Table 3 Outcomes of ablation studies on the BUSI dataset.
From: A hybrid attention network for accurate breast tumor segmentation in ultrasound images
Method | Parameters | Jaccard | Dice | Sensitivity | Accuracy | Precision | Specificity |
|---|---|---|---|---|---|---|---|
U-Net15 | 34,513,345 | 76.54 | 83.13 | 82.83 | 97.91 | 83.94 | 98.81 |
Baseline (BL) | 10,498,945 | 90.44 | 94.37 | 94.33 | 99.46 | 94.43 | 99.69 |
BL + ConvBlock | 11,089,537 | 92.89 | 96.27 | 96.33 | 99.60 | 96.44 | 99.77 |
BL + ConvBlock + SEFB | 15,339,329 | 93.51 | 96.80 | 96.58 | 99.69 | 96.66 | 99.82 |
BL + ConvBlock + SEFB + TAM | 15,427,713 | 94.75 | 97.28 | 97.15 | 99.74 | 97.42 | 99.84 |
