This article discusses the use of a multi-task deep learning model to detect multiple sclerosis (MS) and classify left ventricular ejection fraction (LVEF) using 12-lead ECG scan images. The results show that the multi-task model outperformed single-task models in both tasks and was effective in detecting MS and LVEF <50% in both old and new ECG formats. The model also achieved high accuracy in predicting MS in a prevalence-matched population.
