Table 3. Patient-level performance evaluation for predicting clinical heart failure or severe tissue pathology from H&E stained whole-slide images for validation folds of the training data set.

Metric	Random Forest	Deep Learning	p-value
Image-level results
Accuracy	0.869 ± 0.05	0.954 ± 0.03	0.05
Sensitivity	0.866 ± 0.07	0.968 ± 0.03	n.s.
Specificity	0.872 ± 0.04	0.943 ± 0.05	n.s.
Positive predictive value	0.848 ± 0.05	0.935 ± 0.05	0.05
AUC	0.944 ± 0.04	0.977 ± 0.02	0.05
Patient-level results
Accuracy	0.923 ± 0.03	0.962 ± 0.02	n.s.
Sensitivity	0.917 ± 0.07	0.979 ± 0.04	n.s.
Specificity	0.930 ± 0.06	0.947 ± 0.05	n.s.
Positive predictive value	0.919 ± 0.07	0.942 ± 0.06	n.s.
AUC	0.963 ± 0.05	0.960 ± 0.05	n.s.

The results are presented as the Mean ± SD of three models. Each model was trained on ~770 images from ~70 patients. These models were evaluated on the validation fold of ~35 patients. The patient-level diagnosis is the majority vote over all the images from a single patient. Statistics were determined by an unpaired two-sample t-test with an N of three folds.