Table 7 Results on the Camelyon+ dataset with CONCH-V1.5-extracted features.
From: Comprehensive Benchmark Dataset for Pathological Lymph Node Metastasis in Breast Cancer Sections
Methods | Acc (%) | AUC (%) | F1 (%) | Recall (%) | Precision (%) | Kappa |
|---|---|---|---|---|---|---|
CONCH-V1.57 (WSIs pre-trained) | ||||||
Max-MIL | 84.7 ± 2.50 | 87.6 ± 3.29 | 57.4 ± 2.97 | 57.2 ± 3.24 | 59.1 ± 4.59 | 0.67 ± 0.05 |
Mean-MIL | 79.6 ± 2.11 | 80.4 ± 3.53 | 47.7 ± 2.65 | 47.4 ± 2.81 | 50.6 ± 5.34 | 0.62 ± 0.03 |
ABMIL19 | 84.3 ± 2.43 | 89.3 ± 2.28 | 57.5 ± 4.38 | 57.1 ± 6.17 | 63.7 ± 6.84 | 0.68 ± 0.03 |
Gate-ABMIL19 | 84.4 ± 2.12 | 89.5 ± 2.70 | 57.2 ± 2.55 | 56.1 ± 3.34 | 58.1 ± 4.10 | 0.69 ± 0.01 |
CLAM-SB25 | 86.2 ± 1.95 | 89.1 ± 3.57 | 60.8 ± 2.57 | 60.8 ± 3.82 | 63.6 ± 7.13 | 0.70 ± 0.04 |
CLAM-MB25 | 85.7 ± 3.15 | 89.5 ± 2.99 | 63.7 ± 4.21 | 63.1 ± 3.29 | 63.4 ± 2.38 | 0.65 ± 0.11 |
DSMIL34 | 84.0 ± 3.29 | 89.0 ± 2.72 | 61.0 ± 2.05 | 60.3 ± 2.32 | 62.1 ± 5.04 | 0.66 ± 0.12 |
TransMIL24 | 80.9 ± 4.45 | 90.4 ± 1.89 | 59.0 ± 5.94 | 56.4 ± 5.58 | 61.4 ± 4.61 | 0.61 ± 0.15 |
DTFD35 | 82.3 ± 2.18 | 86.9 ± 2.96 | 55.7 ± 4.51 | 53.2 ± 5.49 | 54.1 ± 5.53 | 0.66 ± 0.06 |
AMD-MIL33 | 85.8 ± 2.93 | 90.7 ± 2.55 | 63.9 ± 6.76 | 63.0 ± 7.81 | 65.5 ± 7.06 | 0.68 ± 0.07 |
WiKG36 | 84.8 ± 2.84 | 89.8 ± 1.89 | 59.7 ± 4.94 | 59.0 ± 6.45 | 61.2 ± 6.67 | 0.66 ± 0.09 |
FR-MIL37 | 84.9 ± 1.15 | 90.1 ± 1.50 | 59.0 ± 2.02 | 59.2 ± 2.69 | 61.1 ± 3.43 | 0.69 ± 0.03 |
