Will Two Do? Varying Dimensions in Electrocardiography: The PhysioNet/Computing in Cardiology Challenge 2021
Matthew A. Reyna, Nadi Sadr, Erick Andres Perez Alday, Annie Gu, Amit Shah, Chad Robichaux, Ali Bahrami Rad, Andoni Elola, Salman Seyedi, Sardar Ansari, Hamid Ghanbari, Qiao Li, Ashish Sharma, Gari D. Clifford
Abstract
The PhysioNet/Computing in Cardiology Challenge 2021 focused on the identification of cardiac abnormalities from electrocardiograms (ECGs) and assessed the diagnostic potential of reduced-lead ECGs relative to the standard but less-accessible twelve-lead ECG. We sourced 131,155 recordings with clinical diagnoses from seven institutions in four countries, sharing 88,253 annotated recordings publicly and withholding the remaining recordings for validation and testing. We asked the Challenge participants to design working, open-source algorithms for identifying cardiac abnormalities from twelve-lead, six-lead, four-lead, three-lead, and two-lead ECG recordings. By sourcing data from diverse populations, requiring the submission of reusable training code, and designing an evaluation metric specifically for this task, we encouraged the development of generalizable, reproducible, and clinically relevant algorithms for identifying cardiac abnormalities from ECGs. A total of 68 teams submitted a total of 1056 algorithms during the Challenge. Of these, 39 teams were ultimately successful, representing a diversity of approaches from both academia and industry.