Marinka Oudkerk pool receives Best abstract award with the working group e-cardiology on the European Society of Cardiology conference 2020 for her abstract titled “Distinguishing sinus rhythm from atrial fibrillation on single-lead ECGs using a deep neural network”.

The abstract describes the use of a residual neural network, tailored to single-lead ECG analysis. The network was trained on 8528 available ECGs of the PhysioNet / Computing in Cardiology Challenge of 2017. Testing was performed on a seperate set containing 300 sinus rhythm ECGs and 300 atrial fibrillation ECGs, resulting in an F1-score of 0.91. 

Due to the popularity of smart watches and smart devices, acquisition of single-lead ECGs on these devices is possible for the general population. Up to now, these single-lead ECGs are predominantly used for screening of atrial fibrillation. However, it is possible to extract more information out of these ECGs, including various pathologies. Therefore, in the Cardiac arrhythmia in patients with congenital heart disease project, we develop novel AI algorithms to analyze single-lead ECGs to distinguish multiple disease classes.