The subject of the PhysioNet/Computing in Cardiology Challenge 2020 was the identification of cardiac abnormalities in 12-lead electrocardiogram (ECG) recordings. A total of 66,405 recordings were sourced from hospital systems from four distinct countries and annotated with clinical diagnoses, including 43,101 annotated recordings that were posted publicly.For this Challenge, we asked participants to design working, open-source algorithms for identifying cardiac abnormalities in 12-lead ECG recordings. This Challenge provided several innovations. First, we sourced data from multiple institutions from around the world with different demographics, allowing us to assess the generalizability of the algorithms. Second, we required participants to submit both their trained models and the code for reproducing their trained models from the training data, which aids the generalizability and reproducibility of the algorithms. Third, we proposed a novel evaluation metric that considers different misclassification errors for different cardiac abnormalities, reflecting the clinical reality that some diagnoses have similar outcomes and varying risks.Over 200 teams submitted 850 algorithms (432 of which successfully ran) during the unofficial and official phases of the Challenge, representing a diversity of approaches from both academia and industry for identifying cardiac abnormalities. The official phase of the Challenge is ongoing.

Cardiac auscultation is one of the most costeffective techniques used to detect and identify many heart conditions. Computer-assisted decision systems based on auscultation can support physicians in their decisions. Unfortunately, the application of such systems in clinical trials is still minimal since most of them only aim to detect the presence of extra or abnormal waves in the phonocardiogram signal, i.e., only a binary ground truth variable (normal vs abnormal) is provided. This is mainly due to the lack of large publicly available datasets, where a more detailed description of such abnormal waves (e.g., cardiac murmurs) exists.To pave the way to more effective research on healthcare recommendation systems based on auscultation, our team has prepared the currently largest pediatric heart sound dataset. A total of 5282 recordings have been collected from the four main auscultation locations of 1568 patients, in the process, 215780 heart sounds have been manually annotated. Furthermore, and for the first time, each cardiac murmur has been manually annotated by an expert annotator according to its timing, shape, pitch, grading, and quality. In addition, the auscultation locations where the murmur is present were identified as well as the auscultation location where the murmur is detected more intensively. Such detailed description for a relatively large number of heart sounds may pave the way for new machine learning algorithms with a real-world application for the detection and analysis of murmur waves for diagnostic purposes.

No abstract

Objective: Cardiac auscultation is an accessible diagnostic screening tool that can help to identify patients with heart murmurs for follow-up diagnostic screening and treatment, especially in resource-constrained environments. However, experts are needed to interpret the heart sound recordings, limiting the accessibility of auscultation for cardiac care. The George B. Moody PhysioNet Challenge 2022 invites teams to develop automated approaches for detecting abnormal heart function from multi-location phonocardiogram (PCG) recordings of heart sounds. Approach: For the Challenge, we sourced 5272 PCG recordings from 1568 pediatric patients in rural Brazil. We required the Challenge participants to submit the complete code for training and running their models, improving the transparency, reproducibility, and utility of the diagnostic algorithms. We devised a cost-based evaluation metric that captures the costs of screening, treatment, and diagnostic errors, allowing us to investigate the benefits of algorithmic pre-screening and facilitate the development of more clinically relevant algorithms. Main results: So far, over 80 teams have submitted over 600 algorithms during the course of the Challenge, representing a diversity of approaches in academia and industry. We will update this manuscript to share an analysis of the Challenge after the end of the Challenge. Significance: The use of heart sound recordings for both heart murmur detection and clinical outcome identification allowed us to explore the potential of automated approaches to provide accessible pre-screening of less-resourced populations. The submission of working, open-source algorithms and the use of novel evaluation metrics supported the reproducibility, generalizability, and relevance of the researched conducted during the Challenge.

The automatic detection of pulse during out-of-hospital cardiac arrest (OHCA) is necessary for the early recognition of the arrest and the detection of return of spontaneous circulation (end of the arrest). The only signal available in every single defibrillator and valid for the detection of pulse is the electrocardiogram (ECG). In this study we propose two deep neural network (DNN) architectures to detect pulse using short ECG segments (5 s), i.e., to classify the rhythm into pulseless electrical activity (PEA) or pulse-generating rhythm (PR). A total of 3914 5-s ECG segments, 2372 PR and 1542 PEA, were extracted from 279 OHCA episodes. Data were partitioned patient-wise into training (80%) and test (20%) sets. The first DNN architecture was a fully convolutional neural network, and the second architecture added a recurrent layer to learn temporal dependencies. Both DNN architectures were tuned using Bayesian optimization, and the results for the test set were compared to state-of-the art PR/PEA discrimination algorithms based on machine learning and hand crafted features. The PR/PEA classifiers were evaluated in terms of sensitivity (Se) for PR, specificity (Sp) for PEA, and the balanced accuracy (BAC), the average of Se and Sp. The Se/Sp/BAC of the DNN architectures were 94.1%/92.9%/93.5% for the first one, and 95.5%/91.6%/93.5% for the second one. Both architectures improved the performance of state of the art methods by more than 1.5 points in BAC.

The subject of the PhysioNet/Computing in Cardiology Challenge 2020 was the identification of cardiac abnormalities in 12-lead electrocardiogram (ECG) recordings. A total of 66,405 recordings were sourced from hospital systems from four distinct countries and annotated with clinical diagnoses, including 43,101 annotated recordings that were posted publicly. For this Challenge, we asked participants to design working, open-source algorithms for identifying cardiac abnormalities in 12-lead ECG recordings. This Challenge provided several innovations. First, we sourced data from multiple institutions from around the world with different demographics, allowing us to assess the generalizability of the algorithms. Second, we required participants to submit both their trained models and the code for reproducing their trained models from the training data, which aids the generalizability and reproducibility of the algorithms. Third, we proposed a novel evaluation metric that considers different misclassification errors for different cardiac abnormalities, reflecting the clinical reality that some diagnoses have similar outcomes and varying risks. Over 200 teams submitted 850 algorithms (432 of which successfully ran) during the unofficial and official phases of the Challenge, representing a diversity of approaches from both academia and industry for identifying cardiac abnormalities. The official phase of the Challenge is ongoing.

Accurate shock decision methods during 1 piston-driven cardiopulmonary resuscitation (CPR) would 2 contribute to improve therapy and increase cardiac arrest 3 survival rates. The best current methods are computationally 4 demanding, and their accuracy could be improved. The objective 5 of this work was to introduce a computationally efficient 6 algorithm for shock decision during piston-driven CPR with 7 increased accuracy. Methods: The study dataset contains 201 8 shockable and 844 nonshockable ECG segments from 230 cardiac 9 arrest patients treated with the LUCAS-2 mechanical CPR 10 device. Compression artifacts were removed using state of the 11 art adaptive filters, and shock/no-shock discrimination features 12 were extracted from the stationary wavelet transform analysis of 13 the filtered ECG, and fed to a support vector machine (SVM) 14 classifier. Quasi-stratified patient wise nested cross-validation was 15 used for feature selection and SVM hyperparameter optimization. 16 The procedure was repeated 50 times to statistically characterize 17 the results. Results: Best results were obtained for a 6 feature 18 classifier with mean (standard deviation) sensitivity, specificity, 19 and total accuracy of 97.5 (0.4), 98.2 (0.4) and 98.1 (0.3), 20 respectively. The algorithm presented a five-fold reduction in 21 computational demands when compared to the best available 22 methods, while improving their balanced accuracy by 3-points.23Conclusions: The accuracy of the best available methods was 24 improved while drastically reducing the computational demands.

Sudden cardiac arrest is one of the leading causes of death in the industrialized world. Pulse detection is essential for the recognition of the arrest and the recognition of return of spontaneous circulation during therapy, and it is therefore crucial for the survival of the patient. This paper introduces the first method based exclusively on the ECG for the automatic detection of pulse during cardiopulmonary resuscitation. Random forest classifier is used to efficiently combine up to nine features from the time, frequency, slope, and regularity analysis of the ECG. Data from 191 cardiac arrest patients was used, and 1177 ECG segments were processed, 796 with pulse and 381 without pulse. A leave-one-patient out cross validation approach was used to train and test the algorithm. The statistical distributions of sensitivity (SE) and specificity (SP) for pulse detection were estimated using 500 patient-wise bootstrap partitions. The mean (std) SE/SP for nine-feature classifier was 88.4 (1.8) %/89.7 (1.4) %, respectively. The designed algorithm only requires 4-s-long ECG segments and could be integrated in any commercial automated external defibrillator. The method permits to detect the presence of pulse accurately, minimizing interruptions in cardiopulmonary resuscitation therapy, and could contribute to improve survival from cardiac arrest.