Fig. 3

Participant selection and workflow for deep learning model development. The flowchart illustrates the process of collecting and analyzing heart-sound data for the assessment of the severity of mitral regurgitation in participants. Of note, the severe group initially included 167 cases; however, later in the study, two cases previously classified as late-stage were merged into the severe group owing to their small sample size. The recordings were then segmented into training, test, and validation datasets for model development. MINE score, mitral insufficiency echocardiographic score