Fig. 1
From: Weakly supervised classification of aortic valve malformations using unlabeled cardiac MRI sequences
Weak supervision scale up performance metrics. Metrics include a positive predictive value (precision); b sensitivity (recall); c area under the ROC curve (AUROC); and d normalized discounted cumulative gain (NDCG). The y-axis is the score in [0,100] and the x-axis is the number of unlabeled MRIs used for training. The dashed horizontal line indicates the expert-labeled baseline model with augmentations. Shaded regions and gray horizontal lines indicate 95% confidence intervals (where n = the number of unlabeled training MRIs)
