Gari D. Clifford

Kshama Kodthalu Shivashankara, Deepanshi, Afagh Mehri Shervedani et al.

Cardiovascular diseases are a major cause of mortality globally, and electrocardiograms (ECGs) are crucial for diagnosing them. Traditionally, ECGs are printed on paper. However, these printouts, even when scanned, are incompatible with advanced ECG diagnosis software that require time-series data. Digitizing ECG images is vital for training machine learning models in ECG diagnosis and to leverage the extensive global archives collected over decades. Deep learning models for image processing are promising in this regard, although the lack of clinical ECG archives with reference time-series data is challenging. Data augmentation techniques using realistic generative data models provide a solution. We introduce ECG-Image-Kit, an open-source toolbox for generating synthetic multi-lead ECG images with realistic artifacts from time-series data. The tool synthesizes ECG images from real time-series data, applying distortions like text artifacts, wrinkles, and creases on a standard ECG paper background. As a case study, we used ECG-Image-Kit to create a dataset of 21,801 ECG images from the PhysioNet QT database. We developed and trained a combination of a traditional computer vision and deep neural network model on this dataset to convert synthetic images into time-series data for evaluation. We assessed digitization quality by calculating the signal-to-noise ratio (SNR) and compared clinical parameters like QRS width, RR, and QT intervals recovered from this pipeline, with the ground truth extracted from ECG time-series. The results show that this deep learning pipeline accurately digitizes paper ECGs, maintaining clinical parameters, and highlights a generative approach to digitization. This toolbox currently supports data augmentation for the 2024 PhysioNet Challenge, focusing on digitizing and classifying paper ECG images.

Matthew A. Reyna, Deepanshi, James Weigle et al.

We introduce the ECG-Image-Database, a large and diverse collection of electrocardiogram (ECG) images generated from ECG time-series data, with real-world scanning, imaging, and physical artifacts. We used ECG-Image-Kit, an open-source Python toolkit, to generate realistic images of 12-lead ECG printouts from raw ECG time-series. The images include realistic distortions such as noise, wrinkles, stains, and perspective shifts, generated both digitally and physically. The toolkit was applied to 977 12-lead ECG records from the PTB-XL database and 1,000 from Emory Healthcare to create high-fidelity synthetic ECG images. These unique images were subjected to both programmatic distortions using ECG-Image-Kit and physical effects like soaking, staining, and mold growth, followed by scanning and photography under various lighting conditions to create real-world artifacts. The resulting dataset includes 35,595 software-labeled ECG images with a wide range of imaging artifacts and distortions. The dataset provides ground truth time-series data alongside the images, offering a reference for developing machine and deep learning models for ECG digitization and classification. The images vary in quality, from clear scans of clean papers to noisy photographs of degraded papers, enabling the development of more generalizable digitization algorithms. ECG-Image-Database addresses a critical need for digitizing paper-based and non-digital ECGs for computerized analysis, providing a foundation for developing robust machine and deep learning models capable of converting ECG images into time-series. The dataset aims to serve as a reference for ECG digitization and computerized annotation efforts. ECG-Image-Database was used in the PhysioNet Challenge 2024 on ECG image digitization and classification.

Mircea Dumitru, Qiao Li, Erick Andres Perez Alday et al.

Objective: Gaussian Processes (GP)-based filters, which have been effectively used for various applications including electrocardiogram (ECG) filtering can be computationally demanding and the choice of their hyperparameters is typically ad hoc. Methods: We develop a data-driven GP filter to address both issues, using the notion of the ECG phase domain -- a time-warped representation of the ECG beats onto a fixed number of samples and aligned R-peaks, which is assumed to follow a Gaussian distribution. Under this assumption, the computation of the sample mean and covariance matrix is simplified, enabling an efficient implementation of the GP filter in a data-driven manner, with no ad hoc hyperparameters. The proposed filter is evaluated and compared with a state-of-the-art wavelet-based filter, on the PhysioNet QT Database. The performance is evaluated by measuring the signal-to-noise ratio (SNR) improvement of the filter at SNR levels ranging from -5 to 30dB, in 5dB steps, using additive noise. For a clinical evaluation, the error between the estimated QT-intervals of the original and filtered signals is measured and compared with the benchmark filter. Results: It is shown that the proposed GP filter outperforms the benchmark filter for all the tested noise levels. It also outperforms the state-of-the-art filter in terms of QT-interval estimation error bias and variance. Conclusion: The proposed GP filter is a versatile technique for preprocessing the ECG in clinical and research applications, is applicable to ECG of arbitrary lengths and sampling frequencies, and provides confidence intervals for its performance.

Mohsen Motie-Shirazi, Reza Sameni, Peter Rohloff et al.

In this study, we present a deep learning framework designed to integrate with our previously developed system that facilitates large-scale 1D fetal Doppler data collection, aiming to enhance data quality. This system, tailored for traditional Indigenous midwives in low-resource communities, leverages a cost-effective Android phone to improve the quality of recorded signals. We have shown that the Doppler data can be used to identify fetal growth restriction, hypertension, and other concerning issues during pregnancy. However, the quality of the signal is dependent on many factors, including radio frequency interference, position of the fetus, maternal body habitus, and usage of the Doppler by the birth attendants. In order to provide instant feedback to allow correction of the data at source, a signal quality metric is required that can run in real-time on the mobile phone. In this study, 191 DUS signals with durations mainly in the range between 5 to 10 minutes were evaluated for quality and classified into five categories: Good, Poor, (Radiofrequency) Interference, Talking, and Silent, at a resolution of 3.75 seconds. A deep neural network was trained on each 3.75-second segment from these recordings and validated using five-fold cross-validation. An average micro F1 = 97.4\% and macro F1 = 94.2\% were achieved, with F1 = 99.2\% for `Good' quality data. These results indicate that the algorithm, which will now be implemented in the midwives' app, should allow a significant increase in the quality of data at the time of capture.

Barun Das, Conor Anderson, Tania Villavicencio et al.

Rapid identification and accurate documentation of interfering and high-risk behaviors in ASD, such as aggression, self-injury, disruption, and restricted repetitive behaviors, are important in daily classroom environments for tracking intervention effectiveness and allocating appropriate resources to manage care needs. However, having a staff dedicated solely to observing is costly and uncommon in most educational settings. Recently, multiple research studies have explored developing automated, continuous, and objective tools using machine learning models to quantify behaviors in ASD. However, the majority of the work was conducted under a controlled environment and has not been validated for real-world conditions. In this work, we demonstrate that the latest advances in video-based group activity recognition techniques can quantify behaviors in ASD in real-world activities in classroom environments while preserving privacy. Our explainable model could detect the episode of problem behaviors with a 77% F1-score and capture distinctive behavior features in different types of behaviors in ASD. To the best of our knowledge, this is the first work that shows the promise of objectively quantifying behaviors in ASD in a real-world environment, which is an important step toward the development of a practical tool that can ease the burden of data collection for classroom staff.

Seyedeh Somayyeh Mousavi, Matthew A. Reyna, Gari D. Clifford et al.

Regular blood pressure (BP) monitoring in clinical and ambulatory settings plays a crucial role in the prevention, diagnosis, treatment, and management of cardiovascular diseases. Recently, the widespread adoption of ambulatory BP measurement devices has been driven predominantly by the increased prevalence of hypertension and its associated risks and clinical conditions. Recent guidelines advocate for regular BP monitoring as part of regular clinical visits or even at home. This increased utilization of BP measurement technologies has brought up significant concerns, regarding the accuracy of reported BP values across settings. In this survey, focusing mainly on cuff-based BP monitoring technologies, we highlight how BP measurements can demonstrate substantial biases and variances due to factors such as measurement and device errors, demographics, and body habitus. With these inherent biases, the development of a new generation of cuff-based BP devices which use artificial-intelligence (AI) has significant potential. We present future avenues where AI-assisted technologies can leverage the extensive clinical literature on BP-related studies together with the large collections of BP records available in electronic health records. These resources can be combined with machine learning approaches, including deep learning and Bayesian inference, to remove BP measurement biases and to provide individualized BP-related cardiovascular risk indexes.

Alireza Rafiei, Anahí Venzor Strader, Esteban Castro Aragón et al.

Hypertensive disorders of pregnancy remain a leading cause of maternal and fetal morbidity worldwide, yet diagnosis relies on intermittent cuff-based blood pressure measurements that are prone to bias and fail to capture continuous physiological dynamics. Growing evidence suggests that fetal cardiovascular activity is associated with maternal-placental hemodynamics and may encode markers of maternal hypertension. To analyze this, we collected a large-scale dataset of fetal one-dimensional Doppler ultrasound recordings paired with maternal blood pressure from 3,255 pregnant women across 8,170 antenatal visits in rural Guatemala. We developed AutoHyPE, a hierarchical attention network that models short- and long-term signal structure, incorporating a novel prototype-based contrastive learning and multi-view strategy to enhance representation robustness under long-tailed class distribution and biological variability. AutoHyPE achieved an AUROC of 0.80 for maternal hypertension detection, outperforming baseline approaches while maintaining balanced performance across classes, with no performance degradation in an edge deployment scenario. Our findings demonstrated that fetal cardiac mechanical activity contains hemodynamic features indicative of maternal hypertension status. This supports a promising paradigm shift toward continuous, objective monitoring of maternal health using existing, low-cost ultrasound technology and introduces a complementary approach to traditional methods based on blood pressure measurements, advancing scalable prenatal care.

Yadhu Kartha, Conor Anderson, Jenny Foster et al.

Autism Spectrum Disorder (ASD) is characterized by challenges with social interaction and communication and by restricted or repetitive patterns of thought and behavior, with significant variability in presentation. Approximately a quarter of children with ASD are classified as having profound autism, who often exhibit challenging behaviors, such as self-injurious behavior, aggression, elopement, or pica, that pose serious safety risks and disrupt learning in educational settings. Prior work has applied wearable sensors and machine learning to detect challenging behaviors, but has been largely confined to controlled laboratory environments. This work demonstrates that predicting challenging behavior episodes is feasible in a real-world special education classroom. We collected approximately 110.7 hours of labeled multimodal wearable data comprising accelerometry, electrodermal activity (EDA), and skin temperature from 9 children and young adults aged 10 to 21 years across standard classroom sessions. We fine-tuned state-of-the-art foundation models for multimodal wearable time-series analysis and show that challenging behavior episodes can be predicted up to 10 minutes in advance with an AUC-ROC of 0.78. These results establish a concrete foundation for developing proactive in-class intervention systems that enable teachers to minimize the safety risks of challenging behaviors in special education classrooms

Sepideh Nikookar, Edward Tian, Harrison Hoffman et al.

The accelerometer has become an almost ubiquitous device, providing enormous opportunities in healthcare monitoring beyond step counting or other average energy estimates in 15-60 second epochs. Objective: To develop an open data set with associated open-source code for processing 50 Hz tri-axial accelerometry-based to classify patient activity levels and natural types of movement. Approach: Data were collected from 23 healthy subjects (16 males and seven females) aged between 23 and 62 years using an ambulatory device, which included a triaxial accelerometer and synchronous lead II equivalent ECG for an average of 26 minutes each. Participants followed a standardized activity routine involving five distinct activities: lying, sitting, standing, walking, and jogging. Two classifiers were constructed: a signal processing technique to distinguish between high and low activity levels and a convolutional neural network (CNN)-based approach to classify each of the five activities. Main results: The binary (high/low) activity classifier exhibited an F1 score of 0.79. The multi-class CNN-based classifier provided an F1 score of 0.83. The code for this analysis has been made available under an open-source license together with the data on which the classifiers were trained and tested. Significance: The classification of behavioral activity, as demonstrated in this study, offers valuable context for interpreting traditional health metrics and may provide contextual information to support the future development of clinical decision-making tools for patient monitoring, predictive analytics, and personalized health interventions.

Gari D. Clifford

Over the last 35 years, the sharing of medical data and models for research has evolved from sneakernet to the internet - from mailing magnetic tapes and compact discs of a handful of well-curated recordings, to the high-speed download of relatively comprehensive hospital databases. More recently, the fervor around the potential for modern machine learning and 'AI' to catapult us into the next industrial revolution has led to a seemingly insatiable desire to pump almost any source of data into large models. Although this has great potential, it also presents a whole set of new challenges. In this article I examine these trends over the last 30 years, drawing on examples from cardiology, one of the oldest data-intensive fields that is undergoing a renaissance via machine learning. From the early days of computerized cardiology, the Research Resource for Complex Physiologic Signals (PhysioNet) has been at the cutting edge of this field. This article, therefore, includes much of the Resource's history and the contributions drawn from 25 years of firsthand experience of co-developing elements of the Resource with its founders. I identify the most promising future directions for the PhysioNet Resource, and more generally, the growing issues and opportunities around dissemination and use of massive physiological databases, associated open access code, and public competitions, along with potential solutions to the key issues facing our field. Topics range from how we should approach foundation models in the context of the rapidly growing AI carbon footprint, to the potential of Tiny-ML and edge computing. I also cover issues around prizes and incentives, funding models, and scientific repeatability, as well as how we might address these issues by leveraging the PhysioNet Challenges, consistent with the philosophy of open-access from the early days of the PhysioNet Resource.

Matthew A. Reyna, Zuzana Koscova, Jan Pavlus et al.

Objective: Chagas disease is a parasitic infection that is endemic to South America, Central America, and, more recently, the U.S., primarily transmitted by insects. Chronic Chagas disease can cause cardiovascular diseases and digestive problems. Serological testing capacities for Chagas disease are limited, but Chagas cardiomyopathy often manifests in ECGs, providing an opportunity to prioritize patients for testing and treatment. Approach: The George B. Moody PhysioNet Challenge 2025 invites teams to develop algorithmic approaches for identifying Chagas disease from electrocardiograms (ECGs). Main results: This Challenge provides multiple innovations. First, we leveraged several datasets with labels from patient reports and serological testing, provided a large dataset with weak labels and smaller datasets with strong labels. Second, we augmented the data to support model robustness and generalizability to unseen data sources. Third, we applied an evaluation metric that captured the local serological testing capacity for Chagas disease to frame the machine learning problem as a triage task. Significance: Over 630 participants from 111 teams submitted over 1300 entries during the Challenge, representing diverse approaches from academia and industry worldwide.

Sajjad Karimi, Shirin Karimi, Amit J. Shah et al.

Cardiovascular diseases are best diagnosed using multiple modalities that assess both the heart's electrical and mechanical functions. While effective, imaging techniques like echocardiography and nuclear imaging are costly and not widely accessible. More affordable technologies, such as simultaneous electrocardiography (ECG) and phonocardiography (PCG), may provide valuable insights into electromechanical coupling and could be useful for prescreening in low-resource settings. Using physical stress test data from the EPHNOGRAM ECG-PCG dataset, collected from 23 healthy male subjects (age: 25.4+/-1.9 yrs), we investigated electromechanical intervals (RR, QT, systolic, and diastolic) and their interactions during exercise, along with hysteresis between cardiac electrical activity and mechanical responses. Time delay analysis revealed distinct temporal relationships between QT, systolic, and diastolic intervals, with RR as the primary driver. The diastolic interval showed near-synchrony with RR, while QT responded to RR interval changes with an average delay of 10.5s, and the systolic interval responded more slowly, with an average delay of 28.3s. We examined QT-RR, systolic-RR, and diastolic-RR hysteresis, finding narrower loops for diastolic RR and wider loops for systolic RR. Significant correlations (average:0.75) were found between heart rate changes and hysteresis loop areas, suggesting the equivalent circular area diameter as a promising biomarker for cardiac function under exercise stress. Deep learning models, including Long Short-Term Memory and Convolutional Neural Networks, estimated the QT, systolic, and diastolic intervals from RR data, confirming the nonlinear relationship between RR and other intervals. Findings highlight a significant cardiac memory effect, linking ECG and PCG morphology and timing to heart rate history.

Nasim Katebi, Mohammad Ahmad, Mohsen Motie-Shirazi et al.

Perinatal complications, defined as conditions that arise during pregnancy, childbirth, and the immediate postpartum period, represent a significant burden on maternal and neonatal health worldwide. Factors contributing to these disparities include limited access to quality healthcare, socioeconomic inequalities, and variations in healthcare infrastructure. Addressing these issues is crucial for improving health outcomes for mothers and newborns, particularly in underserved communities. To mitigate these challenges, we have developed an AI-enabled smartphone application designed to provide decision support at the point-of-care. This tool aims to enhance health monitoring during pregnancy by leveraging machine learning (ML) techniques. The intended use of this application is to assist midwives during routine home visits by offering real-time analysis and providing feedback based on collected data. The application integrates TensorFlow Lite (TFLite) and other Python-based algorithms within a Kotlin framework to process data in real-time. It is designed for use in low-resource settings, where traditional healthcare infrastructure may be lacking. The intended patient population includes pregnant women and new mothers in underserved areas and the developed system was piloted in rural Guatemala. This ML-based solution addresses the critical need for accessible and quality perinatal care by empowering healthcare providers with decision support tools to improve maternal and neonatal health outcomes.

Weitao Tang, Johann Vargas-Calixto, Nasim Katebi et al.

Fetal sleep is a relatively underexplored yet vital aspect of prenatal neurodevelopment. Understanding fetal sleep patterns could provide insights into early brain maturation and help clinicians detect signs of neurological compromise that arise due to fetal hypoxia or fetal growth restriction. This review synthesizes over eight decades of research on the physiological characteristics, ontogeny, and regulation of fetal sleep. We compare sleep-state patterns in humans and large animal models, highlighting species-specific differences and the presence of sleep-state analogs. We review both invasive techniques in animals and non-invasive modalities in humans. Computational methods for sleep-state classification are also examined, including rule-based approaches (with and without clustering-based preprocessing) and state-of-the-art deep learning techniques. Finally, we discuss how intrauterine conditions such as hypoxia and fetal growth restriction can disrupt fetal sleep. This review provides a comprehensive foundation for the development of objective, multimodal, and non-invasive fetal sleep monitoring technologies to support early diagnosis and intervention in prenatal care.

Xiaofan Mu, Merna Bibars, Salman Seyedi et al.

The aging society urgently requires scalable methods to monitor cognitive decline and identify social and psychological factors indicative of dementia risk in older adults. Our machine learning (ML) models captured facial, acoustic, linguistic, and cardiovascular features from 39 older adults with normal cognition or Mild Cognitive Impairment (MCI), derived from remote video conversations and quantified their cognitive status, social isolation, neuroticism, and psychological well-being. Our model could distinguish Clinical Dementia Rating Scale (CDR) of 0.5 (vs. 0) with 0.77 area under the receiver operating characteristic curve (AUC), social isolation with 0.74 AUC, social satisfaction with 0.75 AUC, psychological well-being with 0.72 AUC, and negative affect with 0.74 AUC. Our feature importance analysis showed that speech and language patterns were useful for quantifying cognitive impairment, whereas facial expressions and cardiovascular patterns were useful for quantifying social and psychological well-being. Our bias analysis showed that the best-performing models for quantifying psychological well-being and cognitive states in older adults exhibited significant biases concerning their age, sex, disease condition, and education levels. Our comprehensive analysis shows the feasibility of monitoring the cognitive and psychological health of older adults, as well as the need for collecting largescale interview datasets of older adults to benefit from the latest advances in deep learning technologies to develop generalizable models across older adults with diverse demographic backgrounds and disease conditions.

Weitao Tang, Johann Vargas-Calixto, Nasim Katebi et al.

Introduction: This study presents FetalSleepNet, the first published deep learning approach to classifying sleep states from the ovine electroencephalogram (EEG). Fetal EEG is complex to acquire and difficult and laborious to interpret consistently. However, accurate sleep stage classification may aid in the early detection of abnormal brain maturation associated with pregnancy complications (e.g. hypoxia or intrauterine growth restriction). Methods: EEG electrodes were secured onto the ovine dura over the parietal cortices of 24 late gestation fetal sheep. A lightweight deep neural network originally developed for adult EEG sleep staging was trained on the ovine EEG using transfer learning from adult EEG. A spectral equalisation-based domain adaptation strategy was used to reduce cross-domain mismatch. Results: We demonstrated that while direct transfer performed poorly, full fine tuning combined with spectral equalisation achieved the best overall performance (accuracy: 86.6 percent, macro F1-score: 62.5), outperforming baseline models. Conclusions: To the best of our knowledge, FetalSleepNet is the first deep learning framework specifically developed for automated sleep staging from the fetal EEG. Beyond the laboratory, the EEG-based sleep stage classifier functions as a label engine, enabling large scale weak/semi supervised labeling and distillation to facilitate training on less invasive signals that can be acquired in the clinic, such as Doppler Ultrasound or electrocardiogram data. FetalSleepNet's lightweight design makes it well suited for deployment in low power, real time, and wearable fetal monitoring systems.

Sajjad Karimi, Amit J. Shah, Gari D. Clifford et al.

Simultaneous electrocardiography (ECG) and phonocardiogram (PCG) provide a comprehensive, multimodal perspective on cardiac function by capturing the heart's electrical and mechanical activities, respectively. However, the distinct and overlapping information content of these signals, as well as their potential for mutual reconstruction and biomarker extraction, remains incompletely understood, especially under varying physiological conditions and across individuals. In this study, we systematically investigate the common and exclusive characteristics of ECG and PCG using the EPHNOGRAM dataset of simultaneous ECG-PCG recordings during rest and exercise. We employ a suite of linear and nonlinear machine learning models, including non-causal LSTM networks, to reconstruct each modality from the other and analyze the influence of causality, physiological state, and cross-subject variability. Our results demonstrate that nonlinear models, particularly non-causal LSTM, provide superior reconstruction performance, with reconstructing ECG from PCG proving more tractable than the reverse. Exercise and cross-subject scenarios present significant challenges, but envelope-based modeling that utilizes instantaneous amplitude features substantially improves cross-subject generalizability for cross-modal learning. Furthermore, we demonstrate that clinically relevant ECG biomarkers, such as fiducial points and QT intervals, can be estimated from PCG in cross-subject settings. These findings advance our understanding of the relationship between electromechanical cardiac modalities, in terms of both waveform characteristics and the timing of cardiac events, with potential applications in novel multimodal cardiac monitoring technologies.

Alireza Rafiei, Gari D. Clifford, Nasim Katebi

Fetal health monitoring through one-dimensional Doppler ultrasound (DUS) signals offers a cost-effective and accessible approach that is increasingly gaining interest. Despite its potential, the development of machine learning based techniques to assess the health condition of mothers and fetuses using DUS signals remains limited. This scarcity is primarily due to the lack of extensive DUS datasets with a reliable reference for interpretation and data imbalance across different gestational ages. In response, we introduce a novel autoregressive generative model designed to map fetal electrocardiogram (FECG) signals to corresponding DUS waveforms (Auto-FEDUS). By leveraging a neural temporal network based on dilated causal convolutions that operate directly on the waveform level, the model effectively captures both short and long-range dependencies within the signals, preserving the integrity of generated data. Cross-subject experiments demonstrate that Auto-FEDUS outperforms conventional generative architectures across both time and frequency domain evaluations, producing DUS signals that closely resemble the morphology of their real counterparts. The realism of these synthesized signals was further gauged using a quality assessment model, which classified all as good quality, and a heart rate estimation model, which produced comparable results for generated and real data, with a Bland-Altman limit of 4.5 beats per minute. This advancement offers a promising solution for mitigating limited data availability and enhancing the training of DUS-based fetal models, making them more effective and generalizable.

Ayse Cakmak, Erik Reinertsen, Shamim Nemati et al.

The pattern of state changes in a biomedical time series can be related to health or disease. This work presents a principled approach for selecting a changepoint detection algorithm for a specific task, such as disease classification. Eight key algorithms were compared, and the performance of each algorithm was evaluated as a function of temporal tolerance, noise, and abnormal conduction (ectopy) on realistic artificial cardiovascular time series data. All algorithms were applied to real data (cardiac time series of 22 patients with REM-behavior disorder (RBD) and 15 healthy controls) using the parameters selected on artificial data. Finally, features were derived from the detected changepoints to classify RBD patients from healthy controls using a K-Nearest Neighbors approach. On artificial data, Modified Bayesian Changepoint Detection algorithm provided superior positive predictive value for state change identification while Recursive Mean Difference Maximization (RMDM) achieved the highest true positive rate. For the classification task, features derived from the RMDM algorithm provided the highest leave one out cross validated accuracy of 0.89 and true positive rate of 0.87. Automatically detected changepoints provide useful information about subject's physiological state which cannot be directly observed. However, the choice of change point detection algorithm depends on the nature of the underlying data and the downstream application, such as a classification task. This work represents the first time change point detection algorithms have been compared in a meaningful way and utilized in a classification task, which demonstrates the effect of changepoint algorithm choice on application performance.

Hyeokhyen Kwon, Chaitra Hegde, Yashar Kiarashi et al.

Camera-based activity monitoring systems are becoming an attractive solution for smart building applications with the advances in computer vision and edge computing technologies. In this paper, we present a feasibility study and systematic analysis of a camera-based indoor localization and multi-person tracking system implemented on edge computing devices within a large indoor space. To this end, we deployed an end-to-end edge computing pipeline that utilizes multiple cameras to achieve localization, body orientation estimation and tracking of multiple individuals within a large therapeutic space spanning $1700m^2$, all while maintaining a strong focus on preserving privacy. Our pipeline consists of 39 edge computing camera systems equipped with Tensor Processing Units (TPUs) placed in the indoor space's ceiling. To ensure the privacy of individuals, a real-time multi-person pose estimation algorithm runs on the TPU of the computing camera system. This algorithm extracts poses and bounding boxes, which are utilized for indoor localization, body orientation estimation, and multi-person tracking. Our pipeline demonstrated an average localization error of 1.41 meters, a multiple-object tracking accuracy score of 88.6\%, and a mean absolute body orientation error of 29\degree. These results shows that localization and tracking of individuals in a large indoor space is feasible even with the privacy constrains.

Ismail Sadiq, Erick A. Perez-Alday, Amit J. Shah et al.

Objective: To determine if a realistic, but computationally efficient model of the electrocardiogram can be used to pre-train a deep neural network (DNN) with a wide range of morphologies and abnormalities specific to a given condition - T-wave Alternans (TWA) as a result of Post-Traumatic Stress Disorder, or PTSD - and significantly boost performance on a small database of rare individuals. Approach: Using a previously validated artificial ECG model, we generated 180,000 artificial ECGs with or without significant TWA, with varying heart rate, breathing rate, TWA amplitude, and ECG morphology. A DNN, trained on over 70,000 patients to classify 25 different rhythms, was modified the output layer to a binary class (TWA or no-TWA, or equivalently, PTSD or no-PTSD), and transfer learning was performed on the artificial ECG. In a final transfer learning step, the DNN was trained and cross-validated on ECG from 12 PTSD and 24 controls for all combinations of using the three databases. Main results: The best performing approach (AUROC = 0.77, Accuracy = 0.72, F1-score = 0.64) was found by performing both transfer learning steps, using the pre-trained arrhythmia DNN, the artificial data and the real PTSD-related ECG data. Removing the artificial data from training led to the largest drop in performance. Removing the arrhythmia data from training provided a modest, but significant, drop in performance. The final model showed no significant drop in performance on the artificial data, indicating no overfitting. Significance: In healthcare, it is common to only have a small collection of high-quality data and labels, or a larger database with much lower quality (and less relevant) labels. The paradigm presented here, involving model-based performance boosting, provides a solution through transfer learning on a large realistic artificial database, and a partially relevant real database.

Salman Seyedi, Zifan Jiang, Allan Levey et al.

The expanding usage of complex machine learning methods like deep learning has led to an explosion in human activity recognition, particularly applied to health. In particular, as part of a larger body sensor network system, face and full-body analysis is becoming increasingly common for evaluating health status. However, complex models which handle private and sometimes protected data, raise concerns about the potential leak of identifiable data. In this work, we focus on the case of a deep network model trained on images of individual faces. Full-face video recordings taken from 493 individuals undergoing an eye-tracking based evaluation of neurological function were used. Outputs, gradients, intermediate layer outputs, loss, and labels were used as inputs for a deep network with an added support vector machine emission layer to recognize membership in the training data. The inference attack method and associated mathematical analysis indicate that there is a low likelihood of unintended memorization of facial features in the deep learning model. In this study, it is showed that the named model preserves the integrity of training data with reasonable confidence. The same process can be implemented in similar conditions for different models.

Ayse S. Cakmak, Samuel Densen, Gabriel Najarro et al.

Objective: Worldwide, heart failure (HF) is a major cause of morbidity and mortality and one of the leading causes of hospitalization. Early detection of HF symptoms and pro-active management may reduce adverse events. Approach: Twenty-eight participants were monitored using a smartphone app after discharge from hospitals, and each clinical event during the enrollment (N=110 clinical events) was recorded. Motion, social, location, and clinical survey data collected via the smartphone-based monitoring system were used to develop and validate an algorithm for predicting or classifying HF decompensation events (hospitalizations or clinic visit) versus clinic monitoring visits in which they were determined to be compensated or stable. Models based on single modality as well as early and late fusion approaches combining patient-reported outcomes and passive smartphone data were evaluated. Results: The highest AUCPr for classifying decompensation with a late fusion approach was 0.80 using leave one subject out cross-validation. Significance: Passively collected data from smartphones, especially when combined with weekly patient-reported outcomes, may reflect behavioral and physiological changes due to HF and thus could enable prediction of HF decompensation.

Salman Seyedi, Li Xiong, Shamim Nemati et al.

The increasing complexity of algorithms for analyzing medical data, including de-identification tasks, raises the possibility that complex algorithms are learning not just the general representation of the problem, but specifics of given individuals within the data. Modern legal frameworks specifically prohibit the intentional or accidental distribution of patient data, but have not addressed this potential avenue for leakage of such protected health information. Modern deep learning algorithms have the highest potential of such leakage due to complexity of the models. Recent research in the field has highlighted such issues in non-medical data, but all analysis is likely to be data and algorithm specific. We, therefore, chose to analyze a state-of-the-art free-text de-identification algorithm based on LSTM (Long Short-Term Memory) and its potential in encoding any individual in the training set. Using the i2b2 Challenge Data, we trained, then analyzed the model to assess whether the output of the LSTM, before the compression layer of the classifier, could be used to estimate the membership of the training data. Furthermore, we used different attacks including membership inference attack method to attack the model. Results indicate that the attacks could not identify whether members of the training data were distinguishable from non-members based on the model output. This indicates that the model does not provide any strong evidence into the identification of the individuals in the training data set and there is not yet empirical evidence it is unsafe to distribute the model for general use.

Matthew A. Reyna, Gari D. Clifford

Although there has been significant research in boosting of weak learners, there has been little work in the field of boosting from strong learners. This latter paradigm is a form of weighted voting with learned weights. In this work, we consider the problem of constructing an ensemble algorithm from 70 individual algorithms for the early prediction of sepsis from clinical data. We find that this ensemble algorithm outperforms separate algorithms, especially on a hidden test set on which most algorithms failed to generalize.

Ayse S. Cakmak, Nina Thigpen, Garrett Honke et al.

Depression and post-traumatic stress disorder (PTSD) are psychiatric conditions commonly associated with experiencing a traumatic event. Estimating mental health status through non-invasive techniques such as activity-based algorithms can help to identify successful early interventions. In this work, we used locomotor activity captured from 1113 individuals who wore a research grade smartwatch post-trauma. A convolutional variational autoencoder (VAE) architecture was used for unsupervised feature extraction from four weeks of actigraphy data. By using VAE latent variables and the participant's pre-trauma physical health status as features, a logistic regression classifier achieved an area under the receiver operating characteristic curve (AUC) of 0.64 to estimate mental health outcomes. The results indicate that the VAE model is a promising approach for actigraphy data analysis for mental health outcomes in long-term studies.

Xingyao Wang, Chengyu Liu, Yuwen Li et al.

Objective: This paper presents a novel heart sound segmentation algorithm based on Temporal-Framing Adaptive Network (TFAN), including state transition loss and dynamic inference for decoding the most likely state sequence. Methods: In contrast to previous state-of-the-art approaches, the TFAN-based method does not require any knowledge of the state duration of heart sounds and is therefore likely to generalize to non sinus rhythm. The TFAN-based method was trained on 50 recordings randomly chosen from Training set A of the 2016 PhysioNet/Computer in Cardiology Challenge and tested on the other 12 independent training and test databases (2099 recordings and 52180 beats). The databases for segmentation were separated into three levels of increasing difficulty (LEVEL-I, -II and -III) for performance reporting. Results: The TFAN-based method achieved a superior F1 score for all 12 databases except for `Test-B', with an average of 96.7%, compared to 94.6% for the state-of-the-art method. Moreover, the TFAN-based method achieved an overall F1 score of 99.2%, 94.4%, 91.4% on LEVEL-I, -II and -III data respectively, compared to 98.4%, 88.54% and 79.80% for the current state-of-the-art method. Conclusion: The TFAN-based method therefore provides a substantial improvement, particularly for more difficult cases, and on data sets not represented in the public training data. Significance: The proposed method is highly flexible and likely to apply to other non-stationary time series. Further work is required to understand to what extent this approach will provide improved diagnostic performance, although it is logical to assume superior segmentation will lead to improved diagnostics.

Justus T. C. Schwabedal, John C. Snyder, Ayse Cakmak et al.

Randomizing the Fourier-transform (FT) phases of temporal-spatial data generates surrogates that approximate examples from the data-generating distribution. We propose such FT surrogates as a novel tool to augment and analyze training of neural networks and explore the approach in the example of sleep-stage classification. By computing FT surrogates of raw EEG, EOG, and EMG signals of under-represented sleep stages, we balanced the CAPSLPDB sleep database. We then trained and tested a convolutional neural network for sleep stage classification, and found that our surrogate-based augmentation improved the mean F1-score by 7%. As another application of FT surrogates, we formulated an approach to compute saliency maps for individual sleep epochs. The visualization is based on the response of inferred class probabilities under replacement of short data segments by partial surrogates. To quantify how well the distributions of the surrogates and the original data match, we evaluated a trained classifier on surrogates of correctly classified examples, and summarized these conditional predictions in a confusion matrix. We show how such conditional confusion matrices can qualitatively explain the performance of surrogates in class balancing. The FT-surrogate augmentation approach may improve classification on noisy signals if carefully adapted to the data distribution under analysis.

Supreeth P. Shashikumar, Amit J. Shah, Gari D. Clifford et al.

Detection of atrial fibrillation (AF), a type of cardiac arrhythmia, is difficult since many cases of AF are usually clinically silent and undiagnosed. In particular paroxysmal AF is a form of AF that occurs occasionally, and has a higher probability of being undetected. In this work, we present an attention based deep learning framework for detection of paroxysmal AF episodes from a sequence of windows. Time-frequency representation of 30 seconds recording windows, over a 10 minute data segment, are fed sequentially into a deep convolutional neural network for image-based feature extraction, which are then presented to a bidirectional recurrent neural network with an attention layer for AF detection. To demonstrate the effectiveness of the proposed framework for transient AF detection, we use a database of 24 hour Holter Electrocardiogram (ECG) recordings acquired from 2850 patients at the University of Virginia heart station. The algorithm achieves an AUC of 0.94 on the testing set, which exceeds the performance of baseline models. We also demonstrate the cross-domain generalizablity of the approach by adapting the learned model parameters from one recording modality (ECG) to another (photoplethysmogram) with improved AF detection performance. The proposed high accuracy, low false alarm algorithm for detecting paroxysmal AF has potential applications in long-term monitoring using wearable sensors.

Tingting Zhu, Nic Dunkley, Joachim Behar et al.

With the rapid increase in volume of time series medical data available through wearable devices, there is a need to employ automated algorithms to label data. Examples of labels include interventions, changes in activity (e.g. sleep) and changes in physiology (e.g. arrhythmias). However, automated algorithms tend to be unreliable resulting in lower quality care. Expert annotations are scarce, expensive, and prone to significant inter- and intra-observer variance. To address these problems, a Bayesian Continuous-valued Label Aggregator(BCLA) is proposed to provide a reliable estimation of label aggregation while accurately infer the precision and bias of each algorithm. The BCLA was applied to QT interval (pro-arrhythmic indicator) estimation from the electrocardiogram using labels from the 2006 PhysioNet/Computing in Cardiology Challenge database. It was compared to the mean, median, and a previously proposed Expectation Maximization (EM) label aggregation approaches. While accurately predicting each labelling algorithm's bias and precision, the root-mean-square error of the BCLA was 11.78$\pm$0.63ms, significantly outperforming the best Challenge entry (15.37$\pm$2.13ms) as well as the EM, mean, and median voting strategies (14.76$\pm$0.52ms, 17.61$\pm$0.55ms, and 14.43$\pm$0.57ms respectively with $p<0.0001$).