Table 1.

Evaluation of patch classification accuracy.

Model Name	LUAD	BRCA	SARC	OV	Other*	13 cancer types**	All
Baseline	73.60%	74.90%	–	–	–	79.56%	–
VGG-16	83.28%	88.38%	94.17%	88.29%	82.52%	83.32%	86.02%
ResNet-34	84.28%	86.24%	91.41%	87.29%	82.10%	82.45%	85.14%
Incep-V4	86.29%	87.16%	96.93%	94.31%	82.53%	83.68%	87.43%

Compare result for each of LUAD, BRCA, SARC, OV, *Other: patches from other cancer types in the set of 23 types used in training, **13 cancer types: subset of test patches belonging to the 13 cancer types the baseline model with human in the loop (Baseline) (33) was trained on, All: all test patches from all the 23 cancer types. Best accuracy in each dataset is indicated in bold.