Interstitial Lung Disease (ILD) Classification on 597-patient ILD cohort (test)
86.46AUCRadiomics-Guided
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Radiomics-GuidedGating/Ensemble Strategy=Validation-Weighted, Backbone Architecture=Radiomics-Guided2025.10 | 86.46 | 0.0467 | 80.6 | 12.5 | 0.031 | |
| Mamba Gated MoEGating/Ensemble Strategy=End-to-End Gated MoE, Backbone Architecture=Mamba2025.10 | 81.64 | 0.0393 | 76.8 | 6.2 | 0.053 | |
| CNN WeightedGating/Ensemble Strategy=Validation-Weighted, Backbone Architecture=CNN2025.10 | 80.33 | 0.0472 | 74.5 | 4.5 | 0.176 | |
| ViT WeightedGating/Ensemble Strategy=Validation-Weighted, Backbone Architecture=ViT2025.10 | 80.15 | 0.0453 | 74.5 | 4.3 | 0.179 | |
| ViT Gated MoEGating/Ensemble Strategy=End-to-End Gated MoE, Backbone Architecture=ViT2025.10 | 80.1 | 0.043 | 74.8 | 4.2 | 0.166 | |
| Mamba WeightedGating/Ensemble Strategy=Validation-Weighted, Backbone Architecture=Mamba2025.10 | 79.95 | 0.0466 | 74.2 | 4 | 0.213 | |
| CNN Gated MoEGating/Ensemble Strategy=End-to-End Gated MoE, Backbone Architecture=CNN2025.10 | 77.6 | 0.0689 | 69.1 | 1 | 0.819 | |
| SwinUNETR (Baseline)Gating/Ensemble Strategy=Baseline, Backbone Architecture=SwinUNETR2025.10 | 76.85 | 0.0759 | 67.4 | — | — | |
| CNN (Baseline)Gating/Ensemble Strategy=Baseline, Backbone Architecture=CNN2025.10 | 75.84 | 0.0919 | 66.7 | — | — | |
| Mamba (Baseline)Gating/Ensemble Strategy=Baseline, Backbone Architecture=Mamba2025.10 | 67.75 | 0.0454 | 63.2 | — | — | |
| ViT (Baseline)Gating/Ensemble Strategy=Baseline, Backbone Architecture=ViT2025.10 | 65.35 | 0.0356 | 61.8 | — | — |