Table 3 Evaluation results on SICAPv2 dataset as Mean ± std and the bold font is the best scores.
From: Enhanced hierarchical attention mechanism for mixed MIL in automatic Gleason grading and scoring
Method | Per-class k-score | K-scores | AUC | |||
|---|---|---|---|---|---|---|
Benign | Grade3 | Grsde4 | Grade5 | |||
ABMIL | – | – | – | – | – | 0.4888 ± 0.0377 |
CLAM | 0.7278 ± 0.0019 | 0.1844 ± 0.0238 | 0.9778 ± 0.0006 | 0.1645 ± 0.032 | 0.4852 ± 0.0033 | 0.8417 ± 0.0238 |
DSMIL | 0.7039 ± 0.0006 | 0.0775 ± 0.0009 | 1.0000 | 0.0207 ± 0.0017 | 0.4207 ± 0.0002 | 0.7621 ± 0.0488 |
ATMIL | – | – | – | – | – | 0.9373 ± 0.0294 |
TransMIL | 0.7214 ± 0.0067 | 0.6031 ± 0.0181 | 0.8894 ± 0.0006 | 0.6406 ± 0.0174 | 0.6347 ± 0.0028 | 0.8407 ± 0.0367 |
Mixed MIL | 0.8458 ± 0.0083 | 0.9060 ± 0.0011 | 0.8905 ± 0.0005 | 0.8048 ± 0.017 | 0.8101 ± 0.0041 | 0.9361 ± 0.0278 |
Ours | 0.8728 ± 0.1548 | 0.9378 ± 0.0027 | 0.8906 ± 0.1628 | 0.8963 ± 0.0147 | 0.8494 ± 0.0505 | 0.9597 ± 0.0290 |
