Table 10.
Comparison of classification accuracy (%) of different classification algorithms for BreaKHis dataset.
| Category | Model | Magnification | |||
|---|---|---|---|---|---|
| 40X | 100X | 200X | 400X | ||
| Eight-class Classification | GoogLeNet (23) | 68.7 | 65.9 | 69.1 | 62.8 |
| ResNet50 (23) | 82.5 | 78.8 | 84.3 | 81 | |
| Inception-ResNet-V2 (23) | 86.7 | 80.3 | 83.5 | 68.5 | |
| CNN (28) | 88.2 | 84.6 | 83.3 | 84 | |
| CNN (28) | 82.70 | 82.15 | 83.37 | 82.40 | |
| PFTAS + SVM (29) | 81.65 | 79.70 | 85.30 | 82.30 | |
| PFTAS + RF (29) | 81.70 | 82.60 | 84.40 | 81.20 | |
| Single-Task CNN (29) | 83.08 | 84.15 | 85.67 | 83.10 | |
| Proposed method | 93.88 | 93.97 | 94.57 | 94.77 | |
| Binary Classification | CNN (30) | 89.6 | 85 | 84 | 80.8 |
| DeCAF features using CNN (31) | 84.6 | 84.8 | 84.2 | 81.6 | |
| Single Task CNN (5) | 83 | 83.1 | 84.6 | 82.1 | |
| Proposed method | 97.68 | 97.98 | 97.88 | 97.79 | |
The bold values indicate the highest accuracy under the same conditions.