Table 4.

Cancer detection results, obtained from ROC curves using all histological images’ patch-level statistics. Reported are the AUC for the validation group and the AUC, F1 score, accuracy, sensitivity, and specificity of the testing group with 95% confidence intervals for all values. Also shown in the right-most column is the ability of the proposed method to distinguish slides that contain cancer from slides that are all normal.

Group	Validation AUC	Test AUC	F1 Score	Accuracy	Sensitivity	Specificity	Slide Level AUC
SCC	0.913 (0.90,0.93)	0.916 (0.90, 0.93)	84.8 ± 1.5%	84.8 ± 1.6%	84.7 ± 2.2%	85.0 ± 2.2%	0.944 (0.91, 0.97)
Thyroid	0.927 (0.92, 0.94)	0.954 (0.94, 0.97)	89.4 ± 1.3%	89.4 ± 1.3%	89.6 ± 1.8%	89.1 ± 1.9%	0.995 (0.99, 1.00)
Lymph Node	0.986 (0.96, 0.99)	0.967 (0.96, 0.98)	91.8 ± 1.3%	93.4 ± 1.2%	90.1 ± 1.8%	93.6 ± 1.6%	0.901 (0.86, 0.94)