Research Article
Classification on Digital Pathological Images of Breast Cancer Based on Deep Features of Different Levels
Table 15
Extracted features from different feature extraction nodes are input into the image-level BAC evaluation results in the ML model (/%).
| Point location | Classifier | Magnifications | 40× | 100× | 200× | 400× |
| 1 | SVM | 80.07 ± 4.28 | 80.90 ± 3.45 | 81.77 ± 3.31 | 80.24 ± 3.98 | LR | 76.80 ± 3.66 | 79.74 ± 4.90 | 80.44 ± 4.88 | 78.75 ± 4.74 | GNB | 51.83 ± 8.18 | 69.25 ± 4.68 | 82.54 ± 4.03 | 69.74 ± 5.36 | DT | 77.64 ± 3.93 | 78.07 ± 5.01 | 82.27 ± 2.84 | 79.47 ± 3.78 | RF | 79.20 ± 3.99 | 78.91 ± 4.52 | 82.65 ± 2.23 | 80.39 ± 3.95 | MFNN | 77.64 ± 3.86 | 78.98 ± 5.46 | 80.18 ± 5.30 | 77.52 ± 4.37 | 2 | SVM | 86.57 ± 4.06 | 87.64 ± 3.74 | 89.42 ± 3.31 | 85.31 ± 5.10 | LR | 82.68 ± 4.60 | 84.05 ± 3.79 | 86.82 ± 3.46 | 79.05 ± 3.46 | GNB | 73.84 ± 6.93 | 82.81 ± 2.31 | 88.62 ± 2.42 | 80.74 ± 5.36 | DT | 80.87 ± 5.54 | 79.50 ± 4.10 | 83.68 ± 3.51 | 79.63 ± 3.88 | RF | 80.40 ± 4.40 | 80.67 ± 4.80 | 84.04 ± 2.97 | 80.81 ± 4.38 | MFNN | 80.33 ± 4.56 | 81.54 ± 5.39 | 87.16 ± 5.02 | 80.93 ± 3.75 | 3 | SVM | 83.21 ± 1.49 | 84.74 ± 3.22 | 87.88 ± 4.76 | 83.88 ± 5.56 | LR | 81.49 ± 1.73 | 82.80 ± 2.93 | 85.75 ± 4.96 | 80.37 ± 5.46 | GNB | 76.45 ± 5.73 | 87.68 ± 1.61 | 87.54 ± 2.45 | 79.95 ± 7.65 | DT | 79.66 ± 3.51 | 78.85 ± 3.60 | 82.76 ± 3.71 | 79.91 ± 5.17 | RF | 79.65 ± 3.97 | 79.27 ± 3.16 | 83.29 ± 3.15 | 80.52 ± 4.96 | MFNN | 81.97 ± 2.03 | 83.29 ± 3.11 | 86.94 ± 5.15 | 82.11 ± 5.14 |
|
|