Table 2.

Performance comparison with state-of-the-art methods on TN3K dataset.

Methods	Dice (%)	IoU (%)	Precision (%)	Recall (%)
U-Net	81.5 ± 0.5	68.8 ± 0.6	82.1 ± 0.7	80.9 ± 0.6
U-Net +  +	82.8 ± 0.6	70.7 ± 0.5	83.5 ± 0.5	82.2 ± 0.7
Attention U-Net	83.7 ± 0.4	72.0 ± 0.6	84.8 ± 0.6	82.7 ± 0.5
TransUNet	83.4 ± 0.7	71.5 ± 0.8	87.8 ± 0.5	79.8 ± 0.9
Swin-UNet	85.3 ± 0.5	74.4 ± 0.4	86.2 ± 0.6	84.5 ± 0.5
UNeXt	83.2 ± 0.6	71.2 ± 0.7	84.1 ± 0.6	82.4 ± 0.7
UNETR +  +	84.9 ± 0.6	73.8 ± 0.5	86.7 ± 0.5	83.4 ± 0.6
MedNeXt	86.1 ± 0.4	75.7 ± 0.5	86.9 ± 0.5	85.4 ± 0.4
USEANet	86.2 ± 0.5	75.9 ± 0.5	86.9 ± 0.6	85.7 ± 0.5
Wave-GMS	86.4 ± 0.4	76.2 ± 0.5	87.0 ± 0.5	85.9 ± 0.5
CEA-Net (Ours)	86.9 ± 0.4	77.0 ± 0.5	87.5 ± 0.6	86.4 ± 0.4

Values are reported as mean ± standard deviation over five-fold cross-validation.