Table 1.

Results obtained training the plain ResNet to perform the cancer type classification.

	Plain ResNet
	5x	10x	20x
Weighted F1	0.86	0.85	0.83
Macro F1	0.79	0.80	0.77
ACC	0.43	0.67	0.64
BLCA	0.74	0.78	0.71
BRCA	0.92	0.91	0.90
CESC	0.82	0.75	0.74
CHOL	0.50	0.55	0.25
COAD	0.86	0.86	0.83
DLBC	0.00	0.00	0.00
ESCA	0.74	0.68	0.61
GBM	0.91	0.91	0.90
HNSC	0.88	0.84	0.80
KICH	0.90	0.97	0.90
KIRC	0.93	0.94	0.90
KIRP	0.79	0.75	0.74
LGG	0.95	0.97	0.94
LIHC	0.89	0.87	0.87
LUAD	0.79	0.77	0.78
LUSC	0.75	0.72	0.70
MESO	0.57	0.67	0.70
OV	0.90	0.91	1.00
PAAD	0.92	0.84	0.82
PCPG	0.96	0.96	0.93
PRAD	0.96	0.94	0.95
SARC	0.80	0.80	0.70
SKCM	0.83	0.87	0.83
STAD	0.84	0.80	0.80
TGCT	0.95	0.97	0.92
THCA	0.95	0.93	0.96
THYM	0.94	0.86	0.83
UCEC	0.88	0.87	0.83
UCS	0.37	0.63	0.72

Rows denote the F1 score for each tumor type individually. The value of the best performing magnification is highlighted in green. The top two rows show the aggregated classification results denoted by the weighted F1 and macro F1 scores for the overall classification across all different tumor types.