Table 5.

Confusion matrix of the classification of superpixels using the optimized Xception network in triple-negative breast cancer patients in six categories: tumor, necrosis, cluster of lymphocytes (Lym), stroma, fat, and lumen/empty space (separate test set of five whole-slide images).

	Tumor	Necrosis	Lym	Stroma	Fat	Empty space
Tumor	1830	13	15	42	0	0
Necrosis	50	1446	2	320	0	0
Lym	4	2	705	10	0	0
Stroma	42	120	20	3836	0	1
Fat	0	0	0	0	562	5
Empty space	0	0	0	0	67	1257

Overall accuracy = 93.1%, average precision = 93.9%, average recall = 93.6%.

The bold values indicate the correct predictions of the network.