Phenotyping on (EHR + CXR)PAIRED (test)
0.77AUROCMedFuse
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MedFuseFine-tuning strategy=OPTIMAL2022.07 | 0.77 | 0.481 | |
| MedFuseTraining set=(EHR + CXR)_PARTIAL2022.07 | 0.77 | 0.481 | |
| Unified2022.07 | 0.765 | 0.461 | |
| MedFuseEncoder Initialization=Pre-trained (PT)2022.07 | 0.756 | 0.466 | |
| Joint FusionTraining Data=(EHR + CXR)PARTIAL2022.07 | 0.754 | 0.458 | |
| Early FusionTraining Data=(EHR + CXR)PAIRED2022.07 | 0.753 | 0.453 | |
| MedFuseEncoder Initialization=Randomly Initialized (RI)2022.07 | 0.748 | 0.452 | |
| Joint FusionTraining Data=(EHR + CXR)PAIRED2022.07 | 0.747 | 0.446 | |
| Early FusionTraining Data=(EHR + CXR)PARTIAL2022.07 | 0.739 | 0.435 | |
| DAFT2022.07 | 0.737 | 0.434 | |
| MMTM2022.07 | 0.734 | 0.428 |