Fig. 4

The performance across different medical subdomains in the baseline and the best interpretable classifiers on iDASH and MGH datasets. All measurements, including precision, recall, F1 score, balanced accuracy, and AUC were compared in the a baseline (white) and the best (black) iDASH classifiers, and the b baseline (white) and the best (black) MGH classifiers. Significantly improved performance is observed in the best classifier, especially in difficult to separate medical subdomains, such as ‘Anesthesiology’, “Pulmonary disease”, “Intensive care” and “Infectious diseases”