Table 8 Results on the Camelyon+ dataset with Ctranspath-extracted features.
From: Comprehensive Benchmark Dataset for Pathological Lymph Node Metastasis in Breast Cancer Sections
Methods | Acc (%) | AUC (%) | F1 (%) | Recall (%) | Precision (%) | Kappa |
|---|---|---|---|---|---|---|
Ctranspath6 (WSIs pre-trained) | ||||||
Max-MIL | 83.9 ± 2.34 | 87.1 ± 4.75 | 57.2 ± 3.55 | 57.2 ± 3.12 | 58.6 ± 2.81 | 0.63 ± 0.06 |
Mean-MIL | 75.8 ± 2.56 | 78.6 ± 4.28 | 44.3 ± 4.14 | 43.9 ± 4.47 | 46.3 ± 4.96 | 0.52 ± 0.05 |
ABMIL19 | 84.3 ± 1.09 | 88.3 ± 4.47 | 60.3 ± 2.48 | 60.4 ± 2.05 | 65.7 ± 9.03 | 0.68 ± 0.03 |
Gate-ABMIL19 | 84.1 ± 1.99 | 88.4 ± 4.26 | 57.6 ± 3.69 | 56.4 ± 3.20 | 58.0 ± 3.14 | 0.68 ± 0.04 |
CLAM-SB25 | 84.4 ± 2.14 | 88.6 ± 3.40 | 58.8 ± 2.51 | 57.8 ± 2.20 | 57.7 ± 2.85 | 0.68 ± 0.03 |
CLAM-MB25 | 87.4 ± 1.38 | 89.6 ± 3.84 | 65.6 ± 5.08 | 65.2 ± 5.53 | 66.1 ± 7.02 | 0.72 ± 0.03 |
DSMIL34 | 84.8 ± 2.18 | 85.7 ± 4.83 | 59.0 ± 3.49 | 59.1 ± 2.81 | 60.1 ± 2.19 | 0.65 ± 0.06 |
TransMIL24 | 85.3 ± 3.28 | 89.5 ± 3.42 | 60.7 ± 9.28 | 58.9 ± 12.26 | 60.8 ± 15.92 | 0.71 ± 0.05 |
DTFD35 | 83.9 ± 2.68 | 86.7 ± 5.76 | 56.7 ± 6.00 | 55.3 ± 6.22 | 56.5 ± 4.28 | 0.68 ± 0.04 |
AMD-MIL33 | 85.9 ± 2.30 | 89.4 ± 3.69 | 63.8 ± 4.96 | 63.7 ± 5.09 | 64.8 ± 6.67 | 0.68 ± 0.08 |
WiKG36 | 85.6 ± 2.21 | 89.1 ± 2.22 | 59.9 ± 3.34 | 59.6 ± 3.21 | 60.3 ± 2.93 | 0.67 ± 0.06 |
FR-MIL37 | 84.7 ± 2.45 | 88.2 ± 3.86 | 61.3 ± 1.81 | 61.4 ± 2.35 | 62.0 ± 1.69 | 0.63 ± 0.06 |
