Table 3 Calinski–Harabasz (CH) Index, and Robustness Index (RI) for different scanners and tissue types for all models.
From: Lightweight self supervised learning framework for domain generalization in histopathology
Models | Tumour CH \(\downarrow\) | Non-Tumour CH \(\downarrow\) | Scanner CH \(\downarrow\) | Tissue CH \(\uparrow\) | RI |
|---|---|---|---|---|---|
HistoLite (ours) | 1567.1 | 617.0 | 1709.4 | 1204.3 | 1.10 |
KimiaNet | 1005.2 | 601.6 | 1426.2 | 309.2 | 0.38 |
PathDino | 1249.6 | 836.1 | 1814.0 | 768.6 | 0.73 |
HIPT | 112.2 | 47.6 | 128.8 | 1802.5 | 22.57 |
iBOT-Path | 755.5 | 567.8 | 1105.6 | 838.7 | 1.26 |
Hibou-B | 1507.3 | 810.6 | 1977.2 | 434.5 | 0.37 |
UNI | 497.8 | 317.4 | 714.4 | 353.4 | 0.86 |
Virchow | 2176.5 | 1212.1 | 2956.9 | 464.4 | 0.27 |
Virchow2 | 615.1 | 305.5 | 774.9 | 584.8 | 1.27 |
Prov-GigaPath | 311.6 | 208.3 | 433.8 | 522.8 | 2.01 |
