Keywords: Personal health systems, Personal/consumer health informatics, Personalized health/precision medicine
Abstract: The new era in medicine focuses on the development of tools and applications that empower citizens towards advanced health care practices and personalized approaches. Despite the vast achievements in pharmaceutical industry with innovative drugs of high specificity regarding their pharmacological mechanism of action, the use of complementary and alternative medicine (CAM) products worldwide, is on the rise. CAM products, which are usually of plant origin (herbal medicines), are generally advertised as natural and thus harmless with minimum side effects and frequently used by citizens as part of their self-medication lifestyle. However, several studies have shown that herbal medicines, as well as other CAM products, can potentially modulate biological mechanisms that are related with the pharmacological action of drugs and thus lead to what is called drug-herb interactions. In this respect, availability of relevant eHealth services and applications providing user-friendly information to patients regarding potential interactions of CAM products that they plan to use with their concurrent administered therapy, is of paramount importance. Such services could be used by both the healthcare providers (doctors, pharmacists, nurses etc.) for advanced and personalized approaches, and by the citizens themselves for minimization of self-medication errors. In this work we present an eHealth web application that utilizes drug-herb interaction databases and provides potential notifications

Keywords: Predictive analytics, Cancer genomics, Neuro genomics, Cardio genomics, Machine learning
Abstract: The majority of cancer drug sensitivity models are built utilizing genomic data measured before drug application to predict the steady state sensitivity of an applied drug. Restricting models to this type of data is limiting and can only explain one small piece of the puzzle. Better characterization of cancer cells can be accomplished through the use of proteomic data as this more directly corresponds to cellular activity. We have implemented models that predict cell viability utilizing protein expression measured post drug application. These models are built utilizing the Random Forest, Elastic Net, Partial Least Square Regression and Support Vector Regression algorithms in addition to stacked models. We have also utilized these same algorithms to predict the average protein inhibition of a cancer drug utilizing cell viability screens as input. Protein expression and cell viability data is taken from the HMS-LINCS database. We have shown that cell viability can be effectively predicted utilizing proteomic data and that we can estimate cancer drug protein inhibition utilizing a small number of cell line screens.

Keywords: Predictive analytics, Disease profiling and personalized treatment, Computational modeling and simulations in biology, physiology and medicine
Abstract: Successful translational studies within the field of Traumatic Brain Injury (TBI) are concerned with determining reliable markers of injury outcome at chronic time points. Determination of injury severity following Fluid Percussion Injury (FPI) has long been limited to the measured atmospheric pressure associated with the delivered pulse. Duration of unresponsiveness to toe pinch (unconsciousness) was next introduced as an extra marker of injury severity. The current study is an effort to assess the utilization of acute injury-induced biological responses (duration of toe pinch unresponsiveness, percent body weight change, quantification of brain edema, and apnea duration) to predict cognitive performance at a subacute time point following developmental brain injury. Cognitive performance, when measured at a subacute phase, after developmental FPI was negatively correlated with the following variables, duration of toe pinch unresponsiveness, percent weight change, and quantified level of brain edema. These finding suggest the potential utilization of reliable severity assessment of injury-induced biological responses in determining outcome measures at subacute time points.

Keywords: Health risk evaluation and modeling, Preventive health
Abstract: Obesity is one of the leading preventable causes of death in the United States (U.S.). Risk factor analysis is a process to identify and understand the risk factors contributing to a particular disease, and is an imperative component in the development of efficient and effective prevention and intervention efforts. Most existing methods usually aim to build a one-size-fits-all model to identify the risk factors at the population-level. However, this type of methods does not take into consideration of heterogeneity in the population. To overcome this limitation, we formulate the subpopulation specific obesity risk factors ranking problem, under the framework of multi-task learning (MTL), to identify a ranked list of obesity risk factors for each subpopulation (task) simultaneously with utilizing appropriate shared information across tasks. By synchronously learning multiple related tasks, MTL provides a paradigm to rank risk factors both at the subpopulation and population-levels.

Keywords: Epidemiology
Abstract: In this paper we introduce CASSIA, an assistant that facilitates rapid identification of antibiotic-bacteria pairs demonstrating decreases in susceptibility that need to be monitored. Specifically, CASSIA detects clinically and statistically significant decreases in susceptibility using antibiogram data. While previous studies have used the chi-squared test to evaluate the statistical significance of changes in antimicrobial susceptibility, these studies have not addressed the detection of clinical significance. CASSIA identifies statistically significant differences in susceptibility from antibiograms using chi-squared testing. CASSIA then proceeds to calculate clinical significant changes in susceptibility by propagating the maximum potential error of one report to multiple reports using the standard error formula. CASSIA is demonstrated on the Massachusetts statewide antibiogram data set. The benefits of CASSIA include immediate, consistent identification of clinically and statistically significant decreases in susceptibility without the need for an experienced domain expert.

Keywords: Deep learningBig data to knowledge, Personalized and pervasive health technologies, Personal health records
Abstract: Stacked Denoising Autoencoders (SDA) are deep networks which have gained popularity owing to their superior performance in image classification applications, but they haven’t been used much in healthcare applications. SDA can be efficiently retrained to adapt to large streams of data, and this property is used in this work to develop a technique for classification of arrhythmias in a patient-specific manner. This approach is particularly useful in continuous remote devices because they gather large amounts of data for longer periods of time. Blockchain is a decentralized distributed ledger which secures transactions with cryptography. It is proposed as an access control manager to securely store and access data required by the classifier during retraining in real-time from an external data storage. This work uses MIT-BIH Arrhythmia database and the results show an increased accuracy for Ventricular Ectopic Beats (VEB) (99.15%) and Supraventricular Ectopic Beats (SVEB) (98.55%), which is higher than published results for deep networks which are not retrained.

Keywords: Wearable systems and sensors, Machine learning, Telemedicine
Abstract: Clinical scales support diagnosis and therapy decisions. In the clinical workflow. In this paper, we present general considerations regarding wearable sensor-based clinical score definitions as well as our own work towards the development of an automatic score in Parkinson’s Disease (PD).

Keywords: Evaluation of health information systems, Clinical information systems, Public health management solutions
Abstract: The current work proposes a Benefits Realisation (BR) framework to identify, plan, deliver and measure potential benefits from the adoption of a plausible healthcare technology in clinical practice, wearables for gait assessment. BR is a recognised and documented concept in the UK-based National Health Service (NHS), yet regional NHS organisations (Trusts) have struggled to develop the required capability to successfully apply these concepts in practice. By critically investigating the actualities of BR it is argued that an approach with a central focus on benefits for patients and frontline staff is placed at the forefront of technology delivery.

Keywords: Mobile and wearable technologies for elderly, Personalized health/precision medicine, Wearable systems and sensors
Abstract: Recent advances in wireless, inexpensive sensor technology as well as in smart devices, is resulting in an explosion of miniature, portable devices that can potentially quantify mobility as well as large, expensive gait and balance laboratories. Very soon, practical, useful commercial systems, tailored to the needs of clinicians and therapists, will become available providing more precise, sensitive, and comprehensive evaluation of mobility in and beyond clinical settings.

Keywords: Information technologies for healthcare delivery and management, Public health management solutions, Multi-sensor data fusion
Abstract: It is clear that the recent advances in technology over the last decade have enabled step-change advances in healthcare delivery and in how health and wellness management is perceived by individuals. Ubiquitous technology – not necessarily primarily used for healthcare – is driving change, in behavior and in expectations. Being “connected” all the time, with information access at the fingertips brings with it societal changes and will continue to have profound effects on the future of health and wellness management. One issue remains, however, and that is how the data-deluge from connected devices is turned into meaningful and effective information that is used in the most effective manner possible.

Keywords: Health information systems, Behavioral informatics, Mobile health
Abstract: Access to mobile technologies such as smartphones and smartwatches, and the development and use of mobile health apps is growing rapidly. Many apps aim to change users’ behavior. Very few available apps are based on behavior change theory or are rigorously tested to ensure they bring about desired outcomes. Here we argue that we need to develop theory driven mHealth systems and rigorously evaluate implementation and effectiveness.

Keywords: IOT – internet of things, Health risk evaluation and modeling, Decision support systems
Abstract: Undoubtedly new bioinformatics findings and models as well as the Internet of Things questions the way healthcare services are currently offered. The disruptive nature of these technologies is driving innovation and makes an impact to societal challenges. This paper postulates that disruption should also take place in education and training of health professionals, if the empowerment brought about by innovative technologies is to be maintained and exploited. Use case are provided as examples of good practice.

Keywords: Behavioral health informatics, Behavioral informatics, Health information systems
Abstract: Emotion suppression is a form of conscious inhibition of emotional expressive behavior when a person's emotions are aroused. In human-to-human interaction, problems occur when speakers suppress their emotions; their internal affective states cannot be expressed to the interlocutor. We propose a new recognition technique of emotional suppression that humans may not be able to perceive. In this study, to detect the suppression of negative emotions, we used functional near-infrared spectroscopy and measured cerebral blood flow while participants watched neutral and disgust emotions and then talked about them. For the time-series data of cerebral blood flow, we performed a statistical test to confirm whether there is a significant difference in the mean values and the changes in the cerebral blood flow between each emotion and applied classification models to automatically detect suppressed emotions. Our models correctly detected them with around 60% accuracy.

Keywords: Emerging IT for efficient/low-cost healthcare delivery, Behavioral health informatics, Mobile and wearable technologies for elderly
Abstract: Falling is a major cause of death among elders. To detect falls, numerous approaches have been proposed in the past decade, including computer-vision based image processing, wearable sensors, and acoustic signal processing, etc. Though the recent advancement in infra-red LED, depth camera, MEMS (Micro-Electro-Mechanical Systems) sensors and machine learning algorithms may have enlarged the ap- plication scope, the privacy intrusion and lack of convenience still remain to be open issues which prevent these systems from large deployment. A novel Sensing-Fi system described in this paper aims to overcome such deficiencies. The system design is comprised of harnessing Wi-Fi Channel State Information (CSI) coupled with ground-mounted accelerometer to detect floor vibration, hence the system is completely passive and non- invasive, i.e., the user is not required to wear the technology contrary to wearable accelerometers. In addition, contrary to the existing standalone Wi-Fi CSI fall detection systems, it overcomes the constraint by which only one user can be present in the room. The system has shown promising results with 95% accuracy.

Keywords: Personalized health/precision medicine, Behavioral informatics, Physiological monitoring
Abstract: This paper investigates variations in head pose features in response to specific stressors. A proper experiment consisting of neutral and stressful states was performed aiming to cover different types of stress affect. Then, features related to head movements and pose were computationally estimated and analyzed. Towards this direction, facial landmarks were fitted using Active Appearance Models (AAM). Using the 2D AAM facial landmarks, a 3D head pose model was estimated revealing head inclinations. Results indicate that specific stress conditions increase head mobility and mobility velocity, in both translational and rotational features. Even though stress modulates head movements and velocities, the most prominent increases are presented during tasks that include participant’s speech. The degree and the intensity of the interaction effect between speech and stress should be investigated in more detail. The analysis reports that specific head pose features can be significant stress indicators that could contribute among other facial cues in reliable stress recognition.

Keywords: Behavioral health informatics, Machine learning, Sensor-based mHealth applications
Abstract: This paper describes a new approach for diagnosing anxiety and depression in young children. Currently, diagnosis in this population requires hours of structured clinical interviews spread over days and weeks. In contrast, we propose the use of a 90-second fear induction task during which time participant motion is monitoring using a commercially available wearable sensor. Machine learning and data extracted from one 20-second phase of the task are used to predict diagnosis in a large sample of children with and without an internalizing diagnosis. We examine the performance of a variety of feature sets and model configurations to identify the best performing approach that provides a diagnostic accuracy of 75%. This accuracy is comparable to existing diagnostic techniques, but at a small fraction of the time and cost currently required. These results point toward the future use of this approach in a clinical setting for diagnosing children with internalizing disorders.

Keywords: Machine learning, Wearable systems and sensors, Behavioral health informatics
Abstract: Actigraphy offers a low-cost alternative to conventional polysomnography (PSG) for screening of sleep-wake patterns. Effective use of actigraphy signals requires reliable methods for detecting sleep-wake states from actigraphy measurements. Hence, there is a growing interest in machine learning methods for training predictive models of sleep-wake states from actigraphy data. Existing work has focused on training a single predictive model for the entire population. However, accounting for individual differences, such as age, biological factors, or lifestyle-related variations, calls for personalized models for reliable identification of sleep-wake states from actigraphy data. This study investigates whether personalized models, trained on individual data, can match the performance of generalized models trained on population data. Using a dataset of 54 individuals, we systematically trained and tested personalized and generalized sleep-wake detectors developed using five commonly used machine learning algorithms. Results of our experiments show that personalized sleep-wake predictors are competitive, in terms of their predictive performance, with their generalized counterparts. Our work demonstrates the feasibility of developing reliable personalized sleep-wake states predictors from actigraphy data. This study lays the groundwork for developing personalized models for sleep-wake states detection that are better equipped to handle individual differences.

Keywords: Personal health systems, Human factors (ergonomics) in health information systems, Health data acquisition, transmission, management and visualization
Abstract: The rates of childhood obesity are at an all-time high, both domestically and around the globe. In an effort to stem current trends, researchers have looked to understand the components of effective and sustainable interventions. Although obesity interventions have often focused on a single lifestyle factors such as activity or nutrition, recent focus has shifted to a broader set of interventions which address multiple aspects of a child's life concurrently. Yet, despite promising outcomes, an understanding of what drives the success or failure of an intervention for a specific child is not yet well understood. To this end we have designed FitSpace, a web-based health and wellness platform designed to help understand how goal setting, activity, and nutrition patterns are tied to a child's social-demographic profile and their overall success in achieving sustainable healthy behavior. The work presented here represents the first step in our research, detailing a usability pilot study of the platform at a local high school.

Keywords: Ontology, Data storage, Health data acquisition, transmission, management and visualization
Abstract: Fully exposing, integrating, linking and exploring health data will have a tremendous impact on improving the integrated diagnosis, treatment and prevention of disease in individuals. In addition, it will allow for the secondary use of healthcare data for research transforming the way in which care is provided. To this direction, this paper describes the approach adapted for managing large and heterogeneous health datasets within the iManageCancer EU project. The project, has the objective to provide a cancer specific self-management platform designed according to the needs of patient groups while in parallel focusing on the wellbeing of the cancer patient with special emphasis on avoiding, early detecting and managing adverse events of cancer therapy but also, importantly, on the psycho-emotional evaluation and self-motivated goals.

Keywords: Healthcare modeling and simulation, Personalized health/precision medicine, Systems Biology
Abstract: The need to connect the research for the generation of new knowledge and its exploitation to improve human health has long been identified. We present a technological platform to facilitate such a bridge between the computational modelling for cancer research and the delivery of care for cancer patients. The focus is on the exploitation of in silico cancer models for clinical decision support and treatment selection but the presented methodology and technical architecture is applicable to other domains too.

Keywords: Personal health records, Information technologies for healthcare delivery and management, Computer games for healthcare
Abstract: Basic research and prospective clinical trials have led to higher cure rates of patients with cancer. Cancer is now frequently managed as a chronic disease. There is an increasing need for cancer patients to take an active, informed and leading role in their ongoing care to improve their physical, psychological, and social aspects of health thus resulting in a better quality of life. Identification of self-management processes for cancer can help to guide future research and clinical practice to improve patient’s outcome. As a chronic disease, there is also a need for the healthcare system to develop efficient strategies in supporting these patients. Educating patients to self-management of disease strengthens health behaviours in all aspects. Advances in ICT, together with the recent spread of portable devices, offer the opportunity to re-design self-management. iManageCancer (iMC) will provide the latest in self-management technology to empower people living with cancer going from a passive recipient of health care services to an active, informed participant in charge of his own well-being and that of the family in case of children. In iMC a Personal Health Record platform (iPHR) is developed, were different tools for self-management are available including serious games for children and adults. Data security and privacy are part of the platform. Currently pilots for children and adults are running. iPHR together with its tools and the serious games will be demonstrated.

Keywords: Computational modeling and simulations in biology, physiology and medicine, Disease profiling and personalized treatment
Abstract: Integrated systems modeling frameworks are useful in clinical cancer modeling due to their ability to incorporate a wide variety of patient data and tumor heterogeneity. Due to differences in time and length scales of individual processes, such an integration is a challenging task. Here we have combined ErbB receptor mediated Ras-MAPK and PI3K/AKT with p53 mediated DNA damage pathways to generate patient specific predictions for different cancer types.

Keywords: Computational modeling and simulations in biology, physiology and medicine, Organ/system/disease-level informatics (e.g. physiome, neuro-informatics, cancer-informatics, and cardiovascular-informatics platforms)
Abstract: In the framework of the large scale EU-US integrating research project CHIC entitled: “CHIC-Computational Horizons in Cancer: Developing Meta- and Hyper-Multiscale Models and Respositories for In Silico Oncology” that was funded by the European Commission, a number of highly innovative meta- and hyper-multiscale models and repositories for in silico oncology were developed, clinically adapted and partly clinically validated. Complex technological systems supporting and facilitating the process of hypermodel building and execution aiming at treatment personalization were also developed and successfully tested.

Keywords: Health information system interoperability, Knowledge modeling, Decision support methods and systems
Abstract: The successful implementation and adoption of AI applications and analytics pipelines in the medical domain often require a combination of computerized tasks and tasks carried out by human actors (to provide feedback, for model validation, etc.). The coordination of the two types of tasks in current practice is often ad-hoc, impacting both the efficiency of the processes and the quality of the results. We propose a workflow framework that facilitates the structuring of analytics processes into formal workflow models and streamlines their execution to ensure efficiency, predictability and to avoid errors.

Keywords: IOT – internet of things, Big Data analytics, Wearable and assistive devices for rehabilitation
Abstract: IEEE BHI 2018, March 4-7, 2018, Session #158 Title “Disruptive use cases empowered by IoT and Collaborative Technologies”

Abstract With rapid advancement in technology, medical device evolution along with collaborative efforts on the part of medical community, new use cases emerged that were not possible before, precision results are now feasible and the desire to design with a common platform and ecosystem will make it possible for scalable and sustainable solutions.

A panel of multi-disciplinary subject matter experts will present a number of unique use cases empowered by IoT and collaborative technologies such as wearables, blockchain, tattoo and Big Data/AI applied to healthcare.

The experts will share their vision, best of practices and challenges in advancing healthcare with paradigm shift from Sensing to Analytics as a new wave in the Digital Healthcare evolution.

Session Chair/Moderator: Fawzi Behmann, President, TelNet Management Consulting, Inc.

Speakers/Presentation Title: Speaker 1: Fawzi Behmann, President, TelNet Management Consulting, Inc.

Speaker 2: Hope Young, MT-BC, President/Owner, Center for Music Therapy

Speaker 3: Shideh Kabiri Ameri, postdoctoral fellow in the University of Texas at Austin

Speaker 4: Richard Hurley, President, GovernanceChain.com

Speaker 5: Ajay Bhargava, CEO, Analytics Advisory Group

Keywords: Big Data analytics, Decision support systems, Health data acquisition, transmission, management and visualization
Abstract: “Data & Analytics: Transformational Journey for largest primary care clinic in Central Texas”

The Medical Home department at Austin Regional Clinic (ARC) embarked on a transformational update involving people, process, technology, and data. With a new and improved Data Warehouse (DWH) and an Analytics Platform, a trained staff with updated skills, ARC has laid a strong foundation for improving clinical, operational, and financial outcomes. In this talk, I will describe how ARC partnered with Analytics Advisory Group to accelerate their data & analytics journey. In addition, will elaborate on a) The holistic approach to accelerate the transformation b) Challenges faced on the way, and steps taken to mitigate them c) Best Practices and Lessons learnt d) Strong data architecture and foundations for Analytics in the future e) Benefits & Value delivered to ARC ecosystem

Keywords: Wearable systems and sensors, Sensor-based mHealth applications
Abstract: During recent years, interests in development of sensors for Internet of things (IoT) have increased. Electronic tattoos (e-tattoos) are a group of newly developed wearables with the thickness range of few to hundreds of micrometer and they can be laminated on skin and be used as display or sensors with applications in e-health, e-skin and human machine interface and IoT. Here we present the thinnest, multimodal, mechanically and optically imperceptible graphene based electronic tattoo (GET) sensors. The overall thickness of GET is less than 500 nm which results in conformal contact to the microscopic features of skin, low electrode skin interface, low susceptibility to motion artifact and high signal to noise ratio. Because of low areal mass density of GET, it attaches to skin without the use of any adhesive or tape just by Van der Waals which make is desirable for application in the area with dedicate skin such as face and around eyes. GET has been successfully used for electroencephalography (EEG), Electrocardiography (ECG), electromyography (EMG) and skin temperature and hydration sensing. Also the application of GET for human machine interface has been demonstrated through using GET for electrooculography (EOG). The electrical signa

Keywords: Health information systems, Informatics for chronic disease management, Information technologies for healthcare delivery and management
Abstract: BHI-BSN 2018 - Special Invited Session # 158

Session Title: “Disruptive use cases empowered by IoT and Collaborative Technologies” Extended – One-Page Abstract Speaker 2 Hope Young, MT-BC, President/Owner, Center for Music Therapy, Inc./Executive Producer, Movement Tracks Project. hope.bass@gmail.com Sub-Title: “Movement Tracks: Moving towards greater user impact and potentials within local and global IoT ecosystem” Sub- Title: “Smarter Steps: The Movement Tracks Project” Abstract: Attendees will be shown use cases of combined music and IOT technology’s as a mobility and health solution for children with Cerebral Palsy and adults with Parkinson’s disease. The presentation will demonstrate music technology coupled with wearable advanced predictive, precise and personalized analytics as a means to reduce falls risks while increasing opportunities for users to increase their active mobility options within an IOT Smart Health/Mobility Ecosystem. These use cases will cover successful local and global integration of music into legacy health technology’s, EMR’s, health information management and healthcare service delivery systems. BIO: The Center for Music Therapy, Inc. was founded by Ms. Young in 1990 and is the first for-profit music therapy facility in the world specifically designed to research and treat neurologic movement conditions and disorders through music.

Keywords: Computational modeling and simulations in biology, physiology and medicine, Computational Biology, 3D visualization
Abstract: In this study we analysed computational model for plaque development and validation with clinical study. Mass transfer within the blood lumen and through the arterial wall was coupled with the blood flow. Plaque progression was modeled using three additional reaction–diffusion partial differential equations. The results for plaque concentration for the right coronary artery are presented. Specific patient biomarkers for baseline and follow up of 6 years were used. Plaque accumulation simulations and measurements were in correlation with low shear stress zones.

Keywords: Computational modeling and simulations in biology, physiology and medicine, Computational Biology, 3D visualization
Abstract: Mass transport within heart is process which occurs through two different domains: networks of blood vessels and surrounding tissue. Each medium is very complex and modeling of drug transport remains a challenge. Recently introduced composite smeared finite element (CSFE) offers new possibilities in computational modeling drug delivery in complex organs such as heart tissue. The main achievement in the CSFE is that a discrete transport (approximately 1D) within capillary system is transformed into a continuum framework. Distribution of drug in heart tissue which is calculated using inlet concentrations of drug in coronary arteries offers new insight drug delivery which is important for efficient treatment of heart diseases.

Keywords: Medical image processing and visualization, Image analysis, processing and classification, Algorithms
Abstract: This paper presents combination of two dimensional original and preprocessed brain tumor computerized tomography (CT) medical images with three-dimensional volumetric model. The methodology included performing segmentation on images using one of the available or newly developed algorithms from 37 patients. User-friendly two- and three-dimensional visualization is proposed. The results obtained in this paper can be new paradigm in the surgical assistance tools and brain tumor treatment.

Keywords: Computational Biology, Computational modeling and simulations in biology, physiology and medicine, Systems Biology
Abstract: We developed computational methodology to evaluate safety, efficacy and long-term performance of peripheral vascular stents. This methodology would allow evaluation under loading conditions truly representative of those in peripheral arteries and commercialization for new peripheral vascular stents without bench-top tests. Initial results present stress distribution during stent deployment procedure.

Keywords: Healthcare modeling and simulation, 3D visualization, Patient tracking
Abstract: The main purpose of this study was to computationally examine the biomechanical benefits within Abdominal Aortic Aneurysm (AAA) after stent graft (SG) implantation. The patient-specific 3D model was created on the basis of CT scan images. The study included finite element (FE) analysis of blood flow through implanted SG and the interaction with SG, intraluminal thrombus (ILT) and aortic wall. The benefits of SG implantation were analyzed by determining the Von Mises stress in modeled SG and aortic wall.

Keywords: Personalized health/precision medicine, Dynamic modeling of biomolecular regulatory and signaling networks, Model building for molecular, cellular and organ pathways and networks
Abstract: Nanoconstructs are particle-based nano-scale systems designed for the specific delivery of therapeutic and imaging agents. The Laboratory of Nanotechnology for Precision Medicine at IIT-GE synthesizes polymeric nanoconstructs presenting different sizes, ranging from a few tens of nanometers to a few microns; shapes, including spherical, elliptical and discoidal; surface properties, with positive, negative, neutral coatings; and softness, varying from that of cells to several tens of MPa. This talk will present a multi-scale hierarchical model for predicting the transport behavior of nanoconstructs within the blood vessels and the extravascular space.

Keywords: Mobile health, Pervasive health, Physiological monitoring
Abstract: Social anxiety disorder affects approximately 7% of the adult population in the U.S., yet a vast majority of these individuals do not seek treatment. Thus, it is critical to examine models that deliver treatment to them. Computerized Cognitive Bias Modification (CBM) training programs can be effective in targeting interpretation bias, a key cognitive mechanism underlying social anxiety, and have potential for widespread dissemination, especially if they can be delivered via smart phones, which are becoming ubiquitous. However, the efficacy of CBM interpretation training paradigms that are adapted to and delivered via smart phones remains unknown. We present a pilot study to investigate if physiologic data can be used to track the changes over a smartphone-based CBM intervention for social anxiety. In a 3-week open trial, pilot study involving 20 high socially anxious participants, self-report affect ratings, heart rate and accelerometer data were collected using a smartphone and smartwatch before, after, and during the CBM intervention. The study focused on the relationship between accelerometer and heart rate to track change following the intervention. Results provide preliminary evidence for the viability of using physiological data to identify the change in mental state influenced by CBM interventions.

Keywords: Mobile health, Telehealth, Cloud computing for healthcare
Abstract: Mobile cloud computing (MCC) has been extensively studied to provide pervasive healthcare services in a more affordable manner. Through offloading computation-intensive tasks from mobile to cloud, a significant portion of energy can be saved to extend the mobile battery life, which is critical to maintaining continuous and uninterrupted healthcare services. However, given the ever-changing clinical severity, personal demands, and environmental conditions, it is essential to explore context-aware approach capable of dynamically determining the optimal task offloading strategies and algorithmic settings, with the goal of achieving a balanced trade-off among energy efficiency, diagnostic accuracy, and processing latency. To this aim, we propose a model-free reinforcement learning based task scheduling approach to adapt to the changing requirements.

Keywords: Patient tracking, Wearable systems and sensors, Unsupervised learning method
Abstract: Motor alteration is an important aspect of the elusive schizophrenia disorder, manifested both throughout the various phases of the disease and as a response to treatment. Tracking of patients' movement, and especially in a closed ward hospital setting, can therefore shed light on the dynamics of the disease, and help alert staff to possible deterioration and adverse effects of medication. In this paper we describe the use of ARIMA-based anomaly detection for monitoring of patient motor activity in a closed ward hospital setting. We demonstrate the utility of the approach in several intriguing case studies.

Keywords: Machine learning, Mobile health, Pervasive health
Abstract: Negative affect is a proxy for mental health in adults. By being able to predict participants' negative affect states unobtrusively, researchers and clinicians will be better positioned to deliver targeted, just-in-time mental health interventions via mobile applications. This work attempts to personalize the passive recognition of negative affect states via group-based modeling of user behavior patterns captured from mobility, communication, and activity patterns. Results show that group models outperform generalized models in a dataset based on two weeks of users' daily lives.

Keywords: Personal health systems, Mobile health, Technology and services for assisted-living and elderly
Abstract: Automated real-time detection of tonic-clonic seizures in a home environment can be done using different modalities. The algorithms from the literature have not been evaluated nor compared on the same dataset. In this study, 3 seizure detection algorithms using electrocardiogram, electromyogram and accelerometers are evaluated on a single dataset. In this dataset, 7 pediatric patients with tonic-clonic seizures are monitored during 224 hours using 7 sensors in total. All unimodal algorithms are evaluated and compared to each other using this dataset. The different unimodal algorithms are also combined using a late integration approach. The best unimodal algorithm was found to be the algorithm using the right wrist accelerometer, leading to 95.5% sensitivity and 0.70 false alarms per hour. The best combination of sensors was found to be electrocardiogram with the right ankle accelerometer with a sensitivity of 90.9% and 0.08 false alarms per hour. The results show that all combinations of multimodal sensors lead to at least 75% less false alarms, showing that such multimodal algorithms should be used in practice for tonic-clonic seizure detection in pediatric patients.

Keywords: Machine learning, Sensor-based mHealth applications, Algorithms
Abstract: Atrial Fibrillation (AFib) is one of the most common cardia arrhythmia and can potentially progress to serious illness. Early detection and prevention of Afib can bring huge benefits to Afib affected population. With the gain in popularity of wearable devices such as smartwatches, it is possible to utilize the devices build-in photo-plethysmography (PPG) sensor and computational power to provide a portable, non-intrusive and low-cost solution for Afib monitoring and detection to general population. However, utilizing signal data collected from wearable device is challenging due to various types of noise affecting the signal quality. We cope with this challenge by proposing an end-to-end deep learning AFib detection system that can filter out poor quality signals and make reliable predictions. Our models achieves over 95% AUC in quality assessment task and over 99% AUC in AFib detection task while having a reasonable size.

Keywords: Big Data analytics, Computational genotyping, Deep learningBig data to knowledge
Abstract: Growing concerns regarding increasing rates of antimicrobial resistance (AMR) call for rapid diagnosis. Next-generation sequence and deep learning offers the potential for rapid near-same day estimation of the prediction of AMR. In this paper, we developed deep learning model to prediction AMR of Mycobacterium tuberculosis (MTB). The proposed model addressed resistance co-occurrence of MTB within the first-line anti-tuberculosis (TB) drugs and discovered subtypes of AMR based on genomic data.

Keywords: Personalized health/precision medicine, Unsupervised learning method, Physiological monitoring
Abstract: With the rapid increase in volume of wearable devices, automated algorithms are employed to label physiological time-series data. However, the algorithms lack reliability due to large inter- and intra- subject variabilities. In real-world scenarios, where only unevenly-sampled continuous or numeric estimates are provided, it is difficult to choose which algorithms to trust or discard, or even how to merge their recommendations to form a final estimate that is precise for an individual. We proposed two parametric fully-Bayesian graphical models for fusing labels from (i) independent and (ii) potentially-correlated algorithms. These unsupervised models aggregate labels and estimate jointly the assumed bias and precision of each algorithm. The fused estimate may then be used to infer the underlying ground truth for an individual. We show that modelling the latent correlations between algorithms improves these estimates with respect to commonly-employed strategies in the literature.

Keywords: Machine learning, Deep learningBig data to knowledge, Healthcare modeling and simulation
Abstract: We will present multiple ways in which healthcare data is acquired and machine learning methods are currently being introduced into clinical settings. This will include: 1) Modeling the prediction of disease, including Sepsis, and ways in which the best treatment decisions for Sepsis patients can be made, from electronic health record (EHR) data using Gaussian processes and deep learning methods. 2) Predicting surgical complications and transfer learning methods for combining databases and 3) Using mobile apps and integrated sensors for improving the granularity of recorded health data for chronic conditions. Current work in these areas will be presented and the future of machine learning contributions to the field will be discussed.

Keywords: Machine learning, Computer-aided decision making, Artificial Intelligence
Abstract: Angiectasia, formerly called angiodysplasia, is one of the most frequent vascular lesions and often the cause of gastrointestinal bleedings. Medical specialists assessing videos of examinations reach a detection performance of 16% for the detection of bleeding to 69% for the detection of angiectasia. In this paper, we present several machine-learningbased approaches for angiectasia detection in wireless video capsule endoscopy images. The most promising results for pixel-wise localization and frame-wise detection are obtained by the proposed deep learning approach using generative adversarial networks (GANs) with a sensitivity of 88% and specificity of 99.9% for pixel-wise localization and a sensitivity of 98% and a specificity of 100% for frame-wise detection, which fits the requirements for automatic angiectasia detection in real clinical settings.

Keywords: Deep learningBig data to knowledge, Computational phenotyping, Machine learning
Abstract: Title: Doctor AI - Interpretable Deep Learning Methods for modeling Electronic Health Records Abstract: Deep neural networks provide great potential to create better models for longitudinal electronic health records (EHRs). In this talk, we will present a series of case studies of deep learning examples for modeling EHR. 1) We illustrate how recurrent neural networks (RNN) can be used to model temporal relations among events in electronic health records (EHRs) to predict heart failures. 2) We present Med2Vec, which not only learns the representations for both medical codes and visits from large EHR datasets with over million visits, but also allows us to interpret the learned representations confirmed positively by clinical experts. 3) We introduce an interpretable predictive model RETAIN which achieves high accuracy while remaining clinically interpretable and is based on a two-level neural attention model that detects influential past visits and significant clinical variables within those visits (e.g. key diagnoses). 4) We propose GRaph-based Attention Model (GRAM) that supplements electronic health records (EHR) with hierarchical information inherent to medical ontologies. Based on the data volume and the ontology structure, GRAM represents a medical concept as a combination of its ancestors in the ontology via an attention mechanism. 5) We present a new approach, medical Generative Adversarial Network (medGAN), to generate realistic synthetic patient records. Bio: Jimeng

Keywords: Deep learningBig data to knowledge, Electronic health records, Machine learning
Abstract: It is widely believed that deep learning and artificial intelligence techniques will fundamentally change healthcare industries in the future. Even though recent development in deep learning has achieved significant successes in many applications, such as computer vision, natural language processing, speech recognition and so on, healthcare applications pose a series of significantly different challenges to existing models. Examples include but not are limited to interpretations for prediction, heterogeneity in data, missing value, multi-rate multi-resolution data, big and small data, and privacy issues. In this paper, we will discuss a series of unique challenges in healthcare that can benefit from deep learning models, as well as recent advances in addressing them. We will also discuss benchmarking results of deep learning models on MIMIC III, the largest publicly available medical dataset.