Table 4 Results of comparison experiments on the BUSI dataset.
From: Multi-branch CNN and grouping cascade attention for medical image classification
Method(year) | Params (M) | FLOPs (G) | Acc | F1 | Precision | Recall | Auc |
|---|---|---|---|---|---|---|---|
ResNet50 (2016) | 23.5 | 4.1 | 0.9000 | 0.8769 | 0.9164 | 0.8492 | 0.8909 |
MobileNetV2 (2018) | 2.2 | 0.3 | 0.8667 | 0.8404 | 0.8849 | 0.8115 | 0.8598 |
EfficientNet-B0 (2019) | 4.0 | 0.4 | 0.8800 | 0.8628 | 0.9063 | 0.8353 | 0.8756 |
RepVGG (2021) | 43.7 | 9.9 | 0.9133 | 0.8964 | 0.9102 | 0.8869 | 0.9162 |
ConvNext-S (2022) | 49.5 | 8.7 | 0.7667 | 0.6717 | 0.8503 | 0.6290 | 0.7283 |
ConvMixer (2023) | 47.9 | 49.1 | 0.9067 | 0.8901 | 0.9006 | 0.8810 | 0.9117 |
InceptionNext-S (2023) | 47.1 | 8.4 | 0.9000 | 0.8906 | 0.9066 | 0.8770 | 0.9060 |
FasterNet (2023) | 13.7 | 1.9 | 0.9067 | 0.8907 | 0.9080 | 0.8790 | 0.9122 |
Swin-S (2021) | 48.8 | 8.6 | 0.8400 | 0.8219 | 0.8515 | 0.8095 | 0.8521 |
CrossViT 18 (2021) | 43.3 | 9.0 | 0.8600 | 0.8308 | 0.8679 | 0.8155 | 0.8619 |
MoCoViT 1.0 (2022) | 7.2 | 0.5 | 0.8600 | 0.8373 | 0.8838 | 0.8095 | 0.8563 |
BiFormer-S (2023) | 56.0 | 9.4 | 0.8733 | 0.8493 | 0.8950 | 0.8194 | 0.8653 |
FasterViT-2 (2023) | 75.2 | 8.9 | 0.8968 | 0.9173 | 0.8867 | 0.8736 | 0.8712 |
Flatten-pvt (2023) | 24.2 | 3.7 | 0.8667 | 0.8497 | 0.8811 | 0.8333 | 0.8720 |
TransxNet (2023) | 25.5 | 4.6 | 0.8733 | 0.8576 | 0.8674 | 0.8492 | 0.8861 |
GroupMixFormer (2023) | 22.1 | 5.1 | 0.8933 | 0.8657 | 0.9285 | 0.8294 | 0.8772 |
Eff-CTNet(Ours) | 25.2 | 6.4 | 0.9333 | 0.9261 | 0.9326 | 0.9226 | 0.9404 |
