Fig. 3

Classification Accuracy and Receiver Operating Characteristic (ROC) Curve. (a) Classification accuracy for the best performing model obtained via exhaustive 1-holdout cross-validation on training data. 87% of patients were correctly classified. (b) Classification accuracy of the best performing model on the validation data. 72% of patients were correctly classified. (c) The corresponding ROC curve shows true and false positive rates for different thresholds of a positive diagnosis based on the highest snippet score. The area under the curve is 0.75