Research Article
Classification on Digital Pathological Images of Breast Cancer Based on Deep Features of Different Levels
Table 10
F1 score and BAC evaluation results at different magnifications (/%).
| Evaluation | Strategy | Magnifications | 40× | 100× | 200× | 400× |
| F1 score | Sliding | 88.61 ± 1.64 | 88.69 ± 0.63 | 90.30 ± 2.15 | 88.07 ± 3.03 | Random | 87.38 ± 1.95 | 89.04 ± 0.96 | 91.13 ± 3.01 | 87.28 ± 2.40 | Sliding + Random | 89.31 ± 2.23 | 90.07 ± 0.82 | 90.97 ± 2.50 | 87.76 ± 1.45 | Sliding + Class Balance Random | 89.99 ± 1.86 | 91.93 ± 0.71 | 92.03 ± 2.77 | 89.41 ± 4.31 | BAC | Sliding | 83.58 ± 2.04 | 83.75 ± 2.02 | 86.47 ± 4.37 | 82.55 ± 5.24 | Random | 81.45 ± 2.45 | 83.92 ± 3.56 | 86.99 ± 4.48 | 81.14 ± 4.33 | Sliding + Random | 84.73 ± 3.00 | 85.86 ± 2.24 | 87.05 ± 4.50 | 82.07 ± 3.43 | Sliding + Class Balance Random | 85.49 ± 2.07 | 86.02 ± 3.31 | 88.00 ± 5.66 | 83.75 ± 6.28 |
|
|