Table 4.

Fivefold cross-validation classification metrics compared between our method and baseline segmentations. Mean values and standard deviations across folds are presented. The threshold is chosen to achieve a sensitivity closest to 0.9

Metric/Segmentation	Our Approach	Baseline
ROC AUC	0.763 ± 0.056	0.710 ± 0.060
Accuracy	0.563 ± 0.059	0.512 ± 0.036
Sensitivity	0.901 ± 0.076	0.901 ± 0.052
Specificity	0.323 ± 0.075	0.234 ± 0.023
PPV${}_3$	0.490 ± 0.067	0.459 ± 0.049