Table 3.

Binary classification of diseases based on all the brain structures.

Case	V-Net		UNETR		FS
	LR	XGBoost	LR	XGBoost	LR	XGBoost
Normal vs. PSP	$\mathbf{0.89}\pm \mathbf{0.07}$	$0.86\pm 0.05$	$0.89\pm 0.08$	$0.84\pm 0.04$	$\mathbf{0.89}\pm \mathbf{0.07}$	$0.87\pm 0.06$
Normal vs. MSA-P	$0.78\pm 0.04$	${0.73\pm 0.003}^{\ast }$	$\mathbf{0.81}\pm \mathbf{0.03}$	$0.77\pm 0.04$	$0.79\pm 0.001$	$\mathbf{0.82}\pm \mathbf{0.01}$
Normal vs. MSA-C	$0.90\pm 0.03$	$\mathbf{0.93}\pm \mathbf{0.04}$	$0.85\pm 0.12$	$0.90\pm 0.10$	$0.88\pm 0.04$	$\mathbf{0.95}\pm \mathbf{0.03}$
Normal vs. PD	$0.60\pm 0.07$	${\mathbf{0.66}\pm \mathbf{0.02}}^{\ast }$	$0.60\pm 0.04$	$0.60\pm 0.03$	$0.65\pm 0.07$	$\mathbf{0.70}\pm \mathbf{0.05}$
PD vs. PSP	$\mathbf{0.80}\pm \mathbf{0.08}$	${0.78\pm 0.001}^{\ast }$	$0.77\pm 0.13$	$0.75\pm 0.01$	$\mathbf{0.77}\pm \mathbf{0.10}$	$0.76\pm 0.03$
PD vs. MSA-P	$\mathbf{0.76}\pm \mathbf{0.07}$	$0.66\pm 0.03$	${0.71\pm 0.02}^{\ast }$	${0.68\pm 0.07}^{\ast }$	$\mathbf{0.79}\pm \mathbf{0.08}$	$0.71\pm 0.02$
PD vs. MSA-C	$\mathbf{0.91}\pm \mathbf{0.04}$	$0.87\pm 0.03$	$0.80\pm 0.19$	$0.80\pm 0.12$	$0.89\pm 0.07$	$\mathbf{0.91}\pm \mathbf{0.05}$

AUC in LR and XGBoost of CNN-based V-Net, ViT-based UNETR, and FS. The AUC is expressed as the mean from threefold cross-validation. LR; logistic regression, XGBoost; eXtreme Gradient Boosting.

The best result for each volume segmentation method based on FS and DL in each binary classification is shown in bold.

*$p<0.05$ indicates a significant difference in AUC between the DL models and FS.