal points were projected outwards (for the LM layer) and inwards (for the CM layer) to create 2 new surfaces. In addition, a computational volume mesh was created and fiber orientation in each tetrahedral element was computed using a Laplace-Dirichlet rule-based algorithm and a simulation was performed to validate the model. The stomach model successfully represented the experimental measurements with a thickness in the range of 11.7 – 52.9 µm and 40.6 – 276.5 µm in the LM and CM layers, respectively, while the variation across the stomach was in agreement with the reported values. Similarly, the generated fiber orientations matched with the investigated fiber data and successfully resembled the observed properties such as the hairpin-like structure formed by the LM fibers in the fundus. Bioelectrical simulation using the developed model was successfully converged and reflected the properties of normal antegrade activity. In conclusion, a 3D computational model of the rat stomach was successfully developed and tested for in-silico studies. The model will be used in future studies to assess parameters in electrical therapies and to investigate the structure-function relationship in gastric motility.

Keywords: Models of organ physiology, Data-driven modeling
Abstract: Smooth muscle is found extensively in the human body, including in blood vessels, airways, the gastrointestinal tract, and the urinary bladder. Although the contractile proteins of smooth muscle are very similar to those of striated muscle, smooth muscle's contractile mechanism has not been studied as extensively as those for cardiac and skeletal muscle. Previous studies developed a lumped model of muscle contraction and applied it to cardiac muscle and to skeletal muscle. In this study, this model is used to quantitatively describe the contractile properties of canine smooth muscle, using data from the literature. Results show that a single equation relating muscle force to muscle length and time, and a single set of model parameters, is able to describe smooth muscle's passive and active isometric forces, isometric twitch contractions, isotonic contractions, and an inverse force-velocity relation. The latter arises from the model without assumption of a particular force-velocity curve embodied as a contractile element. This new constitutive relation may be used to describe smooth muscle within larger physiological models, for instance to describe blood vessel constriction or urinary bladder function.

Keywords: Models of organ physiology, Models of organs and medical devices - Inverse problems in biology, Organ modeling
Abstract: Despite the numerous available clinical investigation tests, the associated alteration of quality of life and the socio-economic cost, it remains difficult for physicians to identify the pathophysiological origins of defecation disorders and therefore to provide the appropriate clinical care.

Based on standardized dynamic X-ray defecography, we developed a 2D patient-specific computational fluid dynamic model of rectal evacuation. X-ray defecography was carried out in a sitting position with a standardized paste whose yield stress matched that of soft human feces. The flow was simulated with lattice-Boltzmann methods for yield stress fluids and moving boundary conditions.

The model was applied for a patient with a normal recto-anal function. We deduced from the flow field that the main flow resistance during the defecation was due to the extrusion of the paste through the anal canal. We calculated also from pressure and stress fields the spatio-temporal evolution of the wall normal stress. This latter highlighted a gradient from the proximal to the distal part of the rectum.

We discussed how this new set of hydrodynamical and biomechanical parameters could be interpreted to gain new insights on the physiology of defecation and to diagnose underlying evacuation disorders.

Clinical relevance— If confirmed, our approach should allow clinicians to obtain other parameters from a classic clinical examination and thus better adapt the response of clinicians to the defecation disorders observed in patients.

Keywords: Models of medical devices
Abstract: Deep brain stimulation is currently the only technique used in the clinical setting to modulate the neural activity of deep brain nuclei. Recently, low-intensity transcranial focused ultrasound (LIFU) has been shown to reversibly modulate brain activity through a transcranial pathway. Transcranial LIFU requires a low-frequency ultrasound of around 0.5 MHz due to skull attenuation, thus providing poor axial and lateral resolution. This paper proposes a new conceptual device that would use a stent to place a highfrequency ultrasound array within the brain vasculature to achieve high axial and lateral spatial resolution. The first part of this work identified the most commonly treated deep brain nuclei and examined the human brain vasculature for stent placement. Next, a finite element analysis was carried out using a piezoelectric array that follows the blood vessels curvature, and its ability to focus ultrasound waves in clinically relevant brain nuclei was evaluated. The analytical solution provided promising results for deep brain stimulation via a stent with ultrasound transducers for high spatial resolution neuromodulation.

Keywords: Models of medical devices, Computational modeling - Analysis of high-throughput systems biology data
Abstract: Bioresorbable Vascular Scaffolds (BVS), developed to allow drug deliver and mechanical support, followed by complete resorption, have revolutionized atherosclerosis treatment. InSilc is a Cloud platform for in silico clinical trials (ISCT) used in the design, development and evaluation pipeline of stents. The platform integrates beyond the state-of-the-art multi-disciplinary and multiscale models, which predict the scaffold’s performance in the short/acute and medium/long term. In this study, a use case scenario of two Bioabsorbable Vascular Stents (BVSs) implanted in the same arterial anatomy is presented, allowing the whole InSilc in silico pipeline to be applied and predict how the different aspects of this intervention affect the success of stenting process.

Keywords: Models of medical devices, Models of organ physiology, Organ modeling
Abstract: Microwave ablation (MWA) is a clinically widespread minimally invasive treatment method for lung tumors. Preoperative planning plays a vital role in MWA therapy. However, previous planning methods are far from satisfactory in clinical practice because they only one-sidedly consider the surgical path or energy parameters of an MWA surgery. In this paper, we propose a novel planning model with a computational model of thermal damage to integrally optimize both the surgical path and energy parameters. To ensure the model can be solved in a reasonable time, we elaborate a search space reducing strategy based on clinical constraints. Simulation and ex vivo experimental results were compared with an average mean absolute error of 0.82 K and an average root mean square error of 1.01 K. Our planning model was evaluated on clinical data, and the experimental results demonstrate the effectiveness of our model.

Keywords: Cardiovascular and respiratory signal processing - Cardiovascular signal processing
Abstract: The analysis of the fetal electrocardiogram (ECG) requires to remove the mother ECG (mECG) from the abdominal ECG signals. Template subtraction is a method that consists in modeling and removing the mECG’s mean period i.e. the signal waveform defined as the Euclidean mean of all periods. This mean period is then subtracted to all periods to extract the fetal ECG (fECG). Such a method is not accurate because each mECG’s period is not correctly aligned with the mean period. We propose to take account of the diffeomorphism of each period to improve the precision of the model and remove the mECG more efficiently. The soft-dynamic time warping (DTW) algorithm is used to compute the mean mECG period and the alignment between the mean period and all periods. Our approach is compared to a classic template subtraction on synthetic and real databases. Results show that considering the dynamic time warping allows a better removal of the mECG.

Keywords: Cardiovascular regulation - Autonomic nervous system, Cardiovascular regulation - Heart rate variability, Cardiovascular and respiratory signal processing - Heart Rate and Blood Pressure Variability
Abstract: Abstract— Autonomic regulation is essential in enabling a healthy pregnancy. In fact, several pregnancy complications are associated with autonomic dysfunction. Better understanding of the maternal autonomic state during healthy pregnancy may aid in the early detection of such complications. One aspect of autonomic regulation is autonomic responsiveness, which can by assessed by phase rectified signal averaging (PRSA). While other areas of research have found blunted physiological responses in pregnancy, this paper presents the first investigation of maternal autonomic responsiveness as assessed by PRSA. We find significantly reduced rates of responses, as well as an attenuated capacity for heart rate acceleration when comparing pregnant women to non-pregnant controls. We hypothesize that this attenuated autonomic control may serve to protect the mother against her imbalanced autonomic state, as increased sympathetic and decreased parasympathetic modulation accompany healthy pregnancies.

Clinical Relevance— Maternal autonomic responsiveness is attenuated in pregnancy in comparison to non-pregnant women. Understanding maternal autonomic state not only improves our knowledge of gestational physiology but also forms the basis for the early detection of pregnancy complications associated with maternal autonomic dysfunction.

Keywords: Respiratory transport, mechanics and control - Respiratory variability, Respiratory transport, mechanics and control - Work of breathing, Respiratory transport, mechanics and control - Pulmonary mechanics in disease
Abstract: Globally, an estimated 2.4 million newborns died in the first month of life in 2019. With quality care at birth and treat immediately after birth millions of newborn deaths are expected to be averted. In 2019, most neonatal deaths were due to preterm birth and intrapartum related complications (birth asphyxia, lack of respiration at birth). For these newborns, healthcare providers provide positive pressure ventilation. However, the lungs of newborns are stiff and small, and proper ventilation can be difficult to achieve. Flow sensors can improve the technique because they can convey ventilation parameters. In this study, we have developed a flow sensor that can be used for neonatal resuscitation. The results showed that the sensor had low flow resistance (0.18kPa@10L/min), the measurement accuracy at low flow rates was higher than commercially available flow sensors, was robust to noise when intubation tube was connected, and did not increase the dead space. In the future, we will examine display and usability, and develop devices useful for resuscitation of newborns.

Keywords: Pulmonary and critical care - Bioengineering applications in Intensive care, Pulmonary and critical care - Ventilatory Assist Devices
Abstract: During mechanical ventilation of the neonate the main goal is to stabilize respiratory function of the often premature lungs. Ventilating the patient without inflicting harm is then the subordinated next goal. Ideally the arterial partial pressure of CO2 lays within a normocapnic range and fluctuations are kept minimal. By closely monitoring CO2 and controlling ventilation parameters accordingly, CO2 levels in the blood can be managed. We present an approach consisting of a cascaded controller for arterial CO2 by approximating arterial partial pressure PaCO2 from end-tidal PetCO2. As a proof of concept, feasibility of the controller was first evaluated on a mathematical patient model and subsequently in-vivo in lamb experiments. The controller is able to regulate CO2 into a normocapnic range in both setups with satisfactory stationarity within the target range. Estimation of the arterial partial pressure of CO2 remains a critical aspect that needs to be further investigated.

Keywords: Cardiovascular and respiratory signal processing - Lung Sounds, Cardiovascular and respiratory signal processing - Cardiovascular signal processing, Pulmonary and critical care - Pulmonary disease
Abstract: Neonatal respiratory distress is a common condition that if left untreated, can lead to short- and long-term complications. This paper investigates the usage of digital stethoscope recorded chest sounds taken within 1min post-delivery, to enable early detection and prediction of neonatal respiratory distress. Fifty-one term newborns were included in this study, 9 of whom developed respiratory distress. For each newborn, 1min anterior and posterior recordings were taken. These recordings were pre-processed to remove noisy segments and obtain high-quality heart and lung sounds. The random undersampling boosting (RUSBoost) classifier was then trained on a variety of features, such as power and vital sign features extracted from the heart and lung sounds. The RUSBoost algorithm produced specificity, sensitivity, and accuracy results of 85.0%, 66.7% and 81.8%, respectively.

Keywords: Cardiac electrophysiology - Pacemakers, Cardiac electrophysiology - Defibrillation, ablation, and cardioversion
Abstract: Infants and children with congenital heart defects often receive a cardiac implantable electronic device (CIED). Because transvenous access to the heart is difficult in patients with small veins, the majority of young children receive epicardial CIEDs. Unfortunately, however, once an epicardial CIED is placed, patients are no longer eligible to receive magnetic resonance imaging (MRI) exams due to the unknown risk of MRI-induced radiofrequency (RF) heating of the device. Although many studies have assessed the role of device configuration in RF heating of endocardial CIEDs in adults, such case for epicardial devices in pediatric patients is relatively unexplored. In this study, we evaluated the variation in RF heating of an epicardial lead due to changes in the lateral position and orientation of the implantable pulse generator (IPG). We found that changing the orientation and position of the IPG resulted in a five-fold variation in the RF heating at the lead’s tip. Maximum heating was observed when the IPG was moved to a left lateral abdominal position of patient, and minimum heating was observed when the IPG was positioned directly under the heart.

Keywords: Therapeutic devices and systems - ablation systems and technologies, Image-guided therapies - RF and microwave ablation
Abstract: Gastric ablation has recently emerged as a promising potential therapy for correcting bioelectrical dysrhythmias that underpin many gastrointestinal motility disorders. Despite similarities to well-developed cardiac radiofrequency (RF) ablation, gastric RF ablation is in its infancy and has thus far been limited to temperature-controlled, non-irrigated settings. The potential benefits of power-controlled and irrigated RF ablation have not been investigated in gastric tissue. In this study, RF ablation was performed in vivo in pigs (n=5) using a range of power-control (10-30 W, 10 s per point) and irrigation (2-5 mL/min) settings and compared to known temperature-controlled (65^oC), non-irrigated settings. Excised tissue was stained with H&E. Lesion surface area was calculated and tissue damage was quantitatively ranked by visual assessment. The results demonstrated that irrigation allowed greater energy delivery to tissue with reduced interface temperatures compared to non-irrigated settings. Power settings above 10 W created lesions that extended through the full-thickness of the muscle layer, which suggests the parameter range that can now be used to correct gastric dysrhythmias. Clinical Relevance— This work presents the results of power-controlled, irrigated RF ablation settings applied to the in vivo porcine stomach. The relationships of both lesion area and depth to ablation dose provides an improved insight into which energy doses could provide a safe and effective therapeutic response.

Keywords: Image-guided therapies - Interstitial thermal therapy, Image-guided therapies - MRI-compatible instrumentation and device management, Therapeutic devices and systems - ablation systems and technologies
Abstract: Magnetic Resonance Thermometry Imaging (MRTI) holds great potential in laser ablation (LA) monitoring. It provides the real-time multidimensional visualization of the treatment effect inside the body, thus enabling accurate intraoperative prediction of the thermal damage induced. Despite its great potential, thermal maps obtained with MRTI may be affected by numerous artifacts. Among the sources of error producing artifacts in the images, the cavitation phenomena which could occur in the tissue during LA induces dipole-structured artifacts. In this work, an analysis of the cavitation artifacts occurring during LA in a gelatin phantom in terms of symmetry in space and symmetry of temperature values was performed. Results of 2 W and 4 W laser power were compared finding higher symmetry for the 2 W case in terms of both dimensions of artifact-lobes and difference in temperature values extracted in specular pixels in the image. This preliminary investigation of artifact features may provide a step forward in the identification of the best strategy to correct and avoid artifact occurrence during thermal therapy monitoring.

Keywords: Therapeutic devices and systems - ablation systems and technologies, Computer modeling for treatment planning
Abstract: Laser ablation is a rising technique used to induce a localized temperature increment for tumor ablation. The outcomes of the therapy depend on the tissue thermal history. Monitoring devices help to assess the tissue thermal response, and their combination with a control strategy can be used to promptly address unexpected temperature changes and thus reduce unwanted thermal effects. In this application, numerical simulations can drive the selection of the laser control settings (i.e., laser power and gain parameters) and allow evaluating the thermal effects of the control strategies. In this study, the influence of different control strategies (On-Off and PID-based controls) is quantified considering the treatment time and the thermal effect on the tissue. Finite element model-based simulations were implemented to model the laser-tissue interaction, the heat-transfer, and the consequent thermal damage in liver tissue with tumor. The laser power was modulated based on the temperature feedback provided within the tumor safety margin. Results show that the chosen control strategy does not have a major influence on the extent of thermal damage but on the treatment duration; the percentage of necrosis within the tumor domain is 100% with both strategies, while the treatment duration is 630 s and 786 s for On-Off and PID, respectively. The choice of the control strategy is a trade-off between treatment duration and unwanted temperature overshoot during closed-loop laser ablation.

Keywords: Therapeutic devices and systems - ablation systems and technologies, Image-guided therapies - MRI-compatible instrumentation and device management, Image-guided therapies - RF and microwave ablation
Abstract: The MRI hybrid ablation system is an approach to use the MR (magnetic resonance) scanner’s radiofrequency amplifier itself as power source for ablation. Hereby, an electrode is connected to the MR internal radiofrequency amplifier. An average RF power is provided through a train of short RF pulses, which is sufficient to destroy tissue thermally. However, ablation with too high power values can cause tissue carbonizations, thus impeding the ablation procedure. Therefore, monitoring of the energy and the power absorbed inside the tissue is necessary. For this purpose, a measurement system was designed to measure the energy applied to the tissue when using the concept of an MRI hybrid ablation system. The system consists of a dual-directional coupler, RF-to-RMS sensors, and a microcontroller. The gradient calculation of the measured energy curve provides information about the absorbed RF power in the tissue. Validation measurements of the system were performed and compared with measurements from the MR-internal power measurement system. The energy measurement system was also tested in an ablation experiment with ex-vivo animal tissue using the MRI hybrid ablation system. The measurements showed that the applied RF power can be monitored in real-time. It has been shown that the mean RF power absorbed in the patient decreased during an ablation procedure due to an occurring impedance mismatch and tissue changes. In further work, the influence of the monitoring system on the quality of the MR images should be investigated.

Keywords: Therapeutic devices and systems - ablation systems and technologies, Neuromuscular systems - Muscle stimulation, Models and simulations of therapeutic devices and systems
Abstract: Irreversible electroporation (IRE), or pulsed field ablation, employs microsecond-duration pulsed electric fields to generate targeted cellular damage without injury to the underlying tissue architecture. Biphasic, burst-type waveforms (termed high-frequency IRE, or H-FIRE) have garnered attention for their ability to elicit clinically relevant ablation volumes while reducing several undesirable side effects (muscle contractions/electrochemical effects) seen with monophasic pulses. Pulse width is generally the main (or only) parameter considered during burst construction, with little attention given to the delays within the burst. In this work, we tested the hypothesis that H-FIRE waveforms could be further optimized by manipulating only the interpulse delay between biphasic pulses within each burst. Using benchtop, ex vivo, and in vivo models, we demonstrate that extended interpulse delays (i.e., ~100 µs) reduce the severity of induced muscle contractions, alleviate mechanical tissue destruction, and minimize the chances of electrical arcing.

Clinical Relevance— This proof-of-concept study shows that H-FIRE waveforms with extended interpulse delays provide several therapeutic benefits over conventional waveforms.

Keywords: Image segmentation, Machine learning / Deep learning approaches, Image analysis and classification - Machine learning / Deep learning approaches
Abstract: The use of total marrow and lymphoid irradiation (TMLI) as part of conditioning regimens for bone marrow transplantation is trending due to its advantages in disease control and low toxicity. Accurate contouring of target structures such as bone and lymph nodes plays an important role in irradiation planning. However, this process is often time-consuming and prone to inter-observer variation. Recently, deep learning methods such as convolutional neural networks (CNNs) and vision transformers have achieved tremendous success in medical image segmentation, therefore enabling fast semi-automatic radiotherapy planning. In this paper, we propose a dual-encoder U-shaped model named DE-Net, to automatically segment the target structures for TMLI. To enhance the learned features, the encoder of DE-Net is composed of parallel CNNs and vision transformers, which can model both local and global contexts. The multi-level features from the two branches are progressively fused by intermediate modules, therefore effectively preserving low-level details. Our experiments demonstrate that the proposed method achieves state-of-the-art results and a significant improvement in lymph node segmentation compared with existing methods.

Keywords: Image segmentation, Machine learning / Deep learning approaches, Magnetic resonance imaging - Other organs
Abstract: In our comprehensive experiments and evaluations, we show that it is possible to generate multiple contrast (even all synthetically) and use synthetically generated images to train an image segmentation engine. We showed promising segmentation results tested on real multi-contrast MRI scans when delineating muscle, fat, bone and bone marrow, all trained on synthetic images. Based on synthetic image training, our segmentation results were as high as 93.91%, 94.11%, 91.63%, 95.33%, for muscle, fat, bone, and bone marrow delineation, respectively. Results were not significantly different from the ones obtained when real images were used for segmentation training: 94.68%, 94.67%, 95.91%, and 96.82%, respectively.

Keywords: Image segmentation, Optical imaging, Machine learning / Deep learning approaches
Abstract: Semantic segmentation of surgery scenarios is a fundamental task for computer-aided surgery systems. Precise segmentation of surgical instruments and anatomies contributes to capturing accurate spatial information for tracking. However, uneven reflection and class imbalance lead the segmentation in cataract surgery to a challenging task. To desirably conduct segmentation, a network with multi-view decoders (MVD-Net) is proposed to present a generalizable segmentation for cataract surgery. Two discrepant decoders are implemented to achieve multi-view learning with the backbone of U-Net. The experiment is carried out on the Cataract Dataset for Image Segmentation (CaDIS). The ablation study verifies the effectiveness of the proposed modules in MVD-Net, and superior performance is provided by MVD-Net in the comparison with the state-of-the-art methods. The source code will be publicly released.

Keywords: Image segmentation, Machine learning / Deep learning approaches, CT imaging
Abstract: Unet architectures are promising deep learning networks exploited to perform the automatic segmentation of bone CT images, in line with their ability to deal with pathological deformations and size-varying anatomies. However, bone degeneration, like the development of irregular osteophytes as well as mineral density alterations might interfere with this automated process and demand extensive manual refinement. The aim of this work is to implement an innovative Unet variant, the CEL-Unet, to improve the femur and tibia segmentation outcomes in osteoarthritic knee joints. In this network the decoding path is split into a region and contour-aware branch to increase the prediction reliability in such pathological conditions. The comparison between the segmentation results achieved with a standard Unet and its novel variant (CEL-Unet) was performed as follows: the Unet was trained with 5 different loss functions: Dice Loss, Focal Loss, Exponential Logarithmic Loss, Double Cross Entropy Loss and Distanced Cross Entropy loss. The CEL-Unet was instead trained with two loss functions, one for each of the network outputs, namely Mask and Edge, yielding the so-called Combined Edge Loss (CEL) function. A set of 259 knee CT scans was used to train the model and test segmentation performance. The CEL-Unet outperformed all other Unet-based models, reaching the highest Jaccard values of about 0.97 and 0.96 on femur and tibia, respectively.

Keywords: Image segmentation, Machine learning / Deep learning approaches, Image feature extraction
Abstract: Label scarcity has been a long-standing issue for biomedical image segmentation, due to high annotation costs and professional requirements. Recently, active learning (AL) strategies strive to reduce annotation costs by querying a small portion of data for annotation, receiving much traction in the field of medical imaging. However, most of the existing AL methods have to initialize models with some randomly selected samples followed by active selection based on various criteria, such as uncertainty and diversity. Such random-start initialization methods inevitably introduce under-value redundant samples and unnecessary annotation costs. For the purpose of addressing the issue, we propose a novel self-supervised assisted active learning framework in the cold-start setting, in which the segmentation model is first warmed up with self-supervised learning (SSL), and then SSL features are used for sample selection via latent feature clustering without accessing labels. We assess our proposed methodology on skin lesions segmentation task. Extensive experiments demonstrate that our approach is capable of achieving promising performance with substantial improvements over existing baselines.

Keywords: Image segmentation, Cardiac imaging and image analysis, Magnetic resonance imaging - Cardiac imaging
Abstract: While convolutional neural networks (CNNs) have shown potential in segmenting cardiac structures from magnetic resonance (MR) images, their clinical applications still fall short of providing reliable cardiac segmentation. As a result, it is critical to quantify segmentation uncertainty in order to identify which segmentations might be troublesome. Moreover, quantifying uncertainty is critical in real-world scenarios, where input distributions are frequently moved from the training distribution due to sample bias and non-stationarity. Therefore, well-calibrated uncertainty estimates provide information on whether a model's output should (or should not) be trusted in such situations. In this work, we used a Bayesian version of our previously proposed CondenseUNet framework featuring both a learned group structure and a regularized weight-pruner to reduce the computational cost in volumetric image segmentation and help quantify predictive uncertainty. Our study further showcases the potential of our deep-learning framework to evaluate the correlation between the uncertainty and the segmentation errors for a given model. The proposed model was trained and tested on the Automated Cardiac Diagnosis Challenge (ACDC) dataset featuring 150 cine cardiac MRI patient dataset for the segmentation and uncertainty estimation of the left ventricle (LV), right ventricle (RV), and myocardium (Myo) at end-diastole (ED) and end-systole (ES) phases.

Keywords: Image classification, Image feature extraction, Image analysis and classification - Machine learning / Deep learning approaches
Abstract: Automated skin cancer diagnosis is challenging due to inter-class uniformity, intra-class variation, and the complex structure of dermoscopy images. Convolutional neural networks (CNN) have recently made considerable progress in melanoma classification, even in the presence of limited skin images. One of the drawbacks of these methods is the loss of image details caused by downsampling high-resolution skin images to a low resolution. Further, most approaches extract features only from the whole skin image. This paper proposes an ensemble feature fusion and sparse autoencoder (SAE) based framework to overcome the above issues and improve melanoma classification performance. The proposed method extracts features from two streams, local and global, using a pre-trained CNN model. The local stream extracts features from image patches, while the global stream derives features from the whole skin image, preserving both local and global representation. The features are then fused, and an SAE framework is subsequently designed to enrich the feature representation further. The proposed method is validated on ISIC 2016 dataset and the experimental results indicate the superiority of the proposed approach.

Keywords: Magnetic resonance imaging - Diffusion tensor, diffusion weighted and diffusion spectrum imaging, Brain imaging and image analysis, Image reconstruction and enhancement - Filtering
Abstract: Whole-brain tractograms generated from diffusion MRI digitally represent the white matter structure of the brain and are composed of millions of streamlines. Such tractograms can have false positive and anatomically implausible streamlines. To obtain anatomically relevant streamlines and tracts, supervised and unsupervised methods can be used for tractogram clustering and tract extraction. Here we propose FiberNeat, an unsupervised white matter tract filtering method. FiberNeat takes an input set of streamlines that could either be unlabeled clusters or labeled tracts. Individual clusters/tracts are projected into a latent space using nonlinear dimensionality reduction techniques, t-SNE and UMAP, to find spurious and outlier streamlines. In addition, outlier streamline clusters are detected using DBSCAN and then removed from the data in streamline space. We performed quantitative comparisons with expertly delineated tracts. We ran FiberNeat on 131 participants' data from the ADNI3 dataset. We show that applying FiberNeat as a filtering step after bundle segmentation improves the quality of extracted tracts and helps improve tractometry.

Keywords: Machine learning / Deep learning approaches, Image reconstruction and enhancement - Image synthesis
Abstract: Multi-modality magnetic resonance (MR) images provide complementary information for disease diagnoses. However, modality missing is quite usual in real-life clinical practice. Current methods usually employ convolutionbased generative adversarial network (GAN) or its variants to synthesize the missing modality. With the development of vision transformer, we explore its application in the MRI modality synthesis task in this work. We propose a novel supervised deep learning method for synthesizing a missing modality, making use of a transformer-based encoder. Specifically, a model is trained for translating 2D MR images from T1-weighted to T2-weighted based on conditional GAN (cGAN). We replace the encoder with transformer and input adjacent slices to enrich spatial prior knowledge. Experimental results on a private dataset and a public dataset demonstrate that our proposed model outperforms state-of-the-art supervised methods for MR image synthesis, both quantitatively and qualitatively.

Keywords: Machine learning / Deep learning approaches, Image segmentation, Magnetic resonance imaging - Other organs
Abstract: The aim of the study is to present and tune a fully automatic deep learning algorithm to segment colorectal cancers (CRC) on MR images, based on a U-Net structure. It is a multicenter study, including 3 different Italian institutions, that used 4 different MRI scanners. Two of them were used for training and tuning the systems, while the other two for the validation. The implemented algorithm consists of a pre-processing step to normalize and to highlight the tumoral area, followed by the CRC segmentation using different U-net structures. Automatic masks were compared with manual segmentations performed by three experienced radiologists, one at each center. The two best performing systems (called mdl2 and mdl3), obtained a median Dice Similarity Coefficient of 0.68(mdl2) - 0.69(mdl3), precision of 0.75(mdl2) - 0.71(mdl3), and recall of 0.69(mdl2) - 0.73(mdl3) on the validation set. Both systems reached high detection rates, having, respectively, 1 or 2 false negatives in the validation set. These encouraging results, if confirmed on larger dataset, might improve the management of patients with CRC, since it can be used as a fast and precise tool for further radiomics analyses.

Clinical Relevance— To provide a reliable tool able to automatically segment CRC tumors that can be used as first step in future radiomics studies aimed at predicting response to chemotherapy and personalizing treatment.

Keywords: Machine learning / Deep learning approaches, X-ray radiography
Abstract: This study developed and evaluated deep learning models for prediction of hip knee ankle angle (HKAA) measurements on postoperative full-limb radiographs of total knee arthroplasty (TKA) patients. The process involved extracting regions of interest (RoI) on 1899 radiographs, applying landmark detection by regressing heatmaps based on the extracted RoI, and finally calculating the HKAA. We used mean and standard deviation of the differences between HKAA angle predictions and annotations as the evaluation metric. Postoperative HKAA difference between model predictions and annotations was 0.65° ± 0.82° and the percentage of difference smaller than 1.5° was 95.0%. In conclusion we developed a fully automated tool to measure HKAA accurately and precisely on postoperative full-limb radiographs of TKA patients.

Keywords: Image feature extraction, Image segmentation, Image classification
Abstract: Self-supervised pretext tasks have been introduced as an effective strategy when learning target tasks on small annotated data sets. However, while current research focuses on exploring novel pretext tasks for meaningful and reusable representation learning for the target task, the study of its robustness and generalizability has remained relatively under-explored. Specifically, it is crucial in medical imaging to proactively investigate performance under different perturbations for reliable deployment of clinical applications. In this work, we revisit medical imaging networks pre-trained with self-supervised learnings and categorically evaluate robustness and generalizability compared to vanilla supervised learning. Our experiments on pneumonia detection in X-rays and multi-organ segmentation in CT yield conclusive results exposing the hidden benefits of self-supervision pre-training for learning robust feature representations.

Keywords: Neural signals - Machine learning & Classification, Neural signal processing, Neural interfaces - Implantable systems
Abstract: The peripheral nervous system is a key target for the development of neural interfaces. However, recording from the peripheral nerves can be challenging especially when chronic implantation is desired. Nerve cuffs are frequently employed using either two or three contacts to provide a single recording channel. Advancements in manufacturing technology have enabled multi-contact cuffs, enabling measurement of both temporal (i.e., velocity) and spatial information (i.e., spatial location). Selective techniques have been developed with different time resolutions but it is unclear how the number of contacts and their spatial configuration affect their performance. Thus, this paper investigates two extraneural recording techniques (LDA and spatiotemporal signaturess) and compares them using recordings made from the sciatic nerve of rats using high density (HD, 56 contact), reduced-HD (16 channels), and low density (LD, 16 contact) datasets. Performance of the two techniques was evaluated using classification accuracy and F1-score. Both techniques show an expected improvement in classification accuracy with the spatiotemporal signature approach showing a 21.6 (LD to HD) – 24.6% (reduced HD to HD) increase and the LDA approach showing a 2.9 (reduced HD to HD) – 36.8% (LD to HD) and had comparable results with each other in both the LD and HD datasets.

Keywords: Neural stimulation, Neurorehabilitation, Neural signal processing
Abstract: Temporal interference stimulation has been suggested as a method to reach deep targets during transcutaneous electrical stimulation. Despite its growing use in transcutaneous stimulation therapies, the mechanism of its operation is not fully understood. Recent efforts to fill that gap have focused on computational modelling, in vitro experiments and in vivo experiments relying on physical observations – e.g., sensation or movement. This paper expands the current range of experimental methods by demonstrating in vivo extraneural recordings from the ulnar nerve of a pig while applying temporal interference stimulation at a location targeting a distal part of the nerve. The main aim of the experiment was to compare neural activation using sinusoidal stimulation (100 Hz, 2 kHz, 4 kHz) and temporal interference stimulation (2kHz and 4kHz). The recordings showed a significant increase in the magnitude of stimulation artefacts at higher frequencies. While those artefacts could be removed and provided an indication of the depth of modulation, they resulted in the saturation of the amplifiers, limiting the stimulation currents and amplifier gains used. The results of the 100 Hz sine wave stimulation showed clear neural activity correlated to the stimulation waveform. However, this was not observed with temporal interference stimulation. The results suggesting that, despite its greater penetration, higher currents might be required to observe a neural response with temporal interference stimulation and more complex artefact rejection techniques may be required to validate the method.

Keywords: Neuromuscular systems - EMG processing and applications, Neural stimulation
Abstract: Multi-contact epineural electrical stimulation is a technique that can be used to restore grip movements in people with complete tetraplegia. However, neural stimulation can induce undesired H-reflex. This reflex is known to induce a global lower recruitment threshold together with a steepest recruitment curve leading to a degraded selective response. In this study, during stimulation of the median nerve using a multi-contact cuff electrode, a H-reflex response was observed for one muscle (the pronator teres i.e. PT) among the five recorded. As both M-wave and H-wave were separately recorded, we compared the changes of recruitment, recruitment order and selectivity with and without the H-reflex and found that blocking the reflex would have enhance the selectivity and increase the range of the intensity amplitude while providing a higher level of gripping force. Thus, blocking H-reflex is an important issue to further enhance epineural multicontact selective stimulation.

Keywords: Neural interfaces - Biomaterials, Neural stimulation, Neural interfaces - Tissue-electrode interface
Abstract: Electrical stimulation after peripheral nerve injury (PNI) has the potential to promote more rapid and complete recovery of damaged fiber tracts. While permanently implanted devices are commonly used to treat chronic or persistent conditions, they are not ideal solutions for transient medical therapies due to high costs, increased risk of surgical injury, irritation, infection, and persistent inflammation at the site of the implant. Furthermore, removal of temporary leads placed on or around peripheral nerves may have unacceptable risk for nerve injury, which is counterproductive in developing therapies for PNI treatment. Transient devices which provide effective clinical stimulation while being capable of harmless bioabsorption may overcome key challenges in these areas. However, current bioabsorbable devices are limited in their robustness and require complex fabrication strategies and novel materials which may complicate their clinical translation pathway. In this study, we present a simple bioabsorbable / biodegradable electrode fabricated by modifying standard absorbable sutures, and we present data characterizing our prototype’s stability in vitro and in vivo.

Keywords: Neural interfaces - Microelectrode technology, Neural interfaces - Implantable systems, Sensory neuroprostheses - Auditory
Abstract: The Utah Electrode Array (UEA) and its variants (e.g., the Utah Slanted Electrode Array, or USEA) have been prominent contributors to advances in the field of neural engineering over the past decade. The most common means of inserting UEA and USEA devices into neural tissue is pneumatic insertion performed by an insertion wand and a pneumatic controller. As design changes from the well-established standards occur to better suit specialized surgical applications, it becomes essential to verify that the alterations do not compromise the structural integrity of the device during insertion. This paper characterizes and demonstrates the reliability of specialized USEAs and insertion wands designed for auditory nerve implants following pneumatic insertion into a rat sciatic nerve. The results show that proposed changes in the USEA form factor and pneumatic insertion ergonomics do not compromise implant stability and device structural viability.

Keywords: Neural stimulation, Motor learning, neural control, and neuromuscular systems, Neuromuscular systems - Learning and adaption
Abstract: Neurological disorders such as stroke remain leading causes of disability worldwide. A current thrust in the neurorehabilitation of such disorders involves exogenous neuromodulation of cranial nerves in order to enhance neuroplasticity and maximize recovery of function. Here we present preliminary results on the effects of kilohertz range electrical stimulation of the trigeminal nerve (TNS) on motor learning,using an upper extremity visuomotor adaptation paradigm. Twenty-five (25) healthy adult subjects were randomly assigned to 2 groups: 3kHz stimulation (n=13) and sham (n=12). Participants performed a visuomotor rotation task that involved center-out reaching movements to eight vertically arranged targets. Four blocks of trials were performed: two baseline blocks with veridical visual feedback, one adaptation block involving a 30 degree CCW rotation of hand visual feedback, and one washout block with no rotation. TNS was applied for 20 minutes before the 2nd baseline block using two electrodes targeting the ophthalmic branches of the trigeminal nerve. Early in the rotation block, learning rates were similar between the 3kHz and sham groups but gradually diverged, with the 3kHz group demonstrating slightly faster rates than sham later in the rotation block. The results provide new information on the potential use of TNS in neurorehabilitation.

Keywords: Neurorehabilitation
Abstract: This study examines longitudinal data of subjects initially examined in the early subacute period of recovery following a stroke with a test of reach to grasp (RTG) kinematics in an attempt to identify changes in movement patterns during the period of heightened neural recovery following a stroke. Subjects (n=8) were a convenience sample of persons with stroke that participated in an intervention trial. Baseline Upper Extremity Fugl Meyer Assessment (UEFMA) scores ranged between 31 and 52 and ages were between 49 and 83. The UEFMA and RTG test were collected prior to intervention, immediately after the intervention (approximately 18 days later post baseline) and one month after the intervention. RTG data for the uninvolved UE was collected at the one-month session. Subjects reached for objects placed on a table 10 cm from their sternums, picking them up and placing them on a target 30 cm from their acromioclavicular joints. Data was collected using an optical motion capture system. Active makers were placed on each fingertip, metacarpophalangeal, and proximal interphalangeal joint. Four additional passive markers were placed on the dorsum of the hand, the elbow, the shoulder, and the sternum. Subjects demonstrated statistically significant improvements in reaching duration, reaching trajectory smoothness, time after peak velocity and peak grip aperture. All of these measures correlated significantly with improvements in UEFMA.

Keywords: Neural interfaces - Microelectrode technology, Neuromuscular systems - EMG processing and applications, Neural interfaces - Bioelectric sensors
Abstract: High signal-to-noise ratio (SNR) electromyography (EMG) recordings are essential for identifying and analyzing single motor unit activity. While high-density electrodes allow for greater spatial resolution, the smaller electrode area translates to a higher impedance and lower SNR. In this study, we developed an implantable and flexible 3D microelectrode array (MEA) with low impedance that enables high-quality EMG recording. With polyimide micro-cones realized by standard photolithography process and PEDOT:PSS coating, this design can increase effective surface area by up to 250% and significantly improve electrical performance for electrode sites with various geometric surface areas, where the electrode impedance is at most improved by 99.3%. Acute EMG activity from mice was recorded by implanting the electrodes in vivo, and we were able to detect multiple individual motor units simultaneously and with high resolution (SNR >> 100). The charge storage capacity was measured to be 34.2 mC/cm^2, indicating suitability of the electrodes for stimulation applications as well.

Keywords: Neuromuscular systems - EMG processing and applications, Motor learning, neural control, and neuromuscular systems, Neural signals - Machine learning & Classification
Abstract: Recently, there has been a surge of significant interest on application of Deep Learning (DL) models to autonomously perform hand gesture recognition using surface Electromyogram (sEMG) signals. Many of the existing DL models are, however, designed to be applied on sparse sEMG signals. Furthermore, due to the complex structure of these models, typically, we are faced with memory constraint issues, require large training times and a large number of training samples, and; there is the need to resort to data augmentation and/or transfer learning. In this paper, for the first time (to the best of our knowledge), we investigate and design a Vision Transformer (ViT) based architecture to perform hand gesture recognition from High Density (HD-sEMG) signals. Intuitively speaking, we capitalize on the recent breakthrough role of the transformer architecture in tackling different complex problems together with its potential for employing more input parallelization via its attention mechanism. The proposed Vision Transformer-based Hand Gesture Recognition (ViTHGR) framework can overcome the aforementioned training time problems and can accurately classify a large number of hand gestures from scratch without any need for data augmentation and/or transfer learning. The efficiency of the proposed ViT-HGR framework is evaluated using a recently released HD-sEMG dataset consisting of 65 isometric hand gestures. Our experiments with 64-sample (31.25 ms) window size yield average test accuracy of 84.62 ± 3.07%, where only 78,210 learnable parameters are utilized in the model. The compact structure of the proposed ViT-based ViT-HGR framework (i.e., having significantly reduced number of trainable parameters) shows great potentials for its practical application for prosthetic control.

Keywords: Neuromuscular systems - EMG processing and applications, Neural signal processing, Motor learning, neural control, and neuromuscular systems
Abstract: Swallowing is a vital function that serves to safely transport food and fluid to the stomach, while simultaneously protecting our airways. Evaluation of swallowing is important for the diagnosis and rehabilitation of individuals with dysphagia, a disorder of swallowing. Flexible high-density surface electromyography (HD sEMG) arrays were designed and fabricated to span the floor of mouth and neck muscles. These arrays were applied on 6 healthy participants over duplicate recording sessions. During each recording session, participants performed three different swallowing motor tasks. The HD sEMG signals were filtered and tasks extracted. For each task, the RMS amplitude was computed, visualized, and compared. Dynamic motor coordination was evident in the filtered signals traces, with different electrode locations showing unique temporal activations. The 2D topographical maps allowed the location of different RMS intensities to be visualized, revealing qualitatively similar patterns across participants and tasks. Behavioral patterns were also seen within RMS quantifications. The RMS metric had a minimal variation of 3.1 ± 1.9 µV between separate recording sessions and found a significant difference between non-effortful 3 ml and effortful 3 ml swallow tasks (p = 0.006). The HD-sEMG array successfully recorded differences in muscle activations during swallowing and was able to discern between two different motor tasks. The arrays could be promptly applied and unlike current dysphagia assessment methods, offers a spatially detailed non-invasive assessment of the neuromuscular performance of swallowing.

Keywords: Neuromuscular systems - EMG processing and applications, Neurorehabilitation
Abstract: Abstract— Stroke survivors experience muscular pattern alterations of the upper limb that decrease their ability to perform daily-living activities. The Box and Block test (BBT) is widely used to assess the unilateral manual dexterity. Although BBT provides insights into functional performance, it returns limited information about the mechanisms contributing to the impaired movement. This study aims at exploring the BBT by means of muscle synergies analysis during the execution of BBT in a sample of 12 healthy participants with their dominant and non-dominant upper limb. Results revealed that: (i) the BBT can be described by 1 or 2 synergies; the number of synergies (ii) does not differ between dominant and non-dominant sides and (iii) varies considering each phase of the task; (iv) the transfer phase requires more synergies. Clinical Relevance— This preliminary study characterizes muscular synergies during the BBT task in order to establish normative patterns that could assist in understanding the neuromuscular demands and support future evaluations of stroke deficits.

Keywords: Human performance - Sensory-motor, Human performance - Modelling and prediction
Abstract: Handwriting is an important component of academic curricula and grapho-motor skills (GMS) support learning, reading, memory and self-confidence. Teachers and clinicians report increase in children experiencing problems with acquiring fluid and legible handwriting. To date gold-standard tests evaluating children’s GMS, mostly rely on pen and paper tests, requiring extensive coding time and subject to high inter-rater variability. This work presents preliminary data on a new digital platform for Grapho-motor Handwriting Evaluation & Exercise (GHEE), attempting to overcome limitations of available digitalized methods for GMS evalution. In fact, contrary to previous systems, GHEE design originated from comparisons among multiple standardized tests and was based on a human-machine interaction approach. GHEE hardware and software is presented as well as data on preliminary testing. Cursive handwriting data from six adult volunteers was analyzed according to six parameters of relevance, both automatically (i.e., using GHEE software) and manually (i.e., by a human coder). Comparisons among machine and human data sets allowed parsing out parameters to be extracted automatically and parameters requiring human-machine interaction. Results confirmed platform efficacy and feasibility of the proposed approach.

Keywords: Brain physiology and modeling - Neural circuits, Brain physiology and modeling - Sensory-motor, Neurological disorders - Mechanisms
Abstract: Dystonia is a neurological movement disorder characterized by twisting and repetitive movements or abnormal fixed postures. This complex brain disease has usually been associated with damages to the Basal Ganglia. However, recent studies point out the potential role of the cerebellum. Indeed, motor learning is impaired in dystonic patients, e.g. during eyeblink classical conditioning, a typical cerebellum-driven associative learning protocol, and rodents with local cerebellar damages exhibit dystonic movements. Alterations in the olivocerebellar circuit connectivity have been identified as a potential neural substrate of dystonia. Here, we investigated this hypothesis through simulations of eyeblink conditioning driven by a realistic spiking model of the cerebellum. The pathological model was generated by decreasing the signal transmission from the Inferior Olive to cerebellar cortex, as observed in animal experiments. The model was able to reproduce a reduced acquisition of eyeblink motor responses, with also an unproper timing. Indeed, this pathway is fundamental to drive cerebellar cortical plasticity, which is the basis of cerebellum-driven motor learning. Exploring different levels of damage, the model predicted the possible amount of underlying impairment associated with the misbehavior observed in patients. Simulations of other debated lesions reported in mouse models of dystonia will be run to investigate the cerebellar involvement in different types of dystonia. Indeed, the eyeblink conditioning phenotype could be used to discriminate between them, identifying specific deficits in the generation of motor responses. Future studies will also include simulations of pharmacological or deep brain stimulation treatments targeting the cerebellum, to predict their impact in improving symptoms.

Keywords: Sensory neuroprostheses - Visual, Smart neural implants, Human performance - Activities of daily living
Abstract: Abstract— Visual prostheses can improve vision for people with severe vision loss, but low image resolution and lack of peripheral vision limit their effectiveness. To address both problems, we developed a prototype advanced video processing system with a headworn depth camera and feature detection capabilities. We used computer vision algorithms to detect landmarks representing a goal and plan a path towards the goal, while removing unnecessary distractors from the video. If the landmark fell outside the visual prosthesis’s field-of-view (20 degrees central vision) but within the camera’s field-of-view (70 degrees), we provided vibrational cues to the left or right temple to guide the user in pointing the camera. We evaluated an Argus II retinal prosthesis participant with significant vision loss who could not complete the task (finding a door in a large room) with either his remaining vision or his retinal prosthesis. His success rate improved to 57%, 37.5%, and 100% while requiring 52.3, 83.0, and 58.8 seconds to reach the door using only vibration feedback, retinal prosthesis with modified video, and retinal prosthesis with modified video and vibration feedback, respectively. This case study demonstrates a possible means of augmenting artificial vision. Clinical Relevance— Retinal prostheses can be enhanced by adding computer vision and non-visual cues.

Keywords: Human performance - Attention and vigilance, Brain-computer/machine interface, Neural stimulation
Abstract: Attention enhancement can not only improve individual’s study and work performance, but also help to improve such sychological problems as anxiety and depression. Traditional attention enhancement approaches have high requirements on the external environment, and thus have such limitations as long intervention periods, high costs, poor universality, and insignificant therapeutic effects. Virtual Reality (VR) can provide interactive and immersive environments, which can help to break through these limitations and effectively enhance users’ attention. In this paper, we propose a novel eurofeedback attentional enhancement approach based on VR. The proposed approach utilizes the α band power in the parieto-occipital regions of the brain as an neurofeedback index of the users’ attention, and prompts users by changing the attributes of the VR environment. Statistical results show that the α band power reduces significantly in neurofeedback group compared with that in control group. Accordingly, task performances in neruofeedback group are improved by 6.44% compared with those of control group. Our results provided evidence for the effectiveness of neurofeedback on VR training environment.

Keywords: Neuromuscular systems - EMG processing and applications, Neuromuscular systems - Postural and balance, Neurological disorders - Traumatic brain injury
Abstract: Balance Dysfunction (BDF) is a severe consequence of Traumatic Brain Injury (TBI) that significantly increases the falls risk. However, the neuromuscular mechanisms of the BDF are not adequately researched. Therefore, in this study, our objective was to investigate the effects of a Computerized Biofeedback–based Balance Intervention (CBBI) on the muscle coactivation patterns in a group of TBI participants. This study presents the findings from 13 TBI individuals randomized into the Intervention group (TBI-INT, N=6) and Control group (TBI-CTL, N=7). Using a computerized posturography platform (Neurocom Balance Master) during baseline and follow-up assessment visits, the participant’s postural response to anterior-posterior balance perturbations was recorded in a multimodal setup including electroencephalography (EEG), electromyography (EMG), and the platform sway in terms of center of pressure (COP). The muscle responses were recorded from lower-limb muscles, including tibialis anterior (TA) and gastrocnemius (GAST), whose coactivation was computed using a metric called Co-Contraction Index (CCI). Clinical outcome measures such as Berg Balance Scale (BBS), 10 Meter Walk Test (10MWT), and Timed Up-and-Go (TUG) tests were used to evaluate functional balance and mobility. The comparison of CCI values across time points (baseline and follow-up) revealed a significant decrease (p<0.01) in the TBI-INT group but not TBI-CTL. The intervention-related changes in CCI correlated with the changes in BBS score (from baseline to follow-up). These preliminary findings demonstrate that the CBBI training may help postural stability by facilitating the coactivation between muscles involved in postural control.

Keywords: Neuromuscular systems - Computational modeling, Motor learning, neural control, and neuromuscular systems, Brain physiology and modeling - Neural dynamics and computation
Abstract: It is currently unknown what coordinate system or systems the primate motor cortex uses to represent movement, although experimental evidence has suggested several candidates. In order to understand how the physical geometry of the arm combines with computational constraints to influence the optimal choice of coordinate system, we construct a two-dimensional, physics-based arm model and couple it to a linear model of the motor cortex. The cortical model is provided with target positions and real time feedback of the current hand position in two different coordinate systems: cartesian and joint angle. We then optimize the parameters of the model subject to penalties on neural connectivity and muscle and neural energy use. We find that the optimized model strongly prefers to work in the joint angle coordinate system, suggesting that for neurons whose activity is closely tied to muscle activation, this is computationally the most efficient coordinate system in which to represent movement.

Keywords: Neural stimulation, Brain physiology and modeling - Neuron modeling and simulation, Brain physiology and modeling
Abstract: Computational methods of determining the response of neural tissue to electrical stimulation have demonstrated value for the development of novel devices and the programming of neuromodulation therapies. Detailed biophysical models are excessively computationally intensive for many applications; simple metrics to approximate activation can speed up progress in this area. The activating function provides such a useful metric. However, this measure, defined for a specific axon orientation, is not immediately applicable to computed electric fields to assess their effects. We demonstrate a method for computation of the activating function generalized to a field in order to allow rapid computation of the effects of stimulation on neural tissue while preserving information on axon orientation.