Table 5 Performance comparison of ViT-based models with different classifiers and unseen testing data.

Model	Accuracy	Precision		Recall		F1-Score
Model	Accuracy	Tumor	No Tumor	Tumor	No Tumor	Tumor	No Tumor
ViT + RF (Trained	0.8708	0.7973	0.9925	0.9943	0.7473	0.8850	0.8526
on CLAHE Images)	0.8708	0.7973	0.9925	0.9943	0.7473	0.8850	0.8526
ViT + SVM (Trained	0.8373	0.8032	0.8715	0.8845	0.7832	0.8419	0.8250
on CLAHE Images)	0.8373	0.8032	0.8715	0.8845	0.7832	0.8419	0.8250
ViT + SVM (Trained	0.4460	0.2077	0.4703	0.0383	0.8538	0.0647	0.6065
on Contour-
Region Masked
Image)
ViT + RF (Trained	0.4798	0.4799	0.4798	0.4823	0.4773	0.4811	0.4785
on HOG Images)	0.4798	0.4799	0.4798	0.4823	0.4773	0.4811	0.4785
ViT + SVM (Trained	0.6048	0.5664	0.7485	0.8939	0.3157	0.6934	0.4440
on HOG Images)	0.6048	0.5664	0.7485	0.8939	0.3157	0.6934	0.4440

Quick links

Search