Cervical Lesion Classification Method Based on Cross-Validation Decision Fusion Method of Vision Transformer and DenseNet

Table 5

Comparison fusion results of DenseNet161 with EfficientNetV2 and Vision Transformer.

Model	Classes	F1	PPV	SEN	SPEC	NPV
DenseNet161 + EfficientNetV2	Cancer	0.900	0.900	0.900	0.989	0.989
	HSIL	0.682	0.667	0.698	0.737	0.764
	LSIL	0.637	0.667	0.609	0.909	0.886
	Negative	0.625	0.625	0.625	0.882	0.882
DenseNet161 + Vision Transformer	Cancer	0.900	0.900	0.900	0.989	0.989
	HSIL	0.689	0.660	0.721	0.719	0.774
	LSIL	0.605	0.650	0.565	0.909	0.875
	Negative	0.638	0.652	0.625	0.895	0.883