The George B. Moody PhysioNet Challenge 2024 invites teams to develop algorithmic approaches for digitizing and classifying electrocardiograms (ECGs) captured from images or scanned paper printouts.
Digital ECG-based devices and methods have the potential to improve access to ECG-based diagnoses and cardiac care, but physical ECG representations have been part of cardiac care for nearly a century, and they remain common in much of the world, particularly in the Global South. There are likely billions of paper ECGs in existence globally with millions more collected daily, especially in underrepresented and underserved populations. These physical ECGs capture the variability and evolution of CVDs across demographics, geography, and time. However, they do not inform or benefit from algorithms that identify cardiac abnormalities in ECGs because such algorithms typically require digital time-series representations of ECG data. Therefore, accessible and high-quality digitization of ECG images, including vast ECG archives that are only available as hardcopies, is critical for improving the accessibility and quality of cardiac care, and leveraging the vast ECG archives.
For the 2024 Challenge, we ask participants to design and implement working, open-source algorithms that reconstruct the time-series in the ECG images and/or classify them. The teams with the best scores on the hidden test set will win the Challenge.
This Challenge provides multiple innovations. First, we developed code for generating synthetic ECG images with various realistic distortions, such as wrinkles, creases, shadows, rotations, and handwriting, to allow teams to create arbitrary large and diverse training sets for their models. Second, we evaluate the models using traditional signal quality metrics and classification metrics as well as clinical measurement-aware metrics to help us better understand the important connection between ECG signals and diagnoses. Many teams have participated in the Challenge so far, representing diverse approaches from both academia and industry worldwide.