Table 3. Comprehensive comparison of metrics for different methods on the MSDPT dataset.

Model	Dice↑	Iou↑	Sens↑	Spec↑	HD95↓
3DUNet	0.4359 ± 0.0527	0.3223 ± 0.0393	0.4371 ± 0.0766	0.9997 ± 0.0001	38.5963 ± 18.2940
SegResNet	0.4732 ± 0.0572	0.3536 ± 0.0492	0.4736 ± 0.0812	0.9997 ± 0.0001	37.6231 ± 15.6751
3DUX-Net	0.4781 ± 0.0683	0.3554 ± 0.0566	0.4922 ± 0.0916	0.9997 ± 0.0001	36.3027 ± 22.6361
AttentionUnet	0.4839 ± 0.0853	0.3608 ± 0.0703	0.5124 ± 0.1399	0.9997 ± 0.0001	36.8667 ± 18.2556
UNETR	0.4017 ± 0.0547	0.2883 ± 0.0449	0.4025 ± 0.0854	0.9996 ± 0.0001	43.8937 ± 17.7325
nnFormer	0.4666 ± 0.1123	0.3513 ± 0.0942	0.4628 ± 0.1328	0.9997 ± 0.0001	45.0921 ± 22.3324
SwinUNETR	0.4781 ± 0.0663	0.3556 ± 0.0546	0.5055 ± 0.0902	0.9996 ± 0.0002	45.6697 ± 23.5726
Ours	0.5210 ± 0.0651	0.4174 ± 0.0184	0.5521 ± 0.0836	0.9997 ± 0.0001	28.3127 ± 5.6131

Note:

Bold values indicate the optimal value of each evaluation index.