Table 10 Results on the Camelyon+ dataset with Gigapath-extracted features.
From: Comprehensive Benchmark Dataset for Pathological Lymph Node Metastasis in Breast Cancer Sections
Methods | Acc (%) | AUC (%) | F1 (%) | Recall (%) | Precision (%) | Kappa |
|---|---|---|---|---|---|---|
Gigapath5 (WSIs pre-trained) | ||||||
Max-MIL | 81.4 ± 6.62 | 86.6 ± 6.96 | 55.4 ± 5.93 | 54.6 ± 5.35 | 54.9 ± 5.53 | 0.65 ± 0.11 |
Mean-MIL | 77.4 ± 3.20 | 79.1 ± 3.17 | 46.3 ± 4.24 | 46.3 ± 4.61 | 53.6 ± 9.24 | 0.53 ± 0.09 |
ABMIL19 | 79.2 ± 2.43 | 92.1 ± 0.27 | 65.7 ± 6.11 | 58.1 ± 5.31 | 58.6 ± 0.99 | 0.68 ± 0.03 |
Gate-ABMIL19 | 84.9 ± 3.00 | 89.7 ± 3.41 | 59.6 ± 1.61 | 58.7 ± 2.70 | 59.5 ± 4.52 | 0.69 ± 0.04 |
CLAM-SB25 | 85.4 ± 2.93 | 91.2 ± 2.47 | 60.8 ± 3.78 | 59.8 ± 3.56 | 63.7 ± 8.90 | 0.72 ± 0.03 |
CLAM-MB25 | 86.7 ± 3.01 | 92.1 ± 2.88 | 69.1 ± 8.13 | 66.7 ± 5.36 | 67.0 ± 3.40 | 0.70 ± 0.07 |
DSMIL34 | 86.5 ± 3.12 | 89.7 ± 4.70 | 62.1 ± 4.48 | 62.3 ± 2.86 | 64.2 ± 2.46 | 0.68 ± 0.06 |
TransMIL24 | 88.2 ± 2.77 | 92.8 ± 2.44 | 67.4 ± 3.54 | 66.9 ± 2.68 | 75.5 ± 11.95 | 0.73 ± 0.07 |
DTFD35 | 83.7 ± 2.33 | 89.4 ± 4.64 | 59.4 ± 3.78 | 58.0 ± 4.24 | 61.3 ± 3.57 | 0.68 ± 0.07 |
AMD-MIL33 | 87.3 ± 1.92 | 92.6 ± 3.11 | 65.8 ± 4.68 | 64.6 ± 4.28 | 64.6 ± 4.42 | 0.72 ± 0.05 |
WiKG36 | 85.0 ± 3.89 | 91.3 ± 2.93 | 59.5 ± 5.13 | 59.4 ± 6.88 | 68.0 ± 16.12 | 0.72 ± 0.06 |
FR-MIL37 | 86.2 ± 2.01 | 90.7 ± 3.11 | 63.2 ± 3.54 | 63.8 ± 2.40 | 75.7 ± 10.20 | 0.71 ± 0.06 |
