Optical Coherence Tomography (OCT) is essential in ophthalmology for cross-sectional imaging of the retina. Pretrained foundation models facilitate task-specific model development by enabling fine-tuning with limited labeled data. However, current foundation models rely on a single B-scan (usually the central slice), overlooking volumetric context. This research investigates video foundation models to capture full 3D retinal structure and improve diagnostic performance. V-JEPA, a state-of-the-art video foundation model, was benchmarked against retinal foundation models (RETFound, VisionFM) and a natural image foundation model (DINOv2). All were fine-tuned to detect Age-related Macular Degeneration or Glaucomatous Optic Neuropathy using five OCT datasets. V-JEPA consistently equaled or outperformed image-based models, achieving an average AUROC of 0.94 (0.80–0.99), versus 0.90 (0.76–0.98) for the best image model, a statistically significant improvement (p < 0.001). To our knowledge, this is the first application of transformer-based video models to volumetric OCT, highlighting their promise in 3D medical imaging.

Background: The integration of artificial intelligence into retinal practice represents more than a technological advancement; it constitutes an anthropological shift fundamentally redefining the centuries-old therapeutic partnership between physician and patient. Objective: To examine how artificial intelligence integration into retinal care transforms the therapeutic partnership, revealing what this transformation illuminates about the nature of medical knowledge itself and identifying frameworks for conscious implementation. Methods: We conducted a structured narrative review examining AI-based diagnosis and monitoring of diabetic retinopathy and age-related macular degeneration (2018–2025), synthesizing clinical deployment evidence with qualitative implementation studies, adopting an anthropological interpretive stance to examine technology as a mediator of human relationships. Results: FDA-approved AI systems demonstrate robust diagnostic performance for diabetic retinopathy and age-related macular degeneration. AI integration shifts encounters from examination-based to screen-mediated, clinical reasoning from individual to algorithm-guided, and physicians from diagnosticians to interpreters. Three insights emerge. First, AI reveals that medical practice always combined mechanistic reasoning with pattern recognition, now separated algorithmically. Second, accountability operates asymmetrically: while all stakeholders derive benefits from algorithmic integration, authority over system selection and deployment remains concentrated among vendors and institutions rather than distributed to frontline clinicians or patients. Third, impact diverges along existing stratification: transformation may create access for excluded populations while potentially eroding relationships for those who had comprehensive care, raising fundamental questions about equitable distribution. Conclusions: The AI transformation of retinal care offers a revealing mirror of medicine’s algorithmic future. Success demands epistemological rigor, robust evaluation competencies and establishing frameworks for shared accountability among the various stakeholders. Our framework maps six fundamental dimensions where synthesis supersedes substitution: expanding algorithmic capabilities while preserving healing relationships, creating access while maintaining continuity. Medicine can embrace algorithmic intelligence while preserving its humanistic core through conscious choices about epistemology, equity, and the character of practice we create.

Heart failure (HF) affects 11.8% of adults aged 65 and older, reducing quality of life and longevity. Preventing HF can reduce morbidity and mortality. We hypothesized that artificial intelligence (AI) applied to 24-hour single-lead electrocardiogram (ECG) data could predict the risk of HF within five years. To research this, the Technion-Leumit Holter ECG (TLHE) dataset, including 69,663 recordings from 47,729 patients, collected over 20 years, was used. Our deep learning model, DeepHHF, trained on 24-hour ECG recordings, achieved an area under the receiver operating characteristic curve of 0.80 that outperformed a model using 30-second segments and a clinical score. High-risk individuals identified by DeepHHF had a two-fold chance of hospitalization or death incidents. Explainability analysis showed DeepHHF focused on arrhythmias and heart abnormalities. This study highlights the feasibility of deep learning to model 24-hour continuous ECG data, capturing paroxysmal events essential for reliable risk prediction. Artificial intelligence applied to single-lead Holter ECG is non-invasive, inexpensive, and widely accessible, making it a promising tool for HF risk prediction.

Introduction: Premature Ventricular Contractions (PVCs) are common cardiac arrhythmias originating from the ventricles. Accurate detection remains challenging due to variability in electrocardiogram (ECG) waveforms caused by differences in lead placement, recording conditions, and population demographics. Methods: We developed uPVC-Net, a universal deep learning model to detect PVCs from any single-lead ECG recordings. The model is developed on four independent ECG datasets comprising a total of 8.3 million beats collected from Holter monitors and a modern wearable ECG patch. uPVC-Net employs a custom architecture and a multi-source, multi-lead training strategy. For each experiment, one dataset is held out to evaluate out-of-distribution (OOD) generalization. Results: uPVC-Net achieved an AUC between 97.8% and 99.1% on the held-out datasets. Notably, performance on wearable single-lead ECG data reached an AUC of 99.1%. Conclusion: uPVC-Net exhibits strong generalization across diverse lead configurations and populations, highlighting its potential for robust, real-world clinical deployment.

Objective. Cardiovascular disease remains one of the leading causes of morbidity and mortality worldwide, highlighting the need for accurate risk prediction. Atrial fibrillation burden (AFB), premature ventricular contraction burden (PVCB), and T-wave alternans (TWA) have been individually associated with adverse outcomes. We hypothesized that these three digital ECG biomarkers provide complementary prognostic information for major cardiovascular endpoints and all-cause mortality (ACM), and developed a machine learning approach combining these biomarkers for risk prediction. Approach. We analyzed 81 362 Holter recordings from 54 395 individuals across 20 primary care centers in Israel. A random forest model using AFB, PVCB, TWA, and age was trained to predict 5 year risk of heart failure (HF), ischemic stroke (IS), and ACM. Main results. On the test set, best AUROCs were for HF 0.75 [95% CI: 0.74–0.77] ( (Formula presented) (Formula presented) ), for IS 0.69 [0.66–0.71] ( (Formula presented) (Formula presented) ), and for ACM 0.79 [0.77–0.80] ( (Formula presented) (Formula presented) ). Combining the three ECG-derived biomarkers showed complementary predictive value and improved discrimination by up to 10% over age alone in individuals aged (Formula presented) (Formula presented) years. Significance. AFB, PVCB, and TWA contribute complementary prognostic information for risk stratification. When combined with age, these biomarkers improve predictive performance compared with age alone.

Objective. Self-supervised learning (SSL) has enabled vision transformers (ViTs) to learn robust representations from large-scale natural image datasets, enhancing their generalization across domains. In retinal imaging, foundation models pretrained on either natural or ophthalmic data have shown promise, but the benefits of in-domain pretraining remain uncertain.Approach. To investigate this, we benchmark six SSL-pretrained ViTs on seven digital fundus image (DFI) datasets totaling 70 000 expert-annotated images for the task of moderate-to-late age-related macular degeneration (AMD) identification.Main results. Our results show that DINOv2, pretrained on natural images, shows similar performance than domain-specific models. These findings highlight the value of foundation models in improving AMD identification, and challenge the assumption that in-domain pretraining is necessary.Significance. We present our model AMDNet, which performs state-of-the-art out-of-domain AUROCs on six public datasets. Furthermore, we release BRAMD, an open-access dataset (n = 587) of DFIs with AMD labels from Brazil. Project page:www.aimlab-technion.com/lirot-ai.

Home sleep apnea tests (HSATs) have emerged as alternatives to in-laboratory polysomnography (PSG), but Type IV HSATs often show limited diagnostic performance. This study clinically validates SleepAI, a novel remote digital health system that applies AI algorithms to raw oximetry data for automated sleep staging and obstructive sleep apnea (OSA) diagnosis. SleepAI algorithms were trained on over 10,000 PSG recordings. The system consists of a wearable oximeter connected via Bluetooth to a mobile app transmitting raw data to a cloud-based platform for AI-driven analysis. Clinical validation was conducted in 53 subjects with suspected OSA, who used SleepAI for three nights at home and one night in a sleep centre alongside PSG. SleepAI's apnea-hypopnea index (AHI) estimates and three-class sleep staging (Wake, REM, NREM) were compared to PSG references. For OSA severity classification (non-OSA, mild, moderate, severe), SleepAI achieved an overall accuracy of 89%, with F1-scores of 1.0, 1.0, 0.9, and 0.88, respectively. The three-stage sleep classification achieved a Cohen's kappa of 0.75. Night-to-night AHI variability showed that 37.5% of participants experienced a one-level severity change across nights at home. No significant differences in sleep metrics were found between the first and subsequent nights at home, indicating no sleep disturbance by SleepAI. These findings support the SleepAI system as a promising and scalable alternative to existing Type IV HSATs, with the potential to address key clinical gaps by improving diagnostic accuracy and accessibility.

Objective. Photoplethysmography, a non-invasive optical technique that measures changes in blood volume in the microvascular bed of tissue, offers a promising approach for monitoring physiological changes during sleep. This study evaluates differential photoplethysmography signal patterns that can distinguish between apneas vs hypopneas, which are key features of sleep-related breathing disorders. Approach. We analyzed data from 263 severe (apnea hypopnea index ⩾30) obstructive sleep apnea patients, using recordings from the Multi-Ethnic Study of Atherosclerosis. Over 57 000 respiratory events occurring during stage N2 sleep were included. A machine learning model was trained on 89 features derived from the photoplethysmography signal, using the pyPPG toolbox, to classify: apneas vs hypopneas in the supine and lateral sleep posture, and posture-specific differences for each respiratory event type. Main results. Results showed that photoplethysmography signal characteristics significantly differed between apneas vs hypopneas. The model achieved an area under the receiver operation characteristic curve of 0.80 in the lateral posture and 0.83 in the supine posture. However, classification performance was low when distinguishing between apneas and hypopneas in the lateral vs the supine position with an area under the receiver operation characteristic curve of 0.62 for apneas and 0.64 for hypopneas. The discriminative signal features were consistent across different periods of the night. Significance. These findings indicate that photoplethysmography can detect meaningful differences in sleep-related breathing events and support its potential as a foundation for wearable diagnostic and monitoring tools that are personalized, accessible, and cost-effective.

Glaucomatous optic neuropathy (GON), affecting an estimated 64.3 million people globally, causes irreversible vision loss when not detected early. Traditional diagnosis requires time-consuming ophthalmic examinations by specialists. Recent deep learning models for automating GON detection from colour fundus photographs (CFP) have shown promise but often suffer from limited generalizability across different ethnicities, disease groups and examination settings. To address these limitations, we introduce GONet, a robust deep learning model developed using seven independent datasets, including over 119 000 CFPs with gold-standard annotations and from patients of diverse geographic backgrounds. GONet consists of a DINOv2 pre-trained self-supervised vision transformer fine-tuned using a multisource domain strategy. GONet demonstrated high out-of-distribution generalizability, with an AUC of 0.88-0.99 in target domains.. GONet performance was similar or superior to state-of-the-art works and the cup-to-disc ratio, by up to 18.4%. GONet is available viaLirot.ai). We also contribute a new dataset consisting of 747 CFPs with GON labels as open access, available at https://doi.org/10.13026/pdxv-m215.

Objective. Deep learning for continuous physiological signals, such as electrocardiography or oximetry, has achieved remarkable success in supervised learning scenarios where training and testing data are drawn from the same distribution. However, when evaluating real-world applications, models often fail to generalize due to distribution shifts between the source domain on which the model was trained and the target domain where it is deployed. A common and particularly challenging shift often encountered in reality is where the source and target domain supports do not fully overlap. In this paper, we propose a novel framework, named Deep Unsupervised Domain adaptation using variable nEighbors (DUDE), to address this challenge. Approach. We introduce a new type of contrastive loss between the source and target domains using a dynamic neighbor selection strategy, in which the number of neighbors for each sample is adaptively determined based on the density observed in the latent space. We use multiple real-world datasets as source and target domains, with target domains that included demographics, ethnicities, geographies, and comorbidities that were not present in the source domain. Main results. The experimental results demonstrate superior DUDE performance compared to baselines and with an improvement of up to 16% over the original Nearest-Neighbor Contrastive Learning of Visual Representations strategy. Significance. Our contribution provides evidence on the potential of using DUDE to bridge the crucial gap of domain adaptation in medicine, potentially transforming patient care through more precise and adaptable diagnostic tools.

Objective. Large vessel occlusion (LVO) stroke presents a major challenge in clinical practice due to the potential for poor outcomes with delayed treatment. Treatment for LVO involves highly specialized care, in particular endovascular thrombectomy, and is available only at certain hospitals. Therefore, prehospital identification of LVO by emergency ambulance services, can be critical for triaging LVO stroke patients directly to a hospital with access to endovascular therapy. Clinical scores exist to help distinguish LVO from less severe strokes, but they are based on a series of examinations that can be time-consuming and may be impractical for patients with dementia or those who cannot follow commands due to their stroke. There is a need for a fast and reliable method to aid in the early identification of LVO. In this study, our objective was to assess the feasibility of using 30 s photoplethysmography (PPG) recording to assist in recognizing LVO stroke. Approach. A total of 88 patients, including 25 with LVO, 27 with stroke mimic (SM), and 36 non-LVO stroke patients (NL), were recorded at the Liverpool Hospital emergency department in Sydney, Australia. Demographics (age, sex), as well as morphological features and beating rate variability measures, were extracted from the PPG. A binary classification approach was employed to differentiate between LVO stroke and NL + SM (NL.SM). A 2:1 train-test split was stratified and repeated randomly across 100 iterations. Main results. The best model achieved a median test set area under the receiver operating characteristic curve of 0.77 (0.71–0.82). Significance. Our study demonstrates the potential of utilizing a 30 s PPG recording for identifying LVO stroke.

Excessive daytime sleepiness (EDS) refers to a physiological state where individuals have difficulty remaining alert during the day. Managing EDS is particularly challenging to study and treat due to its multifaceted nature. Assessment methods include both subjective and objective approaches. Subjective evaluation often relies on simple, widely accepted, and widely used questionnaires; however, these tools are inherently limited by self-reporting bias. Objective assessment, on the other hand, primarily involves two well-known and reliable tests, but these are costly, time-consuming, and impractical for use outside of sleep units. Therefore, developing an objective tool that can quickly and accurately detect a decline in alertness, while remaining reliable, easy to use, and affordable, is of critical importance for sleep clinicians, safety organizations, and researchers. According to PRISMA guidelines, we did a systematic analysis of 95 studies that used photoplethysmography (PPG) for assessing EDS, drowsiness, and/or fatigue during the last 15 years (2010–2025). With advances in wearable technology, particularly through PPG and artificial intelligence, achieving this goal may be attainable. The next essential step is rigorous validation against established gold-standard tests to ensure the tool meets scientific and clinical standards for widespread adoption.

Atrial fibrillation (AF) is a common atrial arrhythmia that impairs quality of life and causes embolic stroke, heart failure and other complications. Recent advancements in machine learning (ML) and deep learning (DL) have shown potential for enhancing diagnostic accuracy. It is essential for DL models to be robust and generalizable across variations in ethnicity, age, sex, and other factors. Although a number of ECG database have been made available to the research community, none includes a Japanese population sample. Saitama Heart Database Atrial Fibrillation (SHDB-AF) is a novel open-sourced Holter ECG database from Japan, containing 128 ECG with detailed clinical information from 122 unique patients. Each record in SHDB-AF is 24 hours long and has two channels, totaling 21.6 million seconds of ECG data. The dataset is available at https://physionet.org/content/shdb-af/.

Objective. sleep staging is essential for diagnosing sleep disorders and managing sleep health. Traditional methods require time-consuming manual scoring. Recent photoplethysmography (PPG)-based deep learning models perform well on local datasets but struggle with external generalization due to data drift. Approach. this study evaluates multi-source domain training for improving out-of-distribution generalization in four-class sleep staging (wake, light, deep, rapid eye movement) from raw PPG time-series. The trained deep learning model is denoted SleepPPG-Net2. Additionally, we examined the impact of demographic factors, ethnicity, and obstructive sleep apnea (OSA) on performance. SleepPPG-Net2 was benchmarked against two state-of-the-art models. Main results. SleepPPG-Net2 outperformed benchmark models, improving generalization performance (Cohen’s kappa) by up to 21%. Performance disparities were observed in relation to age, sex, and OSA severity. Significance. SleepPPG-Net2 enhances PPG-based sleep staging and provides insights into demographic and clinical influences on model performance.

Objective: To develop and validate an automated end-to-end methodology for analyzing retinal vasculature in large datasets of digital fundus images (DFIs), aiming to assess the influence of demographic and clinical factors on retinal microvasculature. Design: This study employs a retrospective cohort design to achieve its objectives. Participants: The research utilized a substantial dataset consisting of 32 768 DFIs obtained from individuals undergoing routine eye examinations. There was no inclusion of a separate control group in this study. Methods: The proposed methodology integrates multiple stages: initial image quality assessment, detection of the optic disc (OD), definition of the region of interest surrounding the OD, automated segmentation of retinal arterioles and venules, and the engineering of digital biomarkers representing vasculature characteristics. To analyze the impact of demographic variables (age, sex) and clinical factors (disc size, primary open-angle glaucoma [POAG]), statistical analyses were performed using linear mixed-effects models. Main Outcome Measures: The primary outcomes measured were changes in the retinal vascular geometry. Special attention was given to evaluating the independent effects of age, sex, disc size, and POAG on the newly engineered microvasculature biomarkers. Results: The analysis revealed significant independent similarities in the retinal vascular geometry alterations associated with both advanced age and POAG. These findings suggest a potential mechanism of accelerated vascular aging in patients with POAG. Conclusions: This novel methodology allows for the comprehensive and quantitative analysis of retinal vasculature, facilitating the investigation of its correlations with specific diseases. By enabling the reproducible analysis of extensive datasets, this approach provides valuable insights into the state of retinal vascular health and its broader implications for cardiovascular and ocular health. The software developed through this research will be made publicly available upon publication, offering a critical tool for ongoing and future studies in retinal vasculature. Financial Disclosure(s): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Introduction: The presence of fibrillatory waves (f-waves) is important in the diagnosis of atrial fibrillation (AF), which has motivated the development of methods for f-wave extraction. We propose a novel approach to benchmarking methods designed for single-lead ECG analysis, building on the hypothesis that better-performing AF classification using features computed from the extracted f-waves implies better-performing extraction. The approach is well-suited for processing large Holter data sets annotated with respect to the presence of AF. Methods: Three data sets with a total of 300 two- or three-lead Holter recordings, performed in the USA, Israel and Japan, were used as well as a simulated single-lead data set. Four existing extraction methods based on either average beat subtraction or principal component analysis (PCA) were evaluated. A random forest classifier was used for window-based AF classification. Performance was measured by the area under the receiver operating characteristic (AUROC). Results: The best performance was found for PCA-based extraction, resulting in AUROCs in the ranges 0.80–0.85, 0.66–0.80, and 0.87–0.92 for the data sets from USA, Israel, and Japan, respectively, when analyzed across leads; the AUROC of the simulated single-lead, noisy data set was 0.98. Conclusions: This study provides a novel approach to evaluating the performance of f-wave extraction methods, offering the advantage of not using ground truth f-waves for evaluation, thus being able to leverage real data sets for evaluation. The code is made open source at github.com/noambenmoshe/fwave.

Background: Sleep staging is critical for diagnosing sleep disorders. Traditional methods in clinical settings involve time-intensive scoring procedures. Recent advancements in data-driven algorithms using photoplethysmogram (PPG) time series have shown promise in automating sleep staging in adults. However, for children, algorithm development is hindered by the limited availability of datasets, with the Childhood Adenotonsillectomy Trial (CHAT) being the only substantial source, comprising recordings from children aged 5-10. This limitation constrains the evaluation of algorithmic generalization performance. Methods: We employed a deep learning model for sleep staging from PPG, initially trained using a large dataset of adult sleep recordings, and fine-tuned it on 80% of the CHAT dataset (CHAT-train) for the task of three-class sleep staging (wake, REM, non-REM). The resulting algorithm performance was compared to the same model architecture but trained from scratch on CHAT-train (benchmark). The algorithms are evaluated on the local test set, denoted CHAT-test, as well as on a newly introduced independent dataset. Results: Our deep learning algorithm achieved a Cohen's Kappa of 0.88 on CHAT-test (versus 0.65), and demonstrated generalization capabilities with a Kappa of 0.72 on the external Ichilov dataset for children above 5 years old (versus 0.64) and 0.64 for those below 5 (versus 0.53). Significance: This research establishes a new state-of-the-art performance for the task of sleep staging in children using raw PPG. The findings underscore the value of transfer learning from the adults to children domain. However, the reduced performance in children under 5 suggests the need for further research and additional datasets covering a broader pediatric age range to fully address generalization limitations.

Objective. Diabetic retinopathy (DR) is a serious diabetes complication that can lead to vision loss, making timely identification crucial. Existing data-driven algorithms for DR staging from digital fundus images (DFIs) often struggle with generalization due to distribution shifts between training and target domains. Approach. To address this, DRStageNet, a deep learning model, was developed using six public and independent datasets with 91 984 DFIs from diverse demographics. Five pretrained self-supervised vision transformers (ViTs) were benchmarked, with the best further trained using a multi-source domain (MSD) fine-tuning strategy. Main results. DINOv2 showed a 27.4% improvement in L-Kappa versus other pretrained ViT. MSD fine-tuning improved performance in four of five target domains. The error analysis revealing 60% of errors due to incorrect labels, 77.5% of which were correctly classified by DRStageNet. Significance. We developed DRStageNet, a DL model for DR, designed to accurately stage the condition while addressing the challenge of generalizing performance across target domains. The model and explainability heatmaps are available at www.aimlab-technion.com/lirot-ai.

The Leuven-Haifa dataset contains 240 disc-centered fundus images of 224 unique patients (75 patients with normal tension glaucoma, 63 patients with high tension glaucoma, 30 patients with other eye diseases and 56 healthy controls) from the University Hospitals of Leuven. The arterioles and venules of these images were both annotated by master students in medicine and corrected by a senior annotator. All senior segmentation corrections are provided as well as the junior segmentations of the test set. An open-source toolbox for the parametrization of segmentations was developed. Diagnosis, age, sex, vascular parameters as well as a quality score are provided as metadata. Potential reuse is envisioned as the development or external validation of blood vessels segmentation algorithms or study of the vasculature in glaucoma and the development of glaucoma diagnosis algorithms. The dataset is available on the KU Leuven Research Data Repository (RDR).

Objective. This study aims to automate the segmentation of retinal arterioles and venules (A/V) from digital fundus images (DFI), as changes in the spatial distribution of retinal microvasculature are indicative of cardiovascular diseases, positioning the eyes as windows to cardiovascular health. Approach. We utilized active learning to create a new DFI dataset with 240 crowd-sourced manual A/V segmentations performed by 15 medical students and reviewed by an ophthalmologist. We then developed LUNet, a novel deep learning architecture optimized for high-resolution A/V segmentation. The LUNet model features a double dilated convolutional block to widen the receptive field and reduce parameter count, alongside a high-resolution tail to refine segmentation details. A custom loss function was designed to prioritize the continuity of blood vessel segmentation. Main Results. LUNet significantly outperformed three benchmark A/V segmentation algorithms both on a local test set and on four external test sets that simulated variations in ethnicity, comorbidities and annotators. Significance. The release of the new datasets and the LUNet model (www.aimlab-technion.com/lirot-ai) provides a valuable resource for the advancement of retinal microvasculature analysis. The improvements in A/V segmentation accuracy highlight LUNet's potential as a robust tool for diagnosing and understanding cardiovascular diseases through retinal imaging.

Objective. Photoplethysmography is a non-invasive optical technique that measures changes in blood volume within tissues. It is commonly and being increasingly used for a variety of research and clinical applications to assess vascular dynamics and physiological parameters. Yet, contrary to heart rate variability measures, a field which has seen the development of stable standards and advanced toolboxes and software, no such standards and limited open tools exist for continuous photoplethysmogram (PPG) analysis. Consequently, the primary objective of this research was to identify, standardize, implement and validate key digital PPG biomarkers. Approach. This work describes the creation of a standard Python toolbox, denoted pyPPG, for long-term continuous PPG time-series analysis and demonstrates the detection and computation of a high number of fiducial points and digital biomarkers using a standard fingerbased transmission pulse oximeter. Main results. The improved PPG peak detector had an F1-score of 88.19% for the state-of-the-art benchmark when evaluated on 2054 adult polysomnography recordings totaling over 91 million reference beats. The algorithm outperformed the open-source original Matlab implementation by ∼5% when benchmarked on a subset of 100 randomly selected MESA recordings. More than 3000 fiducial points were manually annotated by two annotators in order to validate the fiducial points detector. The detector consistently demonstrated high performance, with a mean absolute error of less than 10 ms for all fiducial points. Significance. Based on these fiducial points, pyPPG engineered a set of 74 PPG biomarkers. Studying PPG time-series variability using pyPPG can enhance our understanding of the manifestations and etiology of diseases. This toolbox can also be used for biomarker engineering in training data-driven models. pyPPG is available on https://physiozoo.com/.

Aims Ventricular tachycardia (VT) is a dangerous cardiac arrhythmia that can lead to sudden cardiac death. Early detection and management of VT is thus of high clinical importance. We hypothesize that it is possible to identify patients with VT during sinus rhythm by leveraging a continuous 24 h Holter electrocardiogram and artificial intelligence. Methods and results We analysed a retrospective Holter data set from the Rambam Health Care Campus, Haifa, Israel, which included 1773 Holter recordings from 1570 non-VT patients and 52 recordings from 49 VT patients. Morphological and heart rate variability features were engineered from the raw electrocardiogram signal and fed, together with demographical features, to a data-driven model for the task of classifying a patient as either VT or non-VT. The model obtained an area under the receiving operative curve of 0.76 ± 0.07. Feature importance suggested that the proportion of premature ventricular beats and beat-to-beat interval variability was discriminative of VT, while demographic features were not. Conclusion This original study demonstrates the feasibility of VT identification from sinus rhythm in Holter.

Atrial fibrillation (AF) is a prevalent clinical arrhythmia disease and is an important cause of stroke, heart failure, and sudden death. Due to the insidious onset and no obvious clinical symptoms of AF, the status of AF diagnosis and treatment is not optimal. Early AF screening or detection is essential. Internet of Things (IoT) and artificial intelligence (AI) technologies have driven the development of wearable electrocardiograph (ECG) devices used for health monitoring, which are an effective means of AF detection. The main challenges of AF analysis using ambulatory ECG include ECG signal quality assessment to select available ECG, the robust and accurate detection of QRS complex waves to monitor heart rate, and AF identification under the interference of abnormal ECG rhythm. Through ambulatory ECG measurement and intelligent detection technology, the probability of postoperative recurrence of AF can be reduced, and personalized treatment and management of patients with AF can be realized. This work describes the status of AF monitoring technology in terms of devices, algorithms, clinical applications, and future directions.

Introduction: Deep learning models for detecting episodes of atrial fibrillation (AF) using rhythm information in long-term ambulatory ECG recordings have shown high performance. However, the rhythm-based approach does not take advantage of the morphological information conveyed by the different ECG waveforms, particularly the f-waves. As a result, the performance of such models may be inherently limited. Methods: To address this limitation, we have developed a deep learning model, named RawECGNet, to detect episodes of AF and atrial flutter (AFl) using the raw, single-lead ECG. We compare the generalization performance of RawECGNet on two external data sets that account for distribution shifts in geography, ethnicity, and lead position. RawECGNet is further benchmarked against a state-of-the-art deep learning model, named ArNet2, which utilizes rhythm information as input. Results: Using RawECGNet, the results for the different leads in the external test sets in terms of the F1 score were 0.91-0.94 in RBDB and 0.93 in SHDB, compared to 0.89-0.91 in RBDB and 0.91 in SHDB for ArNet2. The results highlight RawECGNet as a high-performance, generalizable algorithm for detection of AF and AFl episodes, exploiting information on both rhythm and morphology.

Obstructive sleep apnea (OSA) is a serious medical condition with a high prevalence, although diagnosis remains a challenge. Existing home sleep tests may provide acceptable diagnosis performance but have shown several limitations. In this retrospective study, we used 12,923 polysomnography recordings from six independent databases to develop and evaluate a deep learning model, called OxiNet, for the estimation of the apnea-hypopnea index from the oximetry signal. We evaluated OxiNet performance across ethnicity, age, sex, and comorbidity. OxiNet missed 0.2% of all test set moderate-to-severe OSA patients against 21% for the best benchmark.

Use of non-stationary physiological signals for biometric verification, reduces the ability to forge. Such signals should be simple to acquire with inexpensive equipment. The beat-to-beat information embedded within the time intervals between consecutive heart beats is a non-stationary physiological signal; its potential for biometric verification has not been studied. This work introduces a biometric verification method termed “CompaRR”. Heartbeat was extracted from longitudinal recordings from 30 mice ranging from 6 to 24 months of age (equivalent to ~ 20–75 human years). Fifty heartbeats, which is close to resting human heartbeats in a minute, were sufficient for the verification task, achieving a minimal equal error rate of 0.21. When trained on 6-month-old mice and tested on unseen mice up to 18-months of age (equivalent to ~ 50 human years), no significant change in the verification performance was noted. Finally, when the model was trained on data from drug-treated mice, verification was still possible.

Background: Atrial fibrillation (AF) is one of the most common cardiac arrhythmias that affects millions of people each year worldwide and it is closely linked to increased risk of cardiovas- cular diseases such as stroke and heart failure. Machine learning methods have shown promising results in evaluating the risk of developing atrial fibrillation from the electrocardiogram. We aim to develop and evaluate one such algorithm on a large CODE dataset collected in Brazil. Methods: We used the CODE cohort to develop and test a model for AF risk prediction for individual patients from the raw ECG recordings without the use of additional digital biomarkers. The cohort is a collection of ECG recordings and annotations by the Telehealth Network of Minas Gerais, in Brazil. A convolutional neural network based on a residual network architecture was implemented to produce class probabilities for the classification of AF. The probabilities were used to develop a Cox proportional hazards model and a Kaplan-Meier model to carry out survival analysis. Hence, our model is able to perform risk prediction for the development of AF in patients without the condition. Results: The deep neural network model identified patients without indication of AF in the presented ECG but who will develop AF in the future with an AUC score of 0.845. From our survival model, we obtain that patients in the high-risk group (i.e. with the probability of a future AF case being >0.7) are 50% more likely to develop AF within 40 weeks, while patients belonging to the minimal-risk group (i.e. with the probability of a future AF case being less than or equal to 0.1) have >85% chance of remaining AF free up until after seven years. Conclusion: We developed and validated a model for AF risk prediction. If applied in clinical practice, the model possesses the potential of providing valuable and useful information in decision- making and patient management processes.

Photoplethysmography is a key sensing technology which is used in wearable devices such as smartwatches and fitness trackers. Currently, photoplethysmography sensors are used to monitor physiological parameters including heart rate and heart rhythm, and to track activities like sleep and exercise. Yet, wearable photoplethysmography has potential to provide much more information on health and wellbeing, which could inform clinical decision making. This Roadmap outlines directions for research and development to realise the full potential of wearable photoplethysmography. Experts discuss key topics within the areas of sensor design, signal processing, clinical applications, and research directions. Their perspectives provide valuable guidance to researchers developing wearable photoplethysmography technology.

Objective: Ophthalmological pathologies such as glaucoma, diabetic retinopathy and age-related macular degeneration are major causes of blindness and vision impairment. There is a need for novel decision support tools that can simplify and speed up the diagnosis of these pathologies. A key step in this process is to automatically estimate the quality of the fundus images to make sure these are interpretable by a human operator or a machine learning model. We present a novel fundus image quality scale and deep learning (DL) model that can estimate fundus image quality relative to this new scale. Methods: A total of 1245 images were graded for quality by two ophthalmologists within the range 1–10, with a resolution of 0.5. A DL regression model was trained for fundus image quality assessment. The architecture used was Inception-V3. The model was developed using a total of 89,947 images from 6 databases, of which 1245 were labeled by the specialists and the remaining 88,702 images were used for pre-training and semi-supervised learning. The final DL model was evaluated on an internal test set (n=209) as well as an external test set (n=194). Results: The final DL model, denoted FundusQ-Net, achieved a mean absolute error of 0.61 (0.54–0.68) on the internal test set. When evaluated as a binary classification model on the public DRIMDB database as an external test set the model obtained an accuracy of 99%. Significance: the proposed algorithm provides a new robust tool for automated quality grading of fundus images.

Study Objectives. To examine the feasibility of using digital oximetry biomarkers (OBMs) and body position to identify positional obstructive sleep apnea (POSA) phenotypes. Methods. A multiclass extreme gradient boost (XGBoost) was implemented to classify between three POSA phenotypes, i.e., positional patients (PP), including supine-predominant OSA (spOSA), and supine-isolated OSA (siOSA), and non-positional patients (NPP). A total of 861 individuals with OSA from the multi ethnic study of atherosclerosis (MESA) dataset were included in the study. Overall, 43 OBMs were computed for supine and non-supine positions and used as input features together with demographic and clinical information (META). Feature selection, using mRMR, was implemented, and nested cross validation was used for the model’s performance evaluation. Results. The best performance for the multiclass classification yielded a median weighted F1 of 0.79 with interquartile range (IQR) of 0.06. Binary classification between PP to NPP achieved weighted F1 of 0.87 (0.04). Conclusion. Using OBMs computed in PP and NPP with OSA, it is possible to distinguish between the different phenotypes of POSA. This data-driven algorithm may be embedded in portable home sleep tests.

Sinoatrial node (SAN) beating interval variability (BIV) and the average beating interval (BI) are regulated by a coupled-clock system, driven by Ca²⁺-calmodulin activated adenylyl cyclase, cAMP, and downstream PKA signaling. Reduced responsiveness of the BI and BIV to submaximal, [X]₅₀, β-adrenergic receptor (β-AR) stimulation, and phosphodiesterase inhibition (PDEI) have been documented in aged SAN tissue, whereas the maximal responses, [X]_max, do not differ by age. To determine whether age-associated dysfunction in cAMP signaling leads to altered responsiveness of BI and BIV, we measured cAMP levels and BI in adult (2–4 months n = 27) and aged (22–26 months n = 25) C57/BL6 mouse SAN tissue in control and in response to β-AR or PDEI at X₅₀ and [X]_max. Both cAMP and average BI in adult SAN were reduced at X₅₀, whereas cAMP and BI at X_max did not differ by age. cAMP levels and average BI were correlated both within and between adult and aged SAN. BIV parameters in long- and short-range terms were correlated with cAMP levels for adult SAN. However, due to reduced cAMP within aged tissues at [X]₅₀, these correlations were diminished in advanced age. Thus, cAMP level generated by the coupled clock mechanisms is tightly linked to average BI. Reduced cAMP level at X₅₀ in aged SAN explains the reduced responsiveness of the BI and BIV to β-AR stimulation and PDEI.

To drive health innovation that meets the needs of all and democratize healthcare, there is a need to assess the generalization performance of deep learning (DL) algorithms across various distribution shifts to ensure that these algorithms are robust. This retrospective study is, to the best of our knowledge, an original attempt to develop and assess the generalization performance of a DL model for AF events detection from long term beat-to-beat intervals across geography, ages and sexes. The new recurrent DL model, denoted ArNet2, is developed on a large retrospective dataset of 2,147 patients totaling 51,386 h obtained from continuous electrocardiogram (ECG). The model’s generalization is evaluated on manually annotated test sets from four centers (USA, Israel, Japan and China) totaling 402 patients. The model is further validated on a retrospective dataset of 1,825 consecutives Holter recordings from Israel. The model outperforms benchmark state-of-the-art models and generalized well across geography, ages and sexes. For the task of event detection ArNet2 performance was higher for female than male, higher for young adults (less than 61 years old) than other age groups and across geography. Finally, ArNet2 shows better performance for the test sets from the USA and China. The main finding explaining these variations is an impairment in performance in groups with a higher prevalence of atrial flutter (AFL). Our findings on the relative performance of ArNet2 across groups may have clinical implications on the choice of the preferred AF examination method to use relative to the group of interest.

Study Objectives: We investigated the characteristics of obstructive sleep apnea (OSA) positional patients’ (PP) phenotypes among different ethnic groups in the Multi-Ethnic Study of Atherosclerosis (MESA) dataset. Moreover, we hypothesized the existence of a new OSA PP phenotype we coined “Lateral PP,” for whom the lateral apnea-hypopnea index is at least double the supine apnea-hypopnea index. Methods: From 2,273 adults with sleep information, we analyzed data of 1,323 participants who slept more than 4 hours and had at least 30 minutes of sleep in both the supine and the nonsupine positions. Demographics and clinical information were compared for the different PP and ethnic groups. Results: 861 (65.1%) patients had OSA, and 35 (4.1%) were Lateral PP. Lateral PP patients were mainly females (62.9%), obese (median body mass index: 31.4 kg/m²), had mild–moderate OSA (94.3%), and mostly were non–Chinese American (97.1%). Among all patients with OSA, 550 (63.9%) were Supine PP and 17.7% were supine-isolated OSA. Supine PP and Lateral PP were present in 73.1% and 1.0% of Chinese Americans, 61.0% and 3.4% of Hispanics, 68.3% and 4.7% of White/Caucasian, and 56.2% and 5.2% of Black/African-American patients with OSA. Conclusions: Chinese Americans have the highest prevalence of Supine PP, whereas Black/African-American patients lean toward less Supine PP and higher Lateral PP. Lateral PP appears to be a novel OSA phenotype. However, Lateral PP was observed in a small group of patients with OSA and thus its existence should be further validated.

Liquid biopsy is seen as a prospective tool for cancer screening and tracking. However, the difficulty lies in effectively sieving, isolating, and overseeing cancer biomarkers from the backdrop of multiple disrupting cells and substances. The current study reports on the ability to perform liquid biopsy without the need to physically filter and/or isolate the cancer cells per se. This has been achieved through the detection and classification of volatile organic compounds (VOCs) emitted from the cancer cells found in the headspace of blood or urine samples or a combined data set of both. Spectrometric analysis shows that blood and urine contain complementary or overlapping VOC information on kidney cancer, gastric cancer, lung cancer, and fibrogastroscopy subjects. Based on this information, a nanomaterial-based chemical sensor array in conjugation with machine learning as well as data fusion of the signals achieved was carried out on various body fluids to assess the VOC profiles of cancer. The detection of VOC patterns by either Gas Chromatography−Mass Spectrometry (GC−MS) analysis or our sensor array achieved >90% accuracy, >80% sensitivity, and >80% specificity in different binary classification tasks. The hybrid approach, namely, analyzing the VOC datasets of blood and urine together, contributes an additional discrimination ability to the improvement (>3%) of the model’s accuracy. The contribution of the hybrid approach for an additional discrimination ability to the improvement of the model’s accuracy is examined and reported.

Sleep staging is an essential component in the diagnosis of sleep disorders and management of sleep health. Sleep is traditionally measured in a clinical setting and requires a labor-intensive labeling process. We hypothesize that it is possible to perform automated robust 4-class sleep staging using the raw photoplethysmography (PPG) time series and modern advances in deep learning (DL). We used two publicly available sleep databases that included raw PPG recordings, totalling 2,374 patients and 23,055 hours of continuous data. We developed SleepPPG-Net, a DL model for 4-class sleep staging from the raw PPG time series. SleepPPG-Net was trained end-to-end and consists of a residual convolutional network for automatic feature extraction and a temporal convolutional network to capture long-range contextual information. We benchmarked the performance of SleepPPG-Net against models based on the best-reported state-of-the-art (SOTA) algorithms. When benchmarked on a held-out test set, SleepPPG-Net obtained a median Cohen's Kappa (κ) score of 0.75 against 0.69 for the best SOTA approach. SleepPPG-Net showed good generalization performance to an external database, obtaining a κ score of 0.74 after transfer learning. Overall, SleepPPG-Net provides new SOTA performance. In addition, performance is high enough to open the path to the development of wearables that meet the requirements for usage in clinical applications such as the diagnosis and monitoring of obstructive sleep apnea.

Introduction: Blood vessels can be non-invasively visualized from a digital fundus image (DFI). Several studies have shown an association between cardiovascular risk and vascular features obtained from DFI. Recent advances in computer vision and image segmentation enable automatising DFI blood vessel segmentation. There is a need for a resource that can automatically compute digital vasculature biomarkers (VBM) from these segmented DFI. Methods: In this paper, we introduce a Python Vasculature BioMarker toolbox, denoted PVBM. A total of 11 VBMs were implemented. In particular, we introduce new algorithmic methods to estimate tortuosity and branching angles. Using PVBM, and as a proof of usability, we analyze geometric vascular differences between glaucomatous patients and healthy controls. Results: We built a fully automated vasculature biomarker toolbox based on DFI segmentations and provided a proof of usability to characterize the vascular changes in glaucoma. For arterioles and venules, all biomarkers were significant and lower in glaucoma patients compared to healthy controls except for tortuosity, venular singularity length and venular branching angles. Conclusion: We have automated the computation of 11 VBMs from retinal blood vessel segmentation. The PVBM toolbox is made open source under a GNU GPL 3 license and is available on physiozoo.com (following publication).

Objective: Over the past few years, deep learning (DL) has been used extensively in research for 12-lead electrocardiogram (ECG) analysis. However, it is unclear whether the explicit or implicit claims made on DL superiority to the more classical feature engineering (FE) approaches, based on domain knowledge, hold. In addition, it remains unclear whether combining DL with FE may improve performance over a single modality. <italic>Methods:</italic> To address these research gaps and in-line with recent major experiments, we revisited three tasks: cardiac arrhythmia diagnosis (multiclass-multilabel classification), atrial fibrillation risk prediction (binary classification), and age estimation (regression). We used an overall dataset of 2.3M 12-lead ECG recordings to train the following models for each task: i) a random forest taking FE as input; ii) an end-to-end DL model; and iii) a merged model of FE+DL. <italic>Results:</italic> FE yielded comparable results to DL while necessitating significantly less data for the two classification tasks. DL outperformed FE for the regression task. For all tasks, merging FE with DL did not improve performance over DL alone. These findings were confirmed on the additional PTB-XL dataset. <italic>Conclusion:</italic> We found that for traditional 12-lead ECG based diagnosis tasks, DL did not yield a meaningful improvement over FE, while it improved significantly the nontraditional regression task. We also found that combining FE with DL did not improve over DL alone, which suggests that the FE were redundant with the features learned by DL. <italic>Significance:</italic> Our findings provides important recommendations on 12-lead ECG based machine learning strategy and data regime to choose for a given task. When looking at maximizing performance as the end goal, if the task is nontraditional and a large dataset is available then DL is preferable. If the task is a classical one and/or a small dataset is available then a FE approach may be the better choice.

Non-invasive oxygen saturation (SpO2) is a central vital sign used to shape the management of COVID-19 patients. Yet, there have been no report quantitatively describing SpO2 dynamics and patterns in COVID-19 patients using continuous SpO2 recordings. We performed a retrospective observational analysis of the clinical information and 27 K hours of continuous SpO2 high-resolution (1 Hz) recordings of 367 critical and non-critical COVID-19 patients hospitalised at the Rambam Health Care Campus, Haifa, Israel. An absolute SpO2 threshold of 93% most efficiently discriminated between critical and non-critical patients, regardless of oxygen support. Oximetry-derived digital biomarker (OBMs) computed per 1 h monitoring window showed significant differences between groups, notably the cumulative time below 93% SpO2 (CT93). Patients with CT93 above 60% during the first hour of monitoring, were more likely to require oxygen support. Mechanical ventilation exhibited a strong effect on SpO2 dynamics by significantly reducing the frequency and depth of desaturations. OBMs related to periodicity and hypoxic burden were markedly affected, up to several hours before the initiation of the mechanical ventilation. In summary, OBMs, traditionally used in the field of sleep medicine research, are informative for continuous assessment of disease severity and response to respiratory support of hospitalised COVID-19 patients. In conclusion, OBMs may improve risk stratification and therapy management of critical care patients with respiratory impairment.

Wearable photoplethysmography devices such as smart-watches can detect possible arrhythmias from inter-beat intervals (IBIs). However, photoplethysmogram (PPG) signals are highly susceptible to motion artifact. This study investigated using simultaneous accelerometry signals to determine whether IBIs can be reliably measured from PPG signals. The PPG-DaLiA and WESAD datasets were used. These datasets contain wrist accelerometry and PPG signals collected from 15 subjects during activities of daily living and mental stress tasks. IBIs were estimated from PPG signals using the 'MSPTD' beat detection algorithm. PPG-based IBIs were deemed accurate if the resulting instantaneous heart rate (IHR) was within ± 5 bpm of a reference ECG-derived IHR. The mean absolute deviation (MAD) of the accelerometry signal was able to predict whether PPG-derived IBIs were accurate, with an area under the precision-recall curve (AUPRC) of 0.82 on all data. An optimal MAD threshold of 12.9 milli-gravitational units was identified. However, performance was poorer during stress (AUPRC of 0.54). In conclusion, accelerometry can be used to identify periods when IBIs can be accurately measured from PPG signals during activities associated with movement, but is not reliable during stress.

Premature births can lead to complications, with fetal lung immaturity being a primary concern. Currently, fetal lung maturity (FLM) requires an invasive surfactant extraction procedure between the 32nd and 39th weeks of pregnancy. Unfortunately, there is no non-invasive method for FLM assessment. This work hypothesized that fetal breathing movement (FBM) and surfactant levels are inversely coupled and that FBM can serve as a proxy for FLM estimation. To investigate the correlation between FBM and FLM, antenatal corticosteroid (ACS) was administered to increase fetal pulmonary surfactant levels in a high-risk 35th-week pregnant woman showing intrauterine growth restriction. Synchronous sonographic and phonographic measurements were continuously recorded for 25 minutes before and after the ASC treatments. Before the ACS injection, 268 continuous movements FBM episodes were recorded. The number of continuous FBM episodes significantly decreased to 3, 43, and 79 within 24, 48, and 72 hours, respectively, of the first injection of ACS, suggesting an inversely coupled connection between FBM and surfactant level s. Therefore, FBM may serve as a proxy for FLM estimation. Quantitative confirmation of these findings would suggest that FBM measurements could be used as a non-invasive and widely accessible FLM-assessment tool for high-risk pregnancies and routine examinations.

PhysioZoo is a collaborative platform designed for the analysis of continuous physiological time series. The platform currently comprises four modules, each consisting of a library, a user interface, and a set of tutorials: (1) PhysioZoo HRV, dedicated to studying heart rate variability (HRV) in humans and other mammals; (2) PhysioZoo SPO2, which focuses on the analysis of digital oximetry biomarkers (OBM) using continuous oximetry (SPO2) measurements from humans; (3) PhysioZoo ECG, dedicated to the analysis of electrocardiogram (ECG) time series; (4) PhysioZoo PPG, designed to study photoplethys-mography (PPG) time series. In this proceeding, we introduce the PhysioZoo platform as an open resource for digital physiological biomarkers engineering, facilitating streamlined analysis and data visualization of physiological time series while ensuring the reproducibility of published experiments. We welcome researchers to contribute new libraries for the analysis of various physiological time series, such as electroencephalography, blood pressure, and phonocardiography. You can access the resource at physiozoo.com. We encourage researchers to explore and utilize this platform to advance their studies in the field of continuous physiological time-series analysis.

Efficient and accurate evaluation of long-term photo-plethysmography (PPG) recordings is essential for both clinical assessments and consumer products. In 2021, the top opensource peak detectors were benchmarked on the Multi-Ethnic Study of Atherosclerosis (MESA) database consisting of polysomnography (PSG) recordings and continuous sleep PPG data, where the Automatic Beat Detector (Aboy) had the best accuracy. This work presents Aboy++, an improved version of the original Aboy beat detector. The algorithm was evaluated on 100 adult PPG recordings from the MESA database, which contains more than 4.25 million reference beats. Aboy++ achieved an F1-score of 85.5%, compared to 80.99% for the original Aboy peak detector. On average, Aboy++ processed a 1 hour-long recording in less than 2 seconds. This is compared to 115 seconds (i.e., over 57-times longer) for the open-source implementation of the original Aboy peak detector. This study demonstrated the importance of developing robust algorithms like Aboy++ to improve PPG data analysis and clinical outcomes. Overall, Aboy++ is a reliable tool for evaluating long-term wearable PPG measurements in clinical and consumer contexts. The open-source algorithm is available on the physiozoo.com website.

Objective. Epileptic seizures are relatively common in critically-ill children admitted to the pediatric intensive care unit (PICU) and thus serve as an important target for identification and treatment. Most of these seizures have no discernible clinical manifestation but still have a significant impact on morbidity and mortality. Children that are deemed at risk for seizures within the PICU are monitored using continuous-electroencephalogram (cEEG). cEEG monitoring cost is considerable and as the number of available machines is always limited, clinicians need to resort to triaging patients according to perceived risk in order to allocate resources. This research aims to develop a computer aided tool to improve seizures risk assessment in critically-ill children, using an ubiquitously recorded signal in the PICU, namely the electrocardiogram (ECG). Approach. A novel data-driven model was developed at a patient-level approach, based on features extracted from the first hour of ECG recording and the clinical data of the patient. Main results. The most predictive features were the age of the patient, the brain injury as coma etiology and the QRS area. For patients without any prior clinical data, using one hour of ECG recording, the classification performance of the random forest classifier reached an area under the receiver operating characteristic curve (AUROC) score of 0.84. When combining ECG features with the patients clinical history, the AUROC reached 0.87. Significance. Taking a real clinical scenario, we estimated that our clinical decision support triage tool can improve the positive predictive value by more than 59% over the clinical standard.

The photoplethysmogram (PPG) signal is widely used in pulse oximeters and smartwatches. A fundamental step in analysing the PPG is the detection of heartbeats. Several PPG beat detection algorithms have been proposed, although it is not clear which performs best. Objective: This study aimed to: (i) develop a framework with which to design and test PPG beat detectors; (ii) assess the performance of PPG beat detectors in different use cases; and (iii) investigate how their performance is affected by patient demographics and physiology. Approach: Fifteen beat detectors were assessed against electrocardiogram-derived heartbeats using data from eight datasets. Performance was assessed using the F 1 score, which combines sensitivity and positive predictive value. Main results: Eight beat detectors performed well in the absence of movement with F 1 scores of ≥90% on hospital data and wearable data collected at rest. Their performance was poorer during exercise with F 1 scores of 55%-91%; poorer in neonates than adults with F 1 scores of 84%-96% in neonates compared to 98%-99% in adults; and poorer in atrial fibrillation (AF) with F 1 scores of 92%-97% in AF compared to 99%-100% in normal sinus rhythm. Significance: Two PPG beat detectors denoted 'MSPTD' and 'qppg' performed best, with complementary performance characteristics. This evidence can be used to inform the choice of PPG beat detector algorithm. The algorithms, datasets, and assessment framework are freely available.

Objective. A classifier based on weighted voting of multiple single-lead based models combining deep learning (DL) representation and hand-crafted features was developed to classify 26 cardiac abnormalities from different lead subsets of short-term electrocardiograms (ECG). Approach. A two-stage method was proposed for the multilead prediction. First a lead-agnostic hybrid classifier was trained to predict the pathologies from single-lead ECG signals. The classifier combined fully automated DL features extracted through a convolutional neural network with hand-crafted features through a fully connected layer. Second, a voting of the single-lead based predictions was performed. For the 12-lead subset, voting consisted in an optimised weighting of the output probabilities of all available single lead predictions. For other lead subsets, voting simply consisted in the average of the lead predictions. Main results. This approach achieved a challenge test score of 0.48, 0.47, 0.46, 0.46, 0.45 on the 12, 6, 4, 3, 2-lead subsets respectively on the 2021 Physionet/Computing in Cardiology challenge hidden test set. The use of an hybrid approach and more advanced voting layer improved some individual class classification but did not offer better generalization than our baseline fully DL approach. Significance. The proposed approach showed potential at correctly classifying main cardiac abnormalities and dealt well with reduced lead subsets.

Objective. Arrhythmia is an abnormal cardiac rhythm that affects the pattern and rate of the heartbeat. Wearable devices with the functionality to measure and store heart rate (HR) data are growing in popularity and enable diagnosing and monitoring arrhythmia on a large scale. The typical sampling resolution of HR data available from non-medical grade wearable devices varies from seconds to several minutes depending on the device and its settings. However, the impact of sampling resolution on the performance and quality of arrhythmia detection has not yet been quantified. Approach. In this study, we investigated the detection and classification of three arrhythmias, namely atrial fibrillation, bradycardia, tachycardia, from down-sampled HR data with various temporal resolution (5-, 15-, 30- and 60 s averages) in 1 h segments extracted from an annotated Holter ECG database acquired at the University of Virginia Heart Station. For the classification task, a total of 15 common heart rate variability (HRV) features were engineered based on the HR time series of each patient. Three different types of machine learning classifiers were evaluated, namely logistic regression, support vector machine and random forest. Main results. A decrease in temporal resolution drastically impacted the detection of atrial fibrillation but did not substantially affect the detection of bradycardia and tachycardia. A HR resolution up to 15 s average demonstrated reasonable performance with a sensitivity of 0.92 and a specificity of 0.86 for a multiclass random forest classifier. Significance. HRV features extracted from low resolution long HR recordings have the potential to increase the early detection of arrhythmias in undiagnosed individuals.

Objective. Fetal arrhythmias are a life-threatening disorder occurring in up to 2% of pregnancies. If identified, many fetal arrhythmias can be effectively treated using anti-arrhythmic therapies. In this paper, we present a novel method of detecting fetal arrhythmias in short length non-invasive fetal electrocardiography (NI-FECG) recordings. Approach. Our method consists of extracting a fetal heart rate time series from each NI-FECG recording and computing an entropy profile using a data-driven range of the entropy tolerance parameter r. To validate our approach, we apply our entropy profiling method to a large clinical data set of 318 NI-FECG recordings. Main Results. We demonstrate that our method (TotalSampEn) provides strong performance for classifying arrhythmic fetuses (AUC of 0.83) and outperforms entropy measures such as SampEn (AUC of 0.68) and FuzzyEn (AUC of 0.72). We also find that NI-FECG recordings incorrectly classified using the investigated entropy measures have significantly lower signal quality, and that excluding recordings of low signal quality (13.5% of recordings) increases the classification performance of TotalSampEn (AUC of 0.90). Significance. The superior performance of our approach enables automated detection of fetal arrhythmias and warrants further investigation in a prospective clinical trial.

Introduction: Atrial fibrillation (AF) is the most prevalent heart arrhythmia. AF manifests on the electrocar-diogram (ECG) though irregular beat-to-beat time interval variation, the absence of P-wave and the presence of fibrillatory waves (f-wave). We hypothesize that a deep learning (DL) approach trained on the raw ECG will enable robust detection of AF events and the estimation of the AF burden (AFB). We further hypothesize that the performance reached leveraging the raw ECG will be superior to previously developed methods using the beat-to-beat interval variation time series. Consequently, we develop a new DL algorithm, denoted ArNet-ECG, to robustly detect AF events and estimate the AFB from the raw ECG and bench-mark this algorithms against previous work. Methods: A dataset including 2,247 adult patients and totaling over 53,753 hours of continuous ECG from the University of Virginia (UVAF) was used. Results: ArNet-ECG obtained an F1 of 0.96 and ArNet2 obtained an F1 0.94. Discussion and conclusion: ArNet-ECG outperformed ArNet2 thus demonstrating that using the raw ECG provides added performance over the beat-to-beat interval time series. The main reason found for explaining the higher performance of ArNet-ECG was its high performance on atrial flutter examples versus poor performance on these recordings for ArNet2.

Introduction: 12-lead electrocardiogram (ECG) is recorded during atrial fibrillation (AF) catheter ablation procedure (CAP). It is not easy to determine if CAP was successful without a long follow-up assessing for AF re-currence (AFR). Therefore, an AFR risk prediction algorithm could enable a better management of CAP patients. In this research, we extracted features from 12-lead ECG recorded before and after CAP and train an AFR risk pre-diction machine learning model. Methods: Pre- and post-CAP segments were extracted from 112 patients. The anal-ysis included a signal quality criterion, heart rate variabil-ity and morphological biomarkers engineered from the 12-lead ECG (804 features overall). 43 out of the 112 patients (n) had AFR clinical endpoint available. These were utilized to assess the feasibility of AFR risk prediction, using either pre or post CAP features. A random forest clas-sifier was trained within a nested cross validation frame-work. Results: 36 features were found statistically signif-icant for distinguishing between the pre and post surgery states (n=112). For the classification, an area under the receiver operating characteristic (AUROC) curve was re-ported with AU ROCpre= 0.64 and AU ROCpost= 0.74 (n=43). Discussion and conclusions: This preliminary analysis showed the feasibility of AFR risk prediction. Such a model could be used to improve CAP management.

Introduction: Atrial fibrillation (AF) is the most common heart arrhythmia, characterized by the presence of fibrillatory waves (f-waves) in the ECG. We introduce a voting scheme to estimate the dominant atrial frequency (DAF) of f-waves. Methods: We analysed a subset of Holter recordings obtained from the University of Virginia AF Database. 100 Holter recordings with manually annotated AF events, resulting in a total 363 AF events lasting more than 1 min. The f-waves were extracted using four different template subtraction (TS) algorithms and the DAF was estimated from the first 1-min window of each AF event. A random forest classifier was used. We hypothesized that better extraction of the f-wave meant better AF/non-AF classification using the DAF as the single input feature of the RF model. Results: Performance on the test set, expressed in terms of AF/non-AF classification, was the best when the DAF was computed computed the three best-performing extraction methods. Using these three algorithms in a voting scheme, the classifier obtained AUC=0.60 and the DAFs were mostly spread around 6 Hz, 5.66 (4.83-7.47). Conclusions: This study has two novel contributions: (1) a method for assessing the performance of f-wave extraction algorithms, and (2) a voting scheme for improved DAF estimation.

Introduction: For supervised deep learning (DL) tasks, researchers need a large annotated dataset. In medical data science, one of the major limitations to develop DL models is the lack of annotated examples in large quantity. This is most often due to the time and expertise required to annotate. We introduce Lirot.ai, a novel platform for facilitating and crowd-sourcing image segmentations. Methods: Lirot.ai is composed of three components; an iPadOS client application named Lirot.ai-app, a back-end server named Lirot.ai-server and a python API name Lirot.ai-API. Lirot.ai-app was developed in Swift 5.6 and Lirot.ai-server is a firebase backend. Lirot.ai-API allows the management of the database. Lirot.ai-app can be installed on as many iPadOS devices as needed so that annotators may be able to perform their segmentation simultaneously and remotely. We incorporate Apple Pencil compatibility, making the segmentation faster, more accurate, and more intuitive for the expert than any other computer-based alternative. Results: We demonstrate the usage of Lirot.ai for the creation of a retinal fundus dataset with reference vasculature segmentations. Discussion and future work: We will use active learning strategies to continue enlarging our retinal fundus dataset by including a more efficient process to select the images to be annotated and distribute them to annotators.

Aims: This study aims to assess whether information derived from the raw 12-lead electrocardiogram (ECG) combined with clinical information is predictive of atrial fibrillation (AF) development. Methods and results: We use a subset of the Telehealth Network of Minas Gerais (TNMG) database consisting of patients that had repeated 12-lead ECG measurements between 2010 and 2017 that is 1 130 404 recordings from 415 389 unique patients. Median and interquartile of age for the recordings were 58 (46-69) and 38% of the patients were males. Recordings were assigned to train-validation and test sets in an 80:20% split which was stratified by class, age and gender. A random forest classifier was trained to predict, for a given recording, the risk of AF development within 5 years. We use features obtained from different modalities, namely demographics, clinical information, engineered features, and features from deep representation learning. The best model performance on the test set was obtained for the model combining features from all modalities with an area under the receiver operating characteristic curve (AUROC) = 0.909 against the best single modality model which had an AUROC = 0.839. Conclusion: Our study has important clinical implications for AF management. It is the first study integrating feature engineering, deep learning, and Electronic medical record system (EMR) metadata to create a risk prediction tool for the management of patients at risk of AF. The best model that includes features from all modalities demonstrates that human knowledge in electrophysiology combined with deep learning outperforms any single modality approach. The high performance obtained suggest that structural changes in the 12-lead ECG are associated with existing or impending AF.

A system comprising at least one hardware processor; and a non-transitory computer-readable storage medium having stored thereon program instructions, the program instructions executable by the at least one hardware processor to: receive, as input, a signal representing temporal beat activity in a mammalian, receive indication of user selection of an integration level of said beat rate activity, receive indication of user selection of a species of said mammalian, and determine heart rate variation (HRV) in said signal, based, at least in part, on said type of said beat rate activity and said species of said mammalian.

Objective: The atrial fibrillation burden (AFB) is defined as the percentage of time spent in atrial fibrillation (AF) over a long enough monitoring period. Recent research has suggested the added prognostic value of using the AFB compared to a binary diagnosis. We evaluate, for the first time, the ability to estimate the AFB over long-term continuous recordings, using a deep recurrent neutral network (DRNN) approach. Methods: The models were developed and evaluated on a large database of p = 2,891 patients, totaling t = 68,800 hours of continuous electrocardiography (ECG) recordings from the University of Virginia. Specifically, 24h beat-to-beat time series were obtained from a single portable ECG channel. The network, denoted ArNet, was benchmarked against a gradient boosting (XGB) model, trained on 21 features including the coefficient of sample entropy (CosEn) and AFEvidence that is derived from the number of irregular points revealed by the Lorenz plot. The generalizations of ArNet and XGB were also evaluated on the independent PhysioNet LTAF test database. Results: the absolute AF burden estimation error | EAF(%)} |, median and interquartile, on the test set, was 1.2 (0.1-6.7) for ArNet and 2.8 (0.9-11.7) for XGB for AF individuals. Generalization results on LTAF were consistent with | EAF(% | of 2.7 (1.1-14.7) for ArNet and 3.6 (1.0-16.7) for XGB. Conclusion: This research demonstrates the feasibility of AFB estimation from 24h beat-to-beat interval time series utilizing DRNNs. Significance: The novel data-driven approach enables robust remote diagnosis and phenotyping of AF.

Background: COVID-19 is a newly recognized illness with a predominantly respiratory presentation. It is important to characterize the differences in disease presentation and trajectory between COVID-19 patients and other patients with common respiratory illnesses. These differences can enhance knowledge of pathogenesis and help in guiding treatment. Methods: Data from electronic medical records were obtained from individuals admitted with respiratory illnesses to Rambam Health Care Campus, Haifa, Israel, between October 1st, 2014 and October 1st, 2020. Four groups of patients were defined: COVID-19 (693), influenza (1,612), severe acute respiratory infection (SARI) (2,292), and Others (4,054). The variable analyzed include demographics (7), vital signs (8), lab tests (38), and comorbidities (15) from a total of 8,651 hospitalized adult patients. Statistical analysis was performed on biomarkers measured at admission and for their disease trajectory in the first 48 h of hospitalization, and on comorobidity prevalence. Results: COVID-19 patients were overall younger in age and had higher body mass index, compared to influenza and SARI. Comorbidity burden was lower in the COVID-19 group compared to influenza and SARI. Severely- and moderately-ill COVID-19 patients older than 65 years of age suffered higher rate of in-hospital mortality compared to hospitalized influenza patients. At admission, white blood cells and neutrophils were lower among COVID-19 patients compared to influenza and SARI patients, while pulse rate and lymphoctye percentage were higher. Trajectories of variables during the first 2 days of hospitalization revealed that white blood count, neutrophils percentage and glucose in blood increased among COVID-19 patients, while decreasing among other patients. Conclusions: The intrinsic virulence of COVID-19 appeared higher than influenza. In addition, several critical functions, such as immune response, coagulation, heart and respiratory function, and metabolism were uniquely affected by COVID-19.

Objective. Chronic obstructive pulmonary disease (COPD) is a highly prevalent chronic condition. COPD is a major cause of morbidity, mortality and healthcare costs globally. Spirometry is the gold standard test for a definitive diagnosis and severity grading of COPD. However, a large proportion of individuals with COPD are undiagnosed and untreated. Given the high prevalence of COPD and its clinical importance, it is critical to develop new algorithms to identify undiagnosed COPD. This is particularly true in specific disease groups in which the presence of concomitant COPD increases overall morbidity/mortality such as those with sleep-disordered breathing. To our knowledge, no research has looked at the feasibility of automated COPD diagnosis using a data-driven analysis of the nocturnal continuous oximetry time series. We hypothesize that patients with COPD will exert certain patterns and/or dynamics of their overnight oximetry time series that are unique to this condition and that may be captured using a data-driven approach. Approach. We introduce a novel approach to nocturnal COPD diagnosis using 44 oximetry digital biomarkers and five demographic features and assess its performance in a population sample at risk of sleep-disordered breathing. A total of n=350 unique patients' polysomnography (PSG) recordings were used. A random forest (RF) classifier was trained using these features and evaluated using nested cross-validation. Main results. The RF classifier obtained F1 = 0.86 ± 0.02 and AUROC = 0.93 ± 0.02 on the test sets. A total of 8 COPD individuals out of 70 were misclassified. No severe cases (GOLD 3-4) were misdiagnosed. Including additional non-oximetry derived PSG biomarkers resulted in minimal performance increase. Significance. We demonstrated for the first time, the feasibility of COPD diagnosis from nocturnal oximetry time series for a population sample at risk of sleep-disordered breathing. We also highlighted what set of digital oximetry biomarkers best reflect how COPD manifests overnight. The results motivate that overnight single channel oximetry can be a valuable modality for COPD diagnosis, in a population sample at risk of sleep-disordered breathing. Further data is needed to validate this approach on other population samples.

Objective. In this perspective paper, we aim to highlight the potential of sleep as an auspicious time for diagnosis, management and therapy of non-sleep-specific pathologies. Approach. Sleep has a profound influence on the physiology of body systems and biological processes. Molecular studies have shown circadian-regulated shifts in protein expression patterns across human tissues, further emphasizing the unique functional, behavioral and pharmacokinetic landscape of sleep. Thus, many pathological processes are also expected to exhibit sleep-specific manifestations. Modern advances in biosensor technologies have enabled remote, non-invasive recording of a growing number of physiologic parameters and biomarkers promoting the detection and study of such processes. Main results. Here, we introduce key clinical studies in selected medical fields, which leveraged novel technologies and the advantageous period of sleep to diagnose, monitor and treat pathologies. Studies demonstrate that sleep is an ideal time frame for the collection of long and clean physiological time series data which can then be analyzed using data-driven algorithms such as deep learning. Significance.This new paradigm proposes opportunities to further harness modern technologies to explore human health and disease during sleep and to advance the development of novel clinical applications – from sleep medicine to medicine during sleep.

Background: Because of the complexity of the interaction between the internal pacemaker mechanisms, cell interconnected signals, and interaction with other body systems, study of the role of individual systems must be performed under in vivo and in situ conditions. The in situ approach is valuable when exploring the mechanisms that govern the beating rate and rhythm of the sinoatrial node (SAN), the heart’s primary pacemaker. SAN beating rate changes on a beat-to-beat basis. However, to date, there are no standard methods and tools for beating rate variability (BRV) analysis from electrograms (EGMs) collected from different mammals, and there is no centralized public database with such recordings. Methods: We used EGM recordings obtained from control SAN tissues of rabbits (n = 9) and mice (n = 30) and from mouse SAN tissues (n = 6) that were exposed to drug intervention. The data were harnessed to develop a beat detector to derive the beat-to-beat interval time series from EGM recordings. We adapted BRV measures from heart rate variability and reported their range for rabbit and mouse. Results: The beat detector algorithm performed with 99% accuracy, sensitivity, and positive predictive value on the test (mouse) and validation (rabbit and mouse) sets. Differences in the frequency band cutoff were found between BRV of SAN tissue vs. heart rate variability (HRV) of in vivo recordings. A significant reduction in power spectrum density existed in the high frequency band, and a relative increase was seen in the low and very low frequency bands. In isolated SAN, the larger animal had a slower beating rate but with lower BRV, which contrasted the phenomena reported for in vivo analysis. Thus, the non-linear inverse relationship between the average HR and HRV is not maintained under in situ conditions. The beat detector, BRV measures, and databases were contributed to the open-source PhysioZoo software (available at: https://physiozoo.com/). Conclusion: Our approach will enable standardization and reproducibility of BRV analysis in mammals. Different trends were found between beating rate and BRV or HRV in isolated SAN tissue vs. recordings collected under in vivo conditions, respectively, implying a complex interaction between the SAN and the autonomic nervous system in determining HRV in vivo.

Pulse oximetry is routinely used to non-invasively monitor oxygen saturation levels. A low oxygen level in the blood means low oxygen in the tissues, which can ultimately lead to organ failure. Yet, contrary to heart rate variability measures, a field which has seen the development of stable standards and advanced toolboxes and software, no such standards and open tools exist for continuous oxygen saturation time series variability analysis. The primary objective of this research was to identify, implement and validate key digital oximetry biomarkers (OBMs) for the purpose of creating a standard and associated reference toolbox for continuous oximetry time series analysis. We review the sleep medicine literature to identify clinically relevant OBMs. We implement these biomarkers and demonstrate their clinical value within the context of obstructive sleep apnea (OSA) diagnosis on a total of n = 3806 individual polysomnography recordings totaling 26,686 h of continuous data. A total of 44 digital oximetry biomarkers were implemented. Reference ranges for each biomarker are provided for individuals with mild, moderate, and severe OSA and for non-OSA recordings. Linear regression analysis between biomarkers and the apnea hypopnea index (AHI) showed a high correlation, which reached R¯ ²= 0.82. The resulting python OBM toolbox, denoted “pobm”, was contributed to the open software PhysioZoo (physiozoo.org). Studying the variability of the continuous oxygen saturation time series using pbom may provide information on the underlying physiological control systems and enhance our understanding of the manifestations and etiology of diseases, with emphasis on respiratory diseases.

Introduction: Photoplethysmography (PPG) is fast becoming the signal of choice for the widespread monitoring of sleep metrics obtained by wearable devices. Robust peak detection is critical for the extraction of mean-ingful features from the PPG waveform. There is however no consensus on what PPG peak detection algorithms perform best on nocturnal continuous PPG recordings. We introduce two methods to benchmark the performance of PPG peak detectors. Methods: We make use of data where nocturnal PPG and electrocardiogram (ECG) are measured synchronously. Within this setting, the ECG, a signal for which there are established R -peak detectors, is used as reference. The first method for benchmarking, denoted 'Peak Matching', consists of forecasting the expected position of the PPG peaks using the ECG R-peaks as reference. The second technique, denoted 'IHR-IPR Accuracy', compares the instantaneous pulse rate (IPR) extracted from the PPG with the instantaneous heart rate (IHR) extracted from the ECG. For benchmarking, we used the MESA dataset consisting of 2,055 overnight polysomnography recordings with a combined length of over 16,300 hours. Four open PPG peak detectors were benchmarked. Results: The 'Pulses' detector performed best with a Peak Matching F1-score of 0.94 and an IHR-IPR Accuracy of 89.6%. Discussion and conclusion: We introduced two new methods for benchmarking PPG peak detectors. Among the four detectors evaluated, 'Pulses' performed best. Benchmarking of further PPG detectors and on other data source (e.g. daytime recordings, recordings from patients with arrhythmia) is needed.

Introduction: The electrocardiogram (ECG) is a standard tool used in medical practice for identifying cardiac pathologies. Because the necessary expertise to interpret this tracing is not readily available in all medical institutions or at all in some large areas of developing countries, there is a need to create a data-driven approach that can automatically capture the information contained in this physiological time series. Yet, contrary to heart rate variability measures, a field which has seen the development of standards, advanced toolboxes and software, very little open tools exist for ECG morphological analysis. The primary objective of this work was to identify and implement clinically important digital ECG biomarkers for the purpose of creating a reference toolbox and software for ECG morphological analysis. Methods: The epltd algorithm was used for R-peak detection. We used a zero-phase filter with passband 0.67Hz - 100Hz to remove baseline wander and high frequency noise. We used a Notch filter at 50/60Hz to remove the power-line interference. ECG fiducial points were detected using the well-known open source wavedet algorithm. A total of 22 biomarkers were engineered including 14 extracted from intervals and segments duration and 8 from waves characteristics. Results and discussion: the result of this work consists of a Python toolbox termed 'pebm' and its user interface termed 'Physio'Zoo ECG' for data visualization and analysis. The software is available at physiozoo.com under a GNU GPL licence. The pebm toolbox may be used to provide new physiological information on cardiac conduction as well as used as a source of readily handcrafted features for training machine learning models.

Coronavirus disease (COVID-19), caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is rapidly spreading across the globe. The clinical spectrum of SARS-CoV-2 pneumonia requires early detection and monitoring, within a clinical environment for critical cases and remotely for mild cases, with a large spectrum of symptoms. The fear of contamination in clinical environments has led to a dramatic reduction in on-site referrals for routine care. There has also been a perceived need to continuously monitor non-severe COVID-19 patients, either from their quarantine site at home, or dedicated quarantine locations (e.g. hotels). In particular, facilitating contact tracing with proximity and location tracing apps was adopted in many countries very rapidly. Thus, the pandemic has driven incentives to innovate and enhance or create new routes for providing healthcare services at distance. In particular, this has created a dramatic impetus to find innovative ways to remotely and effectively monitor patient health status. In this paper, we present a review of remote health monitoring initiatives taken in 20 states during the time of the pandemic. We emphasize in the discussion particular aspects that are common ground for the reviewed states, in particular the future impact of the pandemic on remote health monitoring and consideration on data privacy.

Objective: In this research, we introduce a new methodology for atrial fibrillation (AF) diagnosis during sleep in a large population sample at risk of sleep-disordered breathing. Approach: The approach leverages digital biomarkers and recent advances in machine learning (ML) for mass AF diagnosis from overnight-hours of single-channel electrocardiogram (ECG) recording. Four databases, totaling n = 3088 patients and p = 26 913 h of continuous single-channel electrocardiogram raw data were used. Three of the databases (n = 125, p = 2513) were used for training a ML model in recognizing AF events from beat-to-beat time series. Visit 1 of the sleep heart health study database (SHHS1, n = 2963, p = 24 400) was used as the test set to evaluate the feasibility of identifying prominent AF from polysomnographic recordings. By combining AF diagnosis history and a cardiologist’s visual inspection of individuals suspected of having AF (n = 118), a total of 70 patients were diagnosed with prominent AF in SHHS1. Main results: Model prediction on SHHS1 showed an overall Se = 0.97, Sp = 0.99, NPV = 0.99 and PPV = 0.67 in classifying individuals with or without prominent AF. PPV was non-inferior (p = 0.03) for individuals with an apnea-hypopnea index (AHI) ≥15 versus AHI < 15. Over 22% of correctly identified prominent AF rhythm cases were not previously documented as AF in SHHS1. Significance: Individuals with prominent AF can be automatically diagnosed from an overnight single-channel ECG recording, with an accuracy unaffected by the presence of moderate-to-severe obstructive sleep apnea. This approach enables identifying a large proportion of AF individuals that were otherwise missed by regular care.

Background: The 12-lead electrocardiogram (ECG) is a standard tool used in medical practice for identifying cardiac abnormalities. The 2020 PhysioNet/Computing in Cardiology Challenge addresses the topic of automated classification of 12-lead ECG. Methods: Two machine learning strategies were implemented: a feature engineering approach based on the engineering of physiological features (or 'digital biomarkers') and a deep learning approach. Two sets of features were engineered: (1) capturing the interval variation between consecutive heartbeats, commonly called heart rate variability (HRV) measures and (2) using morphological biomarkers (e.g. QT interval, QRS width). A total of 16 HRV and 97 morphological biomarkers were implemented in python for each lead. A random forest (RF) model was trained using 5-fold cross validation to optimize the model hyperparameters. For the deep learning approach, a residual neural network (ResNet) architecture was used. The RF and ResNet were also combined in an ensemble learning (EL). The dataset was divided into 80%-20% stratified training-test sets. Results: on the local test set we achieved a Challenge score of 0.65 using the FE approach, 0.52 using the DL approach and 0.66 using the EL approach. For technical reasons we did not manage to score our models on the Challenge hidden test set.

Catecholaminergic polymorphic ventricular tachycardia (CPVT) is a stress-induced ventricular arrhythmia associated with rhythm disturbance and impaired sinoatrial node cell (SANC) automaticity (pauses). Mutations associated with dysfunction of Ca²⁺-related mechanisms have been shown to be present in CPVT. These dysfunctions include impaired Ca²⁺ release from the ryanodine receptor (i.e., RyR2^R4496C mutation) or binding to calsequestrin 2 (CASQ2). In SANC, Ca²⁺ signaling directly and indirectly mediates pacemaker function. We address here the following research questions: (i) what coupled-clock mechanisms and pathways mediate pacemaker mutations associated with CPVT in basal and in response to β-adrenergic stimulation? (ii) Can different mechanisms lead to the same CPVT-related pacemaker pauses? (iii) Can the mutation-induced deteriorations in SANC function be reversed by drug intervention or gene manipulation? We used a numerical model of mice SANC that includes membrane and intracellular mechanisms and their interconnected signaling pathways. In the basal state of RyR2^R4496C SANC, the model predicted that the Na⁺-Ca²⁺ exchanger current (I_NCX) and T-type Ca²⁺ current (I_CaT) mediate between changes in Ca²⁺ signaling and SANC dysfunction. Under β-adrenergic stimulation, changes in cAMP-PKA signaling and the sodium currents (I_Na), in addition to I_NCX and I_CaT, mediate between changes in Ca²⁺ signaling and SANC automaticity pauses. Under basal conditions in Casq2^−/−, the same mechanisms drove changes in Ca²⁺ signaling and subsequent pacemaker dysfunction. However, SANC automaticity pauses in response to β–AR stimulation were mediated by I_CaT and I_Na. Taken together, distinct mechanisms can lead to CPVT-associated SANC automaticity pauses. In addition, we predict that specifically increasing SANC cAMP-PKA activity by either a pharmacological agent (IBMX, a phosphodiesterase (PDE) inhibitor), gene manipulation (overexpression of adenylyl cyclase 1/8) or direct manipulation of the SERCA phosphorylation target through changes in gene expression, compensate for the impairment in SANC automaticity. These findings suggest new insights for understanding CPVT and its therapeutic approach.

Objective: Portable oximetry has been shown to be a promising candidate for large-scale obstructive sleep apnea screening. In polysomnography (PSG), the gold standard OSA diagnosis test, the oxygen desaturation index (ODI) is usually computed from desaturation events occurring during sleep periods only, i.e. overnight desaturations occurring during or overlapping with a wake state are excluded. However, for unattended home oximetry, all desaturations are taken into account since no reference electroencephalogram is available for sleep staging. We aim to evaluate the hypothesis that the predictive power of oximetry for OSA screening is not impaired when reference sleep stages are not available. Approach: We used a PSG clinical database of 887 individuals from a representative S o Paulo (Brazil) population sample. Using features derived from the oxygen saturation time series and demographic information, OxyDOSA, a published machine learning model, was trained to distinguish between non-OSA and OSA individuals using the ODI computed while including versus excluding overnight desaturations overlapping with a wake period, thus mimicking portable and PSG oximetry analyses, respectively. Main results: When excluding wake desaturations, the OxyDOSA model had an AUROC = 94.9 ± 1.6, Se = 85.9 ± 2.8, Sp = 90.1 ± 2.6 and F₁ = 86.4 ± 2.7. When considering wake desaturations, the OxyDOSA model had an AUROC = 94.4 ± 1.6, Se = 88.0 ± 2.0, Sp = 87.7 ± 2.9 and F₁ = 86.2 ± 2.4. Non-inferiority was demonstrated (p = 0.049) at a tolerance level of 3%. In addition, analysis of the desaturations excluded by PSG oximetry analysis suggests that up to 21% of the total number of desaturations might actually be related to apneas or hypopneas. Significance: This analysis of a large representative population sample provided strong evidence that the predictive power of oximetry for OSA screening using the OxyDOSA model is not impaired when reference sleep stages are not available. This finding motivates the usage of portable oximetry for OSA screening.

Fetal electrocardiography (fECG) is a promising alternative to cardiotocography continuous fetal monitoring. Robust extraction of the fetal signal from the abdominal mixture of maternal and fetal electrocardiograms presents the greatest challenge to effective fECG monitoring. This is mainly due to the low amplitude of the fetal versus maternal electrocardiogram and to the non-stationarity of the recorded signals. In this review, we highlight key developments in advanced signal processing algorithms for non-invasive fECG extraction and the available open access resources (databases and source code). In particular, we highlight the advantages and limitations of these algorithms as well as key parameters that must be set to ensure their optimal performance. Improving or combining the current or developing new advanced signal processing methods may enable morphological analysis of the fetal electrocardiogram, which today is only possible using the invasive scalp electrocardiography method.

Sepsis is a life-threatening reaction to an infection, responsible for 6 million deaths globally each year. Moreover, this condition is one of the major cost to healthcare. Our aim is to develop a new technique for the early detection of the sepsis onset. Such an early detection would allow for the improvement of sepsis outcome.Our technique is based on the assumption that accurate and early prediction of sepsis requires to be able to predict the evolution of the vital signs. This idea was translated in the use of a recurrent neural network, a Long Short-term memory (LSTM) network, which was trained to accomplish two tasks: the prediction of (i) sepsis and (ii) the vital signs at time t+6. We assume that the use of this auxiliary task allows for a better training of the network given the low prevalence of sepsis. The network consists in three modules: (i) an embedding module aiming at providing a compact representation of the inputs, (ii) a recurrent module with three LSTMs layers with highway connection between each layer (iii) the prediction modules consisting in linear layers for the prediction of two tasks.The network achieved a final utility score of 0.309 on the full hidden test set (0.387 on the test set A, 0.365 on the set B , and -0.148 on the set C). The team name was "IADI".Further improvements are required before transferring such an approach into clinical practice.

Despite the improvements made in perinatal medicine during the 20th century, stillbirths still occur for 1 in 160 pregnancies in the US which represents a total of 26, 000 fetal deaths each year. In addition, approximately 1 in 1000 fetuses experience oxygen deprivation during labor which is severe enough to cause brain injury. It is estimated that half of these cases of birth-related injury are preventable. Incorrect cardiotocography (CTG) interpretation is leading the list of causes.Intrapartum CTG is used routinely to measure maternal uterine pressure and fetal heart rate (FHR). Antepartum CTG monitoring is used to identify fetuses at risk of intrauterine hypoxia and acidaemia. As early as 28 weeks of gestation, analysis of the FHR trace is used as a nonstress test to assess the fetal well-being. In the perinatal period, timely, appropriate intervention can avoid fetal neurological damage or death. The CTG is visually assessed by a clinician or interpreted by computer analysis. In the context of labor monitoring, the CTG is used for continuous fetal monitoring. An abnormal heart rate will lead the clinician to perform a cesarean.With the recent advances in machine learning and statistical signal analysis new algorithms for assessing fetal antepartum or intrapartum health status are being elaborated. These algorithms process signals recorded by CTG monitors or alternative monitoring techniques such as scalp electrocardiography or non-invasive fetal electrocardiography. This session discusses the history of fetal monitoring, its current challenges and the prospects opened by recent algorithmic development.

Purpose: It is uncertain whether right ventricular (RV) lead position in cardiac resynchronization therapy impacts response. There has been little detailed analysis of the activation patterns in RV septal pacing (RVSP), especially in the CRT population. We compare left bundle branch block (LBBB) activation patterns with RV pacing (RVP) within the same patients with further comparison between RV apical pacing (RVAP) and RVSP. Methods: Body surface mapping was undertaken in 14 LBBB patients after CRT implantation. Nine patients had RVAP, 5 patients had RVSP. Activation parameters included left ventricular total activation time (LVtat), biventricular total activation time (VVtat), interventricular electrical synchronicity (VVsync), and dispersion of left ventricular activation times (LVdisp). The direction of activation wave front was also compared in each patient (wave front angle (WFA)). In silico computer modelling was applied to assess the effect of RVAP and RVSP in order to validate the clinical results. Results: Patients were aged 64.6 ± 12.2 years, 12 were male, 8 were ischemic. Baseline QRS durations were 157 ± 18 ms. There was no difference in VVtat between RVP and LBBB but a longer LVtat in RVP (102.8 ± 19.6 vs. 87.4 ± 21.1 ms, p = 0.046). VVsync was significantly greater in LBBB (45.1 ± 20.2 vs. 35.9 ± 17.1 ms, p = 0.01) but LVdisp was greater in RVP (33.4 ± 5.9 vs. 27.6 ± 6.9 ms, p = 0.025). WFA did rotate clockwise with RVP vs. LBBB (82.5 ± 25.2 vs. 62.1 ± 31.7 ^op = 0.026). None of the measurements were different to LBBB with RVSP; however, the differences were preserved with RVAP for VVsync, LVdisp, and WFA. In silico modelling corroborated these results. Conclusions: RVAP activation differs from LBBB where RVSP appears similar. Trial registration: (ClinicalTrials.gov identifier: NCT01831518).

Background: The growing awareness for the high prevalence of obstructive sleep apnea (OSA) coupled with the dramatic proportion of undiagnosed individuals motivates the elaboration of a simple but accurate screening test. This study assesses, for the first time, the performance of oximetry combined with demographic information as a screening tool for identifying OSA in a representative (i.e. non-referred) population sample. Methods: A polysomnography (PSG) clinical database of 887 individuals from a representative population sample of São Paulo's city (Brazil) was used. Using features derived from the oxygen saturation signal during sleep periods and demographic information, a logistic regression model (termed OxyDOSA) was trained to distinguish between non-OSA and OSA individuals (mild, moderate, and severe). The OxyDOSA model performance was assessed against the PSG-based diagnosis of OSA (AASM 2017) and compared to the NoSAS and STOP-BANG questionnaires. Findings: The OxyDOSA model had mean AUROC = 0.94 ± 0.02, Se = 0.87 ± 0.04 and Sp = 0.85 ± 0.03. In particular, it did not miss any of the 75 severe OSA individuals. In comparison, the NoSAS questionnaire had AUROC = 0.83 ± 0.03, and missed 23/75 severe OSA individuals. The STOP-BANG had AUROC = 0.77 ± 0.04 and missed 14/75 severe OSA individuals. Interpretation: We provide strong evidence on a representative population sample that oximetry biomarkers combined with few demographic information, the OxyDOSA model, is an effective screening tool for OSA. Our results suggest that sleep questionnaires should be used with caution for OSA screening as they fail to identify many moderate and even some severe cases. The OxyDOSA model will need to be further validated on data recorded using overnight portable oximetry.

Objective: To assess whether noninvasive fetal electrocardiography (NI-FECG) enables the diagnosis of fetal arrhythmias. Methods: A total of 500 echocardiography and NI-FECG recordings were collected from pregnant women during a routine medical visit in this multicenter study. All the cases with fetal arrhythmias (n = 12) and a matching number of control (n = 14) were used. Two perinatal cardiologists analyzed the extracted NI-FECG while blinded to the echocardiography. The NI-FECG-based diagnosis was compared with the reference fetal echocardiography diagnosis. Results: NI-FECG and fetal echocardiography agreed on all cases (Ac = 100%) on the presence of an arrhythmia or not. However, in one case, the type of arrhythmia identified by the NI-FECG was incorrect because of the low resolution of the extracted fetal P-wave, which prevented resolving the mechanism (2:1 atrioventricular conduction) of the atrial tachycardia. Conclusion: It is possible to diagnose fetal arrhythmias using the NI-FECG technique. However, this study identifies that improvement in algorithms for reconstructing the P-wave is critical to systematically resolve the mechanisms underlying the arrhythmias. The elaboration of a NI-FECG Holter device will offer new opportunities for fetal diagnosis and remote monitoring of problematic pregnancies because of its low-cost, noninvasiveness, portability, and minimal setup requirements.

In the heart, mitochondria are arranged in pairs sandwiched between the contractile machinery, which is the major ATP consumer. Thus, in response to the contraction-relaxation cycle of the cell, the mitochondrial membrane should deform accordingly. Membrane deformations in isolated ATP synthesis or in isolated mitochondria affect ATP production. However, it is unknown whether physiological deformation of the mitochondrial membrane in response to the contraction-relaxation cycle can act as a bioenergetic signaling mechanism between ATP demand to supply. We used both experimental and computational tools to reveal whether bioenergetic feedback exists between heart cell contractile machinery and mitochondrial three-dimensional (3D) deformations. We measured the mitochondrial 3D deformation in contracting rabbit cardiac myocytes and used published data on rat cardiac myocytes. These measurements were an input to a novel biophysics model that includes a description of ionic molecules on the mitochondrial membrane, Ca²⁺ cycling, and mitochondrial membrane stress. As is the case for rat cardiomyocytes, in rabbit cardiomyocytes, the mitochondrial length contracted and expanded with a similar dynamic as the sarcomere length. In contrast, the mitochondrial width expanded and then contracted with a similar dynamic as the mitochondrial length. Differences in the extent of deformation and fractional deformation between the width- and thick-axes were quantified and interpreted as the degree anisotropy between those respective axes. Finally, the model predicts that significant bioenergetic feedback between heart cell contractile machinery and mitochondrial 3D deformations does exist in unloaded rabbit and rat cells. However, this feedback is not a dominant mechanism in ATP supply to demand matching.

Background: The time variation between consecutive heartbeats is commonly referred to as heart rate variability (HRV). Loss of complexity in HRV has been documented in several cardiovascular diseases and has been associated with an increase in morbidity and mortality. However, the mechanisms that control HRV are not well understood. Animal experiments are the key to investigating this question. However, to date, there are no standard open source tools for HRV analysis of mammalian electrocardiogram (ECG) data and no centralized public databases for researchers to access. Methods: We created an open source software solution specifically designed for HRV analysis from ECG data of multiple mammals, including humans. We also created a set of public databases of mammalian ECG signals (dog, rabbit and mouse) with manually corrected R-peaks (>170,000 annotations) and signal quality annotations. The platform (software and databases) is called PhysioZoo. Results: PhysioZoo makes it possible to load ECG data and perform very accurate R-peak detection (F1 > 98%). It also allows the user to manually correct the R-peak locations and annotate low signal quality of the underlying ECG. PhysioZoo implements state of the art HRV measures adapted for different mammals (dogs, rabbits, and mice) and allows easy export of all computed measures together with standard data representation figures. PhysioZoo provides databases and standard ranges for all HRV measures computed on healthy, conscious humans, dogs, rabbits, and mice at rest. Study of these measures across different mammals can provide new insights into the complexity of heart rate dynamics across species. Conclusion: PhysioZoo enables the standardization and reproducibility of HRV analysis in mammalian models through its open source code, freely available software, and open access databases. PhysioZoo will support and enable new investigations in mammalian HRV research. The source code and software are available on www.physiozoo.com.

Introduction: Heart rate variability (HRV) analysis tools have been mainly available for analysis of human electrocardiographic derived heart rate. We explore extending HRV analysis to two additional dimensions: (1) analysis across multiple mammalian species and (2) analysis across different levels of integration for example sinoatrial tissue. Methods: We analyzed the beating rate variability (BRV) across the two additional dimensions using the PhysioZoo computer program that we recently introduced. We used published databases of electrocardiograms from four mammal types: human (n=18), dog (n=17, rabbit (n=4) and mouse (n=8) We computed the BRV measures for each. We also show how the PhysioZoo program can be used for the analysis of sinoatrial node tissue BRV. Results: The study of typical mammalian heart and respiration rates (obtained from the dominant high frequency peak) revealed a linear relationship between these two quantities. Analysis of the rabbit sinoatrial node tissue BRV showed that it had reduced overall variability when compared to in vivo heart BRV.

Background: Power spectral density (PSD) analysis of the heartbeat intervals in the three main frequency bands [very low frequency (VLF), low frequency (LF), and high frequency (HF)] provides a quantitative non-invasive tool for assessing the function of the cardiovascular control system. In humans, these frequency bands were standardized following years of empirical evidence. However, no quantitative approach has justified the frequency cutoffs of these bands and how they might be adapted to other mammals. Defining mammal-specific frequency bands is necessary if the PSD analysis of the HR is to be used as a proxy for measuring the autonomic nervous system activity in animal models. Methods: We first describe the distribution of prominent frequency peaks found in the normalized PSD of mammalian data using a Gaussian mixture model while assuming three components corresponding to the traditional VLF, LF and HF bands. We trained the algorithm on a database of human electrocardiogram recordings (n = 18) and validated it on databases of dogs (n = 17) and mice (n = 8). Finally, we tested it to predict the bands for rabbits (n = 4) for the first time. Results: Double-logarithmic analysis demonstrates a scaling law between the GMM-identified cutofffrequencies and the typical heart rate (HR_m): fVLF-LF = 0.0037·HR_m0.58, f_LF-HF = 0.0017·HR_m1.01 and f_HFup = 0.0128·HR_m0.86. We found that the band cutofffrequencies and Gaussian mean scale with a power law of 1/4 or 1/8 of the typical body mass (BM_m), thus revealing allometric power laws. Conclusion: Our automated data-driven approach allowed us to define the frequency bands in PSD analysis of beat-to-beat time series from different mammals. The scaling law between the band frequency cutoffs and the HR_m can be used to approximate the PSD bands in other mammals.

Background: In parallel to the introduction of mobile communication devices with high computational power and internet connectivity, high-quality and low-cost health sensors have also become available. However, although the technology does exist, no clinical mobile system has been developed to monitor the R peaks from electrocardiogram recordings in real time with low false positive and low false negative detection. Implementation of a robust electrocardiogram R peak detector for various arrhythmogenic events has been hampered by the lack of an efficient design that will conserve battery power without reducing algorithm complexity or ease of implementation. Objective: Our goals in this paper are (1) to evaluate the suitability of the MATLAB Mobile platform for mHealth apps and whether it can run on any phone system, and (2) to embed in the MATLAB Mobile platform a real-time electrocardiogram R peak detector with low false positive and low false negative detection in the presence of the most frequent arrhythmia, atrial fibrillation. Methods: We implemented an innovative R peak detection algorithm that deals with motion artifacts, electrical drift, breathing oscillations, electrical spikes, and environmental noise by low-pass filtering. It also fixes the signal polarity and deals with premature beats by heuristic filtering. The algorithm was trained on the annotated non-atrial fibrillation MIT-BIH Arrhythmia Database and tested on the atrial fibrillation MIT-BIH Arrhythmia Database. Finally, the algorithm was implemented on mobile phones connected to a mobile electrocardiogram device using the MATLAB Mobile platform. Results: Our algorithm precisely detected the R peaks with a sensitivity of 99.7% and positive prediction of 99.4%. These results are superior to some state-of-the-art algorithms. The algorithm performs similarly on atrial fibrillation and non-atrial fibrillation patient data. Using MATLAB Mobile, we ran our algorithm in less than an hour on both the iOS and Android system. Our app can accurately analyze 1 minute of real-time electrocardiogram signals in less than 1 second on a mobile phone. Conclusions: Accurate real-time identification of heart rate on a beat-to-beat basis in the presence of noise and atrial fibrillation events using a mobile phone is feasible.

(Biophysical Journal 114, 1176–1189; March 13, 2018) The second author's name, Joachim Behar, was spelled incorrectly in the published article. The journal apologizes for any inconvenience.

Recent data suggest that cardiac pacemaker cell function is determined by numerous time-, voltage-, and Ca-dependent interactions of cell membrane electrogenic proteins (M-clock) and intracellular Ca cycling proteins (Ca-clock), forming a coupled-clock system. Many aspects of the coupled-clock system, however, remain underexplored. The key players of the system are Ca release channels (ryanodine receptors), generating local Ca releases (LCRs) from sarcoplasmic reticulum, electrogenic Na/Ca exchanger (NCX) current, and L-type Ca current (I_CaL). We combined numerical model simulations with experimental simultaneous recordings of action potentials (APs) and Ca to gain further insight into the complex interactions within the system. Our simulations revealed a positive feedback mechanism, dubbed AP ignition, which accelerates the diastolic depolarization (DD) to reach AP threshold. The ignition phase begins when LCRs begin to occur and the magnitude of inward NCX current begins to increase. The NCX current, together with funny current and T-type Ca current accelerates DD, bringing the membrane potential to I_CaL activation threshold. During the ignition phase, I_CaL-mediated Ca influx generates more LCRs via Ca-induced Ca release that further activates inward NCX current, creating a positive feedback. Simultaneous recordings of membrane potential and confocal Ca images support the model prediction of the positive feedback among LCRs and I_CaL, as diastolic LCRs begin to occur below and continue within the voltage range of I_CaL activation. The ignition phase onset (identified within the fine DD structure) begins when DD starts to notably accelerate (∼0.15 V/s) above the recording noise. Moreover, the timing of the ignition onset closely predicted the duration of each AP cycle in the basal state, in the presence of autonomic receptor stimulation, and in response to specific inhibition of either the M-clock or Ca-clock, thus indicating general importance of the new coupling mechanism for regulation of the pacemaker cell cycle duration, and ultimately the heart rate.

Age-related deterioration of pacemaker function has been documented in mammals, including humans. In aged isolated sinoatrial node tissues and cells, reduction in the spontaneous action potential (AP) firing rate was associated with deterioration of intracellular and membrane mechanisms; however, their relative contribution to age-associated deficient pacemaker function is not known. Interestingly, pharmacological interventions that increase posttranslation modification signaling activities can restore the basal and maximal AP firing rate, but the identities of the protein targets responsible for AP firing rate restoration are not known. Here, we developed a numerical model that simulates the function of a single mouse pacemaker cell. In addition to describing membrane and intracellular mechanisms, the model includes descriptions of autonomic receptor activation pathways and posttranslation modification signaling cascades. The numerical model shows that age-related deterioration of pacemaker function is related to impaired intracellular and membrane mechanisms: HCN₄, T-type channels, and phospholamban functions, as well as the node connecting these mechanisms, i.e., intracellular Ca²⁺ and posttranslation modification signaling. To explain the restored maximal beating rate in response to maximal phosphodiesterase (PDE) inhibition, autonomic receptor stimulation, or infused cyclic adenosine monophosphate (cAMP), the model predicts that phospholamban phosphorylation by protein kinase A (PKA) and HCN₄ sensitivity to cAMP are altered in advanced age. Moreover, alteration in PKA and cAMP sensitivity can also explain age-reduced sensitivity to PDE inhibition and autonomic receptor stimulation. Finally, the numerical model suggests two pharmacological approaches and one gene manipulation method to restore the basal beating rate of aged pacemaker cells to that of normal adult cells. In conclusion, our numerical model shows that impaired membrane and intracellular mechanisms and the nodes that couple them can lead to deteriorated pacemaker function. By increasing posttranslation modification signaling, the deteriorated basal and maximal age-associated beating rate can be restored to adult levels.

Catecholaminergic polymorphic ventricular tachycardia (CPVT) is a stress-provoked ventricular arrhythmia, which also manifests sinoatrial node (SAN) dysfunction. We recently showed that SK4 calcium-activated potassium channels are important for automaticity of cardiomyocytes derived from human embryonic stem cells. Here SK4 channels were identified in human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) from healthy and CPVT2 patients bearing a mutation in calsequestrin 2 (CASQ2-D307H) and in SAN cells from WT and CASQ2-D307H knock-in (KI) mice. TRAM-34, a selective blocker of SK4 channels, prominently reduced delayed afterdepolarizations and arrhythmic Ca²⁺ transients observed following application of the β-adrenergic agonist isoproterenol in CPVT2-derived hiPSC-CMs and in SAN cells from KI mice. Strikingly, in vivo ECG recording showed that intraperitoneal injection of the SK4 channel blockers, TRAM-34 or clotrimazole, greatly reduced the arrhythmic features of CASQ2-D307H KI and CASQ2 knockout mice at rest and following exercise. This work demonstrates the critical role of SK4 Ca²⁺-activated K⁺ channels in adult pacemaker function, making them promising therapeutic targets for the treatment of cardiac ventricular arrhythmias such as CPVT.

Introduction: Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia. Its prevalence is 1-2% of the general population and it is associated with increased risk of mortality and morbidity. Methods: The AliveCor mobile electrocardiogram (ECG) device was used to collect data. The Physionet Challenge aimed to create an intelligent algorithm for automated rhythm and quality classification. A database of 8528 single lead ECG was used for training and a closed database of 3658 ECG recordings was used for testing the participants algorithms on the Challenge server. The RR interval time-series was first estimated using a R-peak detector. Signal quality was estimated on a second-by-second basis and the continuous sub-segment with the highest quality was selected for further analysis. A number of features were estimated: heart rate variability (time domain based, fragmentation, coefficient of sample entropy etc.), ECG morphology (QRS length, QT interval etc.) and the presence of ectopic beats. The features were used to train support vector machine classifiers in a one-vs.-rest approach. Results: For the final score of the challenge we obtained an overall F₁ measure on the test set of 0.80. Conclusion: The feature based machine learning approach showed high performance in distinguishing between the different rhythms represented in the Challenge. This opens the horizon for computer automated interpretation of single lead mobile ECG.

Introduction: Awareness of the high prevalence of obstructive sleep apnoea (OSA) coupled with the dramatic proportion of undiagnosed individuals has been motivating research in the past two decades for elaborating sleep questionnaires that could help in early detection of OSA. This work aims to assess the predictive value of a subset of features that are used in the STOP-BANG questionnaire (BMI, age, gender, neck size) through a rigorous statistical analysis. Methods: A clinical database of 856 individuals referred to the sleep clinic for polysomnography was used to estimate the predictive value of the individual and combined demographic features in identifying OSA individuals. Four of the eight STOP-BANG questionnaire features were available in this database. These features were combined using a logistic regression model. The data were divided into train (80%) and test set (20%). Results: Results on the test set were 83.3% sensitivity, 45.8% specificity, 62.2% positive predictive value, 71.7% negative predictive value. The area under the curve (AUC) had a mean of 0.717. The results showed that combining the available subset of STOP-BANG questionnaire features gave similar performance to the full STOP-BANG questionnaire as reported in a number of studies. This highlights that combining features using a machine learning framework may improve the STOP-BANG questionnaire prediction. Alternatively, it may indicate that there are redundant or uninformative questions in the STOP-BANG questionnaire.

BACKGROUND: Complete atrioventricular block in fetuses is known to be mostly associated with autoimmune disease and can be irreversible if no steroids treatment is provided. Conventional methods used in clinical practice for diagnosing fetal arrhythmia are limited since they do not reflect the primary electrophysiological conduction processes that take place in the myocardium. The non-invasive fetal electrocardiogram has the potential to better support fetal arrhythmias diagnosis through the continuous analysis of the beat to beat variation of the fetal heart rate and morphological analysis of the PQRST complex.

CASE PRESENTATION: We present two retrospective case reports on which atrioventricular block diagnosis could have been supported by the non-invasive fetal electrocardiogram. The two cases comprised a 22-year-old pregnant woman with the gestational age of 31 weeks and a 25-year-old pregnant woman with the gestational age of 41 weeks. Both women were admitted to the Department of Maternal and Fetal Medicine at the Kyiv and Kharkiv municipal perinatal clinics. Patients were observed using standard fetal monitoring methods as well as the non-invasive fetal electrocardiogram. The non-invasive fetal electrocardiographic recordings were analyzed retrospectively, where it is possible to identify the presence of the atrioventricular block.

CONCLUSIONS: This study demonstrates, for the first time, the feasibility of the non-invasive fetal electrocardiogram as a supplementary method to diagnose of the fetal atrioventricular block. Combined with current fetal monitoring techniques, non-invasive fetal electrocardiography could support clinical decisions.

Sinoatrial nodal cells (SANCs) generate spontaneous action potentials (APs) that control the cardiac rate. The brain modulates SANC automaticity, via the autonomic nervous system, by stimulating membrane receptors that activate (adrenergic) or inactivate (cholinergic) adenylyl cyclase (AC). However, these opposing afferents are not simply additive. We showed that activation of adrenergic signaling increases AC-cAMP/PKA signaling, which mediates the increase in the SANC AP firing rate (i.e., positive chronotropic modulation). However, there is a limited understanding of the underlying internal pacemaker mechanisms involved in the crosstalk between cholinergic receptors and the decrease in the SANC AP firing rate (i.e., negative chronotropic modulation). We hypothesize that changes in AC-cAMP/PKA activity are crucial for mediating either decrease or increase in the AP firing rate and that the change in rate is due to both internal and membrane mechanisms. In cultured adult rabbit pacemaker cells infected with an adenovirus expressing the FRET sensor AKAR3, PKA activity and AP firing rate were tightly linked in response to either adrenergic receptor stimulation (by isoproterenol, ISO) or cholinergic stimulation (by carbachol, CCh). To identify the main molecular targets that mediate between PKA signaling and pacemaker function, we developed a mechanistic computational model. The model includes a description of autonomic-nervous receptors, post- translation signaling cascades, membrane molecules, and internal pacemaker mechanisms. Yielding results similar to those of the experiments, the model simulations faithfully reproduce the changes in AP firing rate in response to CCh or ISO or a combination of both (i.e., accentuated antagonism). Eliminating AC-cAMP-PKA signaling abolished the core effect of autonomic receptor stimulation on the AP firing rate. Specifically, disabling the phospholamban modulation of the SERCA activity resulted in a significantly reduced effect of CCh and a failure to increase the AP firing rate under ISO stimulation. Directly activating internal pacemaker mechanisms led to a similar extent of changes in the AP firing rate with respect to brain receptor stimulation. Thus, Ca²⁺ and cAMP/PKA-dependent phosphorylation limits the rate and magnitude of chronotropic changes in the spontaneous AP firing rate.

Non-invasive fetal electrocardiography (NI-FECG) is a promising alternative continuous fetal monitoring method that has the potential to allow morphological analysis of the FECG. However, there are a number of challenges associated with the evaluation of morphological parameters from the NI-FECG, including low signal to noise ratio of the NI-FECG and methodological challenges for getting reference annotations and evaluating the accuracy of segmentation algorithms. This work aims to validate the measurement of the fetal QT interval in term laboring women using a NI-FECG electrocardiogram monitor. Fetal electrocardiogram data were recorded from 22 laboring women at term using the NI-FECG and an invasive fetal scalp electrode simultaneously. A total of 105 one-minute epochs were selected for analysis. Three pediatric electrophysiologists independently annotated individual waveforms and averaged waveforms from each epoch. The intervals measured on the averaged cycles taken from the NI-FECG and the fetal scalp electrode showed a close agreement; the root mean square error between all corresponding averaged NI-FECG and fetal scalp electrode beats was 13.6 ms, which is lower than the lowest adult root mean square error of 16.1 ms observed in related adult QT studies. These results provide evidence that NI-FECG technology enables accurate extraction of the fetal QT interval.

We aimed to determine how age-associated changes in mechanisms extrinsic and intrinsic to pacemaker cells relate to basal beating interval variability (BIV) reduction in vivo. Beating intervals (BIs) were measured in aged (23–25 months) and adult (3–4 months) C57BL/6 male mice (i) via ECG in vivo during light anesthesia in the basal state, or in the presence of 0.5 mg mL⁻¹ atropine + 1 mg mL⁻¹ propranolol (in vivo intrinsic conditions), and (ii) via a surface electrogram, in intact isolated pacemaker tissue. BIV was quantified in both time and frequency domains using linear and nonlinear indices. Although the average basal BI did not significantly change with age under intrinsic conditions in vivo and in the intact isolated pacemaker tissue, the average BI was prolonged in advanced age. In vivo basal BIV indices were found to be reduced with age, but this reduction diminished in the intrinsic state. However, in pacemaker tissue BIV indices increased in advanced age vs. adults. In the isolated pacemaker tissue, the sensitivity of the average BI and BIV in response to autonomic receptor stimulation or activation of mechanisms intrinsic to pacemaker cells by broad-spectrum phosphodiesterase inhibition declined in advanced age. Thus, changes in mechanisms intrinsic to pacemaker cells increase the average BIs and BIV in the mice of advanced age. Autonomic neural input to pacemaker tissue compensates for failure of molecular intrinsic mechanisms to preserve average BI. But this compensation reduces the BIV due to both the imbalance of autonomic neural input to the pacemaker cells and altered pacemaker cell responses to neural input.

Cardiac pacemaker cell function is regulated by a coupled-clock system that integrates molecular cues on the cell-membrane surface (i.e., membrane clock) and on the sarcoplasmic reticulum (SR) (i.e., Ca²⁺ clock). A recent study has shown that cotransfection of spontaneous beating cells (HEK293 cells and neonatal rat myocytes) with R524Q-mutant human hyperpolarization-activated cyclic nucleotide-gated molecules (the dominant component of funny channels) increases the funny channel's sensitivity to cAMP and leads to a decrease in spontaneous action potential (AP) cycle length (i.e., tachycardia). We hypothesize that in rabbit pacemaker cells, the same behavior is expected, and because of the coupled-clock system, the resultant steady-state decrease in AP cycle length will embody contributions from both clocks: the initial decrease in the spontaneous AP beating interval, arising from increased sensitivity of the f-channel to cAMP, will be accompanied by an increase in the adenylyl cyclase (AC)-cAMP-PKA-dependent phosphorylation activity, which will further decrease this interval. To test our hypothesis, we used the recently developed Yaniv-Lakatta pacemaker cell numerical model. This model predicts the cAMP signaling dynamics, as well as the kinetics and magnitude of protein phosphorylation in both normal and mutant pacemaker cells. We found that R524Q-mutant pacemaker cells have a shorter AP firing rate than that of wild-type cells and that gain in pacemaker function is the net effect of the R514Q mutation on the functioning of the coupled-clock system. Specifically, our results directly support the hypothesis that changes in Ca²⁺-activated AC-cAMP-PKA signaling are involved in the development of tachycardia in R524Q-mutant pacemaker cells.

Over the past decades, many studies have been published on the extraction of non-invasive foetal electrocardiogram (NI-FECG) from abdominal recordings. Most of these contributions claim to obtain excellent results in detecting foetal QRS (FQRS) complexes in terms of location. A small subset of authors have investigated the extraction of morphological features from the NI-FECG. However, due to the shortage of available public databases, the large variety of performance measures employed and the lack of open-source reference algorithms, most contributions cannot be meaningfully assessed. This article attempts to address these issues by presenting a standardised methodology for stress testing NI-FECG algorithms, including absolute data, as well as extraction and evaluation routines. To that end, a large database of realistic artificial signals was created, totaling 145.8 h of multichannel data and over one million FQRS complexes. An important characteristic of this dataset is the inclusion of several non-stationary events (e.g. foetal movements, uterine contractions and heart rate fluctuations) that are critical for evaluating extraction routines. To demonstrate our testing methodology, three classes of NI-FECG extraction algorithms were evaluated: blind source separation (BSS), template subtraction (TS) and adaptive methods (AM). Experiments were conducted to benchmark the performance of eight NI-FECG extraction algorithms on the artificial database focusing on: FQRS detection and morphological analysis (foetal QT and T/QRS ratio). The overall median FQRS detection accuracies (i.e. considering all non-stationary events) for the best performing methods in each group were 99.9% for BSS, 97.9% for AM and 96.0% for TS. Both FQRS detections and morphological parameters were shown to heavily depend on the extraction techniques and signal-to-noise ratio. Particularly, it is shown that their evaluation in the source domain, obtained after using a BSS technique, should be avoided. Data, extraction algorithms and evaluation routines were released as part of the fecgsyn toolbox on Physionet under an GNU GPL open-source license. This contribution provides a standard framework for benchmarking and regulatory testing of NI-FECG extraction algorithms.

Non-Invasive foetal electrocardiography (NI-FECG) represents an alternative foetal monitoring technique to traditional Doppler ultrasound approaches, that is non-invasive and has the potential to provide additional clinical information. However, despite the significant advances in the field of adult ECG signal processing over the past decades, the analysis of NI-FECG remains challenging and largely unexplored. This is mainly due to the relatively low signal-to-noise ratio of the FECG compared to the maternal ECG, which overlaps in both time and frequency. This article is intended to be used by researchers as a practical guide to NI-FECG signal processing, in the context of the above issues. It reviews recent advances in NI-FECG research including: publicly available databases, NI-FECG extraction techniques for foetal heart rate evaluation and morphological analysis, NI-FECG simulators and the methodology and statistics for assessing the performance of the extraction algorithms. Reference to the most recent work is given, recent findings are highlighted in the form of intermediate summaries, references to open source code and publicly available databases are provided and promising directions for future research are motivated. In particular we emphasise the need and specifications for building a new open reference database of NI-FECG signals, and the need for new algorithms to be benchmarked on the same database, employing the same evaluation statistics. Finally we motivate the need for research in NI-FECG to address morphological analysis, since this represent one of the most promising avenues for this foetal monitoring modality.

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 ± 0.63 ms, significantly outperforming the best Challenge entry (15.37 ± 2.13 ms) as well as the EM, mean, and median voting strategies (14.76 ± 0.52, 17.61 ± 0.55, and 14.43 ± 0.57 ms respectively with p < 0.0001). The BCLA could therefore provide accurate estimation for medical continuous-valued label tasks in an unsupervised manner even when the ground truth is not available.

Automatic processing and accurate diagnosis of pathological electrocardiogram (ECG) signals remains a challenge. As long-term ECG recordings continue to increase in prevalence, driven partly by the ease of remote monitoring technology usage, the need to automate ECG analysis continues to grow. In previous studies, a model-based ECG filtering approach to ECG data from healthy subjects has been applied to facilitate accurate online filtering and analysis of physiological signals. We propose an extension of this approach, which models not only normal and ventricular heartbeats, but also morphologies not previously encountered. A switching Kalman filter approach is introduced to enable the automatic selection of the most likely mode (beat type), while simultaneously filtering the signal using appropriate prior knowledge. Novelty detection is also made possible by incorporating a third mode for the detection of unknown (not previously observed) morphologies, and denoted asX-factor. This new approach is compared to state-of-the-art techniques for the ventricular heartbeat classification in theMIT-BIH arrhythmia and Incart databases. F₁ scores of 98.3% and 99.5% were found on each database, respectively, which are superior to other published algorithms' results reported on the same databases. Only 3% of all the beats were discarded as X-factor, and the majority of these beats contained high levels of noise. The proposed technique demonstrates accurate beat classification in the presence of previously unseen (and unlearned) morphologies and noise, and provides an automated method for morphological analysis of arbitrary (unknown) ECG leads.

The electrocardiogram (ECG) is a well studied signal from which many clinically relevant parameters can be derived, such as heart rate. A key component in the estimation of these parameters is the accurate detection of the R peak in the QRS complex. While corruption of the ECG by movement artefact or sensor failure can result in poor delineation of the R peak, use of synchronously measured signals could allow for resolution of the R peak even scenarios with poor quality ECG recordings. Robust estimation of R peak locations from multimodal signals facilitates real time monitoring and is likely to reduce false alarms due to inaccurate derived parameters. We propose a method which fuses R peaks detected on the ECG using an energy detector with those detected on the arterial blood pressure (ABP) waveform using the length transform. A signal quality index (SQI) for the two signals is then derived. The ECG SQI is based upon the agreement between two distinct peak detectors. The ABP SQI estimates the blood pressure at various phases in the cardiac cycle and only accepts the signal as good quality if the values are physiologically plausible. Detections from these two signals were merged by selecting the R peak detections from the signal with a higher SQI. The approach presented in this paper was evaluated on datasets provided for the Physionet/Computing in Cardiology Challenge 2014. The algorithm achieved a sensitivity of 95.1% and positive predictive value of 89.3% on an external evaluation set, and achieved a score of 91.5%. The method here demonstrated excellent performance across a variety of signal morphologies collected during clinical practice. Fusion of R peaks from other signals has the potential to provide informed estimates of the R peak location in situations where the ECG is noisy or completely absent. Source code for the algorithm is made available freely online.

This editorial reviews the background issues, the design, the key achievements, and the follow-up research generated as a result of the PhysioNet/Computing in Cardiology (CinC) Challenge 2014, published in the concurrent focus issue of Physiological Measurement. Our major focus was to accelerate the development and facilitate the comparison of robust methods for locating heart beats in long-term multi-channel recordings. A public (training) database consisting of 151 032 annotated beats was compiled from records that contained ECGs as well as pulsatile signals that directly reflect cardiac activity, and other signals that may have few or no observable markers of heart beats. A separate hidden test data set (consisting of 152 478 beats) is permanently stored at PhysioNet, and a public framework has been developed to provide researchers with the ability to continue to automatically score and compare the performance of their algorithms. A scoring criteria based on the averaging of gross sensitivity, gross positive predictivity, average sensitivity, and average positive predictivity is proposed. The top three scores (as of March 2015) on the hidden test data set were 93.64%, 91.50%, and 90.70%.

Obstructive sleep apnoea (OSA) is a sleep disorder with long-term consequences. Long-term effects include sleep-related issues and cardiovascular diseases. OSA is often diagnosed with an overnight sleep test called a polysomnogram. Monitoring can be costly with long wait times for diagnosis. In this paper, a novel OSA screening framework and prototype phone application are introduced. A database of 856 patients that underwent at-home polygraphy was collected. Features were derived from audio, actigraphy, photoplethysmography (PPG), and demographics, and used as the inputs of a support vector machine (SVM) classifier. The SVM was trained on 735 patients and tested on 121 patients. Classification on the test set had an accuracy of up to 92.2% when classifying subjects as having moderate or severe OSA versus being healthy or a snorer based on the clinicians' diagnoses. The signal processing and machine learning algorithms were ported to Java and integrated into the phone application - SleepAp. SleepAp records the body position, audio, actigraphy and PPG signals, and implements the clinically validated STOP-BANG questionnaire. It derives features from the signals and classifies the user as having OSA or not using the SVM trained on the clinical database. The resulting software could provide a new, easy-to-use, low-cost, and widely available modality for OSA screening.

Barriers to effective screening and management of NCDs in resource-constrained regions include limited availability of trained personnel, access to affordable automatic medical devices, and longitudinal clinical data. We present an end-to-end mHealth system which takes advantage of the almost universal availability of smartphones in order to address these barriers in a scalable and affordable manner. Our system includes simple, low-cost ($5-$20) and open-source peripherals that allow a minimally trained person to collect high-quality medical data at the point-of-care through a standard smartphone; allows the reliable transmission of clinical data even in the case of high-latency network connections; stores data in a cloud-based system, making patient records accessible anywhere; and enables both crowdsourced diagnostics and generation of annotated data for the research and development of automatic decision support and risk assessment systems. We show examples of the different elements of the system tailored for the management of cardiovascular disease and chronic obstructive pulmonary disease, which includes prototypes of the low-cost peripherals. In a validation study (of 40 volunteers), our smartphone-based blood pressure (BP) monitor was shown to measure BP, heart rate and respiration rate with a mean-absolute-error of less than 5 units from the reference values for 80% of the measurements.

Accurate foetal electrocardiogram (FECG) morphology extraction from non-invasive sensors remains an open problem. This is partly due to the paucity of available public databases. Even when gold standard information (i.e derived from the scalp electrode) is present, the collection of FECG can be problematic, particularly during stressful or clinically important events. In order to address this problem we have introduced an FECG simulator based on earlier work on foetal and adult ECG modelling. The open source foetal ECG synthetic simulator, fecgsyn, is able to generate maternal-foetal ECG mixtures with realistic amplitudes, morphology, beat-to-beat variability, heart rate changes and noise. Positional (rotation and translation-related) movements in the foetal and maternal heart due to respiration, foetal activity and uterine contractions were also added to the simulator. The simulator was used to generate some of the signals that were part of the 2013 PhysioNet Computing in Cardiology Challenge dataset and has been posted on Physionet.org (together with scripts to generate realistic scenarios) under an open source license. The toolbox enables further research in the field and provides part of a standard for industry and regulatory testing of rare pathological scenarios.

Despite significant advances in adult clinical electrocardiography (ECG) signal processing techniques and the power of digital processors, the analysis of non-invasive foetal ECG (NI-FECG) is still in its infancy. The Physionet/Computing in Cardiology Challenge 2013 addresses some of these limitations by making a set of FECG data publicly available to the scientific community for evaluation of signal processing techniques. The abdominal ECG signals were first preprocessed with a band-pass filter in order to remove higher frequencies and baseline wander. A notch filter to remove power interferences at 50Hz or 60Hz was applied if required. The signals were then normalized before applying various source separation techniques to cancel the maternal ECG. These techniques included: template subtraction, principal/independent component analysis, extended Kalman filter and a combination of a subset of these methods (FUSE method). Foetal QRS detection was performed on all residuals using a Pan and Tompkins QRS detector and the residual channel with the smoothest foetal heart rate time series was selected. The FUSE algorithm performed better than all the individual methods on the training data set. On the validation and test sets, the best Challenge scores obtained were E1 = 179.44, E2 = 20.79, E3 = 153.07, E4 = 29.62 and E5 = 4.67 for events 1-5 respectively using the FUSE method. These were the best Challenge scores for E1 and E2 and third and second best Challenge scores for E3, E4 and E5 out of the 53 international teams that entered the Challenge. The results demonstrated that existing standard approaches for foetal heart rate estimation can be improved by fusing estimators together. We provide open source code to enable benchmarking for each of the standard approaches described.

Despite the important advances achieved in the field of adult electrocardiography signal processing, the analysis of the non-invasive fetal electrocardiogram (NI-FECG) remains a challenge. Currently no gold standard database exists which provides labelled FECG QRS complexes (and other morphological parameters), and publications rely either on proprietary databases or a very limited set of data recorded from few (or more often, just one) individuals. The PhysioNet/Computing in Cardiology Challenge 2013 enables to tackle some of these limitations by releasing a set of NI-FECG data publicly to the scientific community in order to evaluate signal processing techniques for NI-FECG extraction. The Challenge aim was to encourage development of accurate algorithms for locating QRS complexes and estimating the QT interval in non-invasive FECG signals. Using carefully reviewed reference QRS annotations and QT intervals as a gold standard, based on simultaneous direct FECG when possible, the Challenge was designed to measure and compare the performance of participants' algorithms objectively. Multiple challenge events were designed to test basic FHR estimation accuracy, as well as accuracy in measurement of inter-beat (RR) and QT intervals needed as a basis for derivation of other FECG features. This editorial reviews the background issues, the design of the Challenge, the key achievements, and the follow-up research generated as a result of the Challenge, published in the concurrent special issue of Physiological Measurement.

The abdominal electrocardiogram (ECG) provides a non-invasive method for monitoring the fetal cardiac activity in pregnant women. However, the temporal and frequency overlap between the fetal ECG (FECG), the maternal ECG (MECG) and noise results in a challenging source separation problem. This work seeks to compare temporal extraction methods for extracting the fetal signal and estimating fetal heart rate. A novel method for MECG cancelation using an echo state neural network (ESN) based filtering approach was compared with the least mean square (LMS), the recursive least square (RLS) adaptive filter and template subtraction (TS) techniques. Analysis was performed using real signals from two databases composing a total of 4 h 22 min of data from nine pregnant women with 37,452 reference fetal beats. The effects of preprocessing the signals was empirically evaluated. The results demonstrate that the ESN based algorithm performs best on the test data with an F1 measure of 90.2% as compared to the LMS (87.9%), RLS (88.2%) and the TS (89.3%) techniques. Results suggest that a higher baseline wander high pass cutoff frequency than traditionally used for FECG analysis significantly increases performance for all evaluated methods. Open source code for the benchmark methods are made available to allow comparison and reproducibility on the public domain data.

For medical applications, the ground truth is ascertained through manual labels by clinical experts. However, significant inter-observer variability and various human biases limit accuracy. A probabilistic framework addresses these issues by comparing aggregated human and automated labels to provide a reliable ground truth, with no prior knowledge of the individual performance. As an alternative to median or mean voting strategies, novel contextual features (signal quality and physiology) were introduced to allow the Probabilistic Label Aggregator (PLA) to weight an algorithm or human based on its performance. As a proof of concept, the PLA was applied to QT interval (pro-arrhythmic indicator) estimation from the electrocardiogram using labels from 20 humans and 48 algorithms crowd-sourced from the 2006 PhysioNet/Computing in Cardiology Challenge database. For automatic annotations, the root mean square error of the PLA was 13.97 ± 0.46 ms, significantly outperforming the best Challenge entry (16.36 ms) as well as mean and median voting strategies (17.67 ± 0.56 ms and 14.44 ± 0.52 ms respectively with p<0.05). When selecting three annotators, the PLA improved the annotation accuracy over median aggregation by 10.7% for human annotators and 14.4% for automated algorithms. The PLA could therefore provide an improved ''gold standard'' for medical annotation tasks even when ground truth is not available.

This article presents a review of signals used for measuring physiology and activity during sleep and techniques for extracting information from these signals. We examine both clinical needs and biomedical signal processing approaches across a range of sensor types. Issues with recording and analysing the signals are discussed, together with their applicability to various clinical disorders. Both univariate and data fusion (exploiting the diverse characteristics of the primary recorded signals) approaches are discussed, together with a comparison of automated methods for analysing sleep.

Introduction: The electrocardiogram (ECG) allows for interpretation of the electrical activity of the heart. The information which can be derived from the foetal ECG (FECG) goes beyond heart rate and heart rate variability. However morphological analysis of the FECG waveform is usually not performed in clinical practice. Methods: A Bayesian Filtering Framework based on an Extended Kalman Filter (EKF) for extracting the FECG from a single abdominal channel is described using a training database of 20, one minute maternal-foetal mixtures and evaluated on 200, one minute mixtures. (Data was generated using the simulator, fecgsyn, used to generate a subset of the signals of the Physionet Challenge 2013.) A single pass of the EKF (EKFS) was performed to cancel out the maternal ECG (MECG) in order to build an average FECG morphology. A dual EKF (EKFD, i.e. where both the MECG and FECG cycle morphology were modelled) was then applied to separate the three sources present in the signal mixture (noise, MECG and FECG). A normalised root mean square error and absolute QT error after EKFS and EKFD were calculated. Results: An SNR improvement of 1.97 dB after EKFS and 14.14 dB after EKFD on the test set were achieved. Median absolute error on QT measurement was 17.0 ms for the EKFS and 4.0 ms for the EKFD. Conclusion: This work is a proof of concept that the EKFD allows accurate beat to beat extraction of the FECG morphology from abdominal recordings.

In foetal electrocardiographic monitoring, assessment of foetal QT (FQT) in identifying foetal hypoxia has been limited mainly due to the lack of available public databases with expert labels. Our proposed platform, CrowdLabel, a web-based open-source annotation system, was developed for crowdsourcing medical labels from multiple expert and/or non-expert annotators. We describe the platform and an example of use; to improve FQT estimation by creating reference labels against which automated algorithms can be benchmarked. A total of 501, 30s segments were extracted from 15 foetal ECG (FECG) recordings from a private database. 23 volunteers participated in the study and provided a total of 7,307 FQT annotations, which were aggregated using a probabilistic label aggregator (PLA). The best annotator identified by the PLA had a standard deviation in the change of FQT annotations of 13.35 ms and 35.52 ms when labelling FECG with 'very good' and 'poor' signal quality respectively. The PLA does not require any ground truth to identify the best annotator or annotations. Annotator accuracy was also shown to be a function of objective signal quality measures. The feasibility of the CrowdLabel annotation system for ECG crowdsourcing with an unknown ground truth, as well as the results of the first experiment conducted using such a platform have been demonstrated.

This paper describes, a graphical user interface (GUI) for synthetic simulation of maternal-foetal ECG mixtures. The GUI is directly linked to the previously introduced fecgsyn ECG model, which was used for producing a subset of data of the Physionet/Computing in Cardiology Challenge 2013. fecgsyngui serves as a tool that facilitates the use of the broad capabilities of fecgsyn. ECG mixtures can be generated with realistic beat-to-beat variability, morphology, amplitude modulation due to respiration and foetal movement. The GUI allows the user to visualise and generate a set of realistic scenarios to stress test and train separation and detection algorithms. Electrodes can be placed on the maternal abdomen using an interactive geometric representation. Simulation parameters and generated signals can be exported to .mat files for further processing. Using the underlying model, noise sources, ectopic beats, and multiple foetuses can be added. A series of pathological scenarios for situations (described in the original publication) for which data collection is problematic in reality, are included. These can be further customised to mirror specific pathological situations.

Introduction: The ECG Bayesian filtering framework has been shown to be a promising method to extract the foetal electrocardiogram (FECG) from abdominal recordings. This framework requires an estimation of the ECG morphology, which is obtained by approximating an average beat with a number of Gaussian kernels. This approximation results in a high dimensional nonlinear optimization problem (finding ideal positions, width and height for these kernels). Methods: Proposed methodologies in the literature initialize the optimization algorithm using fixed positions for the kernel functions. This contribution benchmarks alternative schemes for finding the Gaussian parameters, namely an approach based on the stationary wavelet transform and random search. The goal is minimizing the normalized mean squared error between the average beat and the approximated model, while increasing foetal QRS detection accuracy. Results: The suggested methods are able to produce improved morphology approximations of the averaged beat up 4.05% (depending on the selected method). The proposed method using the stationary wavelet transform improves the goodness of the fit, while reducing the computational load. However, no immediate improvement on the accuracy of FQRS detections was noticed. Such findings render the proposed method a promising tool. However, further research should be directed at transferring the improved fit to an improvement of FQRS detections.

Introduction: Intensive care unit patients are heavily monitored, and a number of clinically relevant parameters are routinely extracted from high resolution signals. In particular, heart rate is derived from intervals between pulses in pseudo-periodic signals such as the electrocardiogram (ECG) or arterial blood pressure (ABP) waveforms. However, poor signal quality and high noise levels can unfortunately lead to false localisation of these pulses (or peaks), resulting in incorrect estimates of heart rate. The goal of the 2014 Physionet/CinC Challenge was to encourage the creation of an intelligent system that fused information from different biosignals to create a robust set of peak detections. Methods: First, a set of peak detectors were evaluated on different cardiovascular signals. The detections were then fused using two different approaches: the first one was based on a calculated measures of signal quality for the ECG and ABP signals and the the second fusion technique was based on the regularity of the derived intervals between subsequent detections made on ECG, ABP, Stroke Volume and Photoplethysmogram signals. These techniques were developed using the MGH/MF database and submitted for scoring on the Challenge test-set. Conclusion: The best entries for the two approaches obtained an overall score of 87:88% and 87:66%, respectively, in phase III of the challenge, which provided the highest official score.

Obstructive Sleep Apnoea (OSA) is a disorder characterised by pauses in breathing during sleep which lead to deoxygenation. It is associated with increased risk of serious conditions including cardiovascular disease, diabetes and stroke. OSA prevalence ranges from 2-7% throughout the world, but it is estimated that up to 90% of cases are undiagnosed and untreated. OSA is usually diagnosed with a costly overnight sleep study, where a large number of physiological signals are recorded and analysed. The recent increase in adoption of smartphones has led to the proliferation of many sleep screening applications, but no existing application is based on scientific evidence. In this project we propose a novel OSA screening framework involving a pulse oximeter, a smartphone and an online server. The screening algorithms have been evaluated on a clinical database of 1007 patients and show 85.6% accuracy on the test set in identifying those diagnosed with OSA. This framework has the potential to provide a simple, cheap and widely-available modality for OSA screening, particularly in developing countries where conventional screening is limited.

A team from Oxford University conducted a 12-month 'smart handpump' trial in Kenya in 2013, funded by the UK Department for International Development (DfID), which tested a new maintenance model for rural handpump services. The service was based around the availability of information on handpump usage and functionality, generated by a GSM transmitter installed within the pump handle. For this trial, a version of the hardware and software was produced, that had to be operational in the field for a year, but produced with limited resources. The trial (67 community pumps) lead to a reduction in pump outage times from an average of 27 days per-trial to under three. This paper discusses some of the design decisions made and lessons learnt during the trial.

A system and method is provided for subject monitoring and acquisition of data for determining the mental and/or physical health of a subject. In an embodiment, the system and method provide automated subject monitoring and signal collection (acquisition) of data, for example activity data, associated with one or more of a plurality of metrics for mental and physical health, processing of the data to extract features associated with mental and/or physical health and then classifying the features extracted from the data using one or more classifiers into behavioral subject categories. The system and method can simultaneously combine multiple metrics to determine mental and/or physical health of a subject, which is robust to artifacts and missing data.

Sleep disorders are a common problem and contribute to a wide range of healthcare issues. The societal and financial costs of sleep disorders are enormous. Sleep-related disorders are often diagnosed with an overnight sleep test called a polysomnogram, or sleep study involving the measurement of brain activity through the electroencephalogram. Other parameters monitored include oxygen saturation, respiratory effort, cardiac activity (through the electrocardiogram), as well as video recording, sound and movement activity. Monitoring can be costly and removes the patients from their normal sleeping environment, preventing repeated unbiased studies. The recent increase in adoption of smartphones, with high quality on-board sensors has led to the proliferation of many sleep screening applications running on the phone. However, with the exception of simple questionnaires, no existing sleep-related application available for smartphones is based on scientific evidence. This paper reviews the existing smartphone applications landscape used in the field of sleep disorders and proposes possible advances to improve screening approaches.

Human bias and significant intra- and inter- observer variance exist in electrocardiogram QT interval evaluation. A Bayesian approach (BA) with an informative prior, that combines measures from multiple humans or algorithms as well as contextual information (such as heart rate and signal quality) was developed for inferring the true QT length. The developed method is compared to the mean and median voting approaches by computing the root-mean-square (RMS) error between the computed QT lengths and the reference annotations provided by the 2006 PhysioNet/Computing in Cardiology Challenge. The BA with features can reduces the human RMS error of QT estimates to 6.04ms and 13.97ms for automated algorithms, out-performing the results in the Challenge of 6.67ms and 16.34ms respectively. For three annotators, the BA had a 10.7% improvement over the next best voting strategy for manual annotations, and 14.4% for automated algorithms. For large numbers of annotators, the BA estimates became approximately equal to the best-performing annotator.

An automated algorithm to assess electrocardiogram (ECG) quality for both normal and abnormal rhythms is presented for false arrhythmia alarm suppression of intensive care unit (ICU) monitors. A particular focus is given to the quality assessment of a wide variety of arrhythmias. Data from three databases were used: the Physionet Challenge 2011 dataset, the MIT-BIH arrhythmia database, and the MIMIC II database. The quality of more than 33 000 single-lead 10 s ECG segments were manually assessed and another 12 000 bad-quality single-lead ECG segments were generated using the Physionet noise stress test database. Signal quality indices (SQIs) were derived from the ECGs segments and used as the inputs to a support vector machine classifier with a Gaussian kernel. This classifier was trained to estimate the quality of an ECG segment. Classification accuracies of up to 99% on the training and test set were obtained for normal sinus rhythm and up to 95% for arrhythmias, although performance varied greatly depending on the type of rhythm. Additionally, the association between 4050 ICU alarms from the MIMIC II database and the signal quality, as evaluated by the classifier, was studied. Results suggest that the SQIs should be rhythm specific and that the classifier should be trained for each rhythm call independently. This would require a substantially increased set of labeled data in order to train an accurate algorithm.

Introduction: Despite significant advances in adult clinical electrocardiography (ECG) signal processing techniques and the power of digital processors, the analysis of non-invasive foetal ECG (NI-FECG) is still in its infancy. The PhysioNet/Computing in Cardiology Challenge 2013 addresses some of these limitations by making publicly available a set of FECG data to the scientific community for evaluation of signal processing techniques.

The PhysioNet/CinC 2013 Challenge aimed to stimulate rapid development and improvement of software for estimating fetal heart rate (FHR), fetal interbeat intervals (FRR), and fetal QT intervals (FQT), from multichannel recordings made using electrodes placed on the mother's abdomen. For the challenge, five data collections from a variety of sources were used to compile a large standardized database, which was divided into training, open test, and hidden test subsets. Gold-standard fetal QRS and QT interval annotations were developed using a novel crowdsourcing framework.

The development of mHealth applications could facilitate the decrease of the healthcare costs in both high income and low to middle income regions. However, it is essential that mHealth software is validated on public databases. Moreover, public scrutiny of the algorithms is likely to lead to faster and lower cost innovation. In this paper, we therefore present a novel Java-based Android application offering advanced Electrocardiogram (ECG) processing techniques, including signal quality analysis and Atrial Fibrillation (AF) screening.

Obstructive Sleep Apnoea (OSA) is a sleep disorder with long term consequences. It is often diagnosed with an overnight sleep study or polysomnogram. Monitoring can be costly with long wait times for diagnosis. In this paper we describe a novel OSA screening framework and prototype phone application (app). A database of 856 patients that underwent at-home polysomnography was collected. Features were derived from audio, actigraphy, photoplethysmography (PPG) and demographics, and used as the inputs of a support vector machine (SVM) classifier. The SVM was trained on 735 patients (368 non-OSA and 567 OSA) and tested on 121 patients (44-77 split). Classification on the test set had an accuracy of up to 92.3%. The signal processing and machine learning algorithms were ported to Java and integrated into the phone app. The app records the audio, actigraphy and PPG signals, implements the clinically validated STOP-BANG questionnaire, derives features from the signals, and finally classifies the patient as needing treatment or not using the trained SVM. The resulting software could provide a new, easy-to-use, low-cost and widely available modality for OSA screening.

A completely automated algorithm to detect poor-quality electrocardiograms (ECGs) is described. The algorithm is based on both novel and previously published signal quality metrics, originally designed for intensive care monitoring. The algorithms have been adapted for use on short (5-10s) single- and multi-lead ECGs. The metrics quantify spectral energy distribution, higher order moments and inter-channel and inter-algorithm agreement. Seven metrics were calculated for each channel (84features in all) and presented to either a multi-layer perceptron artificial neural network or a support vector machine (SVM) for training on a multiple-annotator labelled and adjudicated training dataset. A single-lead version of the algorithm was also developed in a similar manner. Data were drawn from the PhysioNet Challenge 2011 dataset where binary labels were available, on 1500 12-lead ECGs indicating whether the entire recording was acceptable or unacceptable for clinical interpretation. We re-annotated all the leads in both the training set (1000 labelled ECGs) and test dataset (500 12-lead ECGs where labels were not publicly available) using two independent annotators, and a third for adjudication of differences. We found that low-quality data accounted for only 16% of the ECG leads. To balance the classes (between high and low quality), we created extra noisy data samples by adding noise from PhysioNets noise stress test database to some of the clean 12-lead ECGs. No data were shared between training and test sets. A classification accuracy of 98% on the training data and 97% on the test data were achieved. Upon inspection, incorrectly classified data were found to be borderline cases which could be classified either way. If these cases were more consistently labelled, we expect our approach to achieve an accuracy closer to 100%.

We describe a framework for automated electrocardiogram (ECG) quality assessment which works in both normal and arrhythmic situations, on an arbitrary number of ECG leads and for time periods of as short as five seconds. Originally developed for the Physionet/Computing in Cardiology (CinC) Challenge 2011, we present here an extension to our original works with improved quality metrics. We manually annotated the 18000 single lead from the Challenge dataset as well as 9452, 10s segments (of both leads) from every subject in the MIT-BIH arrhythmia database as clinically acceptable or not. To balance the classes, noisy segments from the Noise Stress Test Database were added to clean data segments. A support vector machine was then trained to classify the data as clinically acceptable or not. A 97.1% accuracy was achieved on the test set of the extended database of 10s recordings, dropping almost linearly to 92.4% for 5s recordings. Retraining the classifier on all the challenge data, the classifier gave 93% accuracy on the MIT-BIH arrhythmia database. The results are promising and indicate that our method may be applied to Holter and intensive care unit monitoring.

Diabetes is one of the foremost causes of death in many countries and a leading cause of blindness, renal failure, and non-traumatic amputation. Therefore, diabetic foot ulceration and amputation cause extensive burden on individuals and health care systems in developed and developing countries. Due to the multi-disciplinary requirements for the treatment of diabetic foot ulceration, telemedicine has been introduced to facilitate the access of the patients to specialized health professionals. In this paper the development of an open source mobile health platform is presented, able to support diagnostic algorithms, with the use of a smartphone.

The device has a set of solar collectors (1) fixed or forming an integral part of a cloth (2) of a parasol, under which the collectors cross a post (4), accumulator (3) and an electric outlet (5). An orientation device (6) orients the cloth according to sunshine, and the accumulator and the orientation device are situated in the post. The post includes a thermal insulation such as insulating layer of magnesium carbonate and glass. The post includes an upper part integrating the accumulator and a lower part integrating an inking system (7), where the system is in the form of a corkscrew.