Table 3 Model performance comparison. Shows the performance differences on GUSI-A between the baseline U-Net and modified models. Models combining affine transformations and ASTM exhibit significant improvements in IoU and Dice scores.
Model | Accuracy | Precision | Recall | IoU | Dice | Specificity |
|---|---|---|---|---|---|---|
Baseline U-Net | 96.64±0.20 | 86.11±2.40 | 85.96±2.72 | 75.40±1.26 | 85.97±0.82 | 98.09±0.44 |
Without affine Transform | 96.58±0.63 | 86.71±4.10 | 84.74±4.14 | 74.91±3.92 | 85.60±2.58 | 98.20±0.63 |
Without ASTM | 96.79±0.31 | 86.70±5.24 | 87.42±5.38 | 76.59±1.23 | 86.74±0.78 | 98.07±1.02 |
Ours | 97.39 ± 0.12 | 89.26 ± 0.72 | 89.63 ± 0.42 | 80.87 ± 0.76 | 89.39 ± 0.47 | 98.48 ± 0.11 |
