Histopathological Tissue Segmentation of Lung Cancer with Bilinear CNN and Soft Attention

Table 2

The classification F1 score of tissue types on lung cancer dataset.

Model	TUM	LYM	STR	NOR	VES	BRO	NEC	APC	BAC	OTH
ResNet50 [29]	0.9686	0.9914	0.8687	0.8532	0.8512	0.9615	0.9668	0.9962	0.9959	0.7734
VGG19 [30]	0.9639	0.9913	0.8720	0.8354	0.8821	0.9545	0.9628	0.9939	0.9944	0.7219
EfficientNet [31]	0.9514	0.9789	0.8854	0.8414	0.8768	0.9533	0.9689	0.9955	0.9957	0.7326
DeepTissue Net [14]	0.9331	0.9640	0.8669	0.8635	0.8973	0.9542	0.9589	0.9967	0.9915	0.7852
ResNet50+bilinear pooling module	0.9698	0.9911	0.8753	0.8658	0.8736	0.9538	0.9638	0.9969	0.9936	0.7857
ResNet50+attention module	0.9712	0.9916	0.8862	0.8732	0.8954	0.9582	0.9615	0.9972	0.9931	0.7516
ResNet50+bilinear pooling module+attention module	0.9739	0.9911	0.9056	0.8788	0.9186	0.9586	0.9766	0.9952	0.9935	0.8025