Phenotyping on (EHR + CXR) PARTIAL (test)
0.768AUROCMedFuse (OPTIMAL)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MedFuse (OPTIMAL)Training set=(EHR + CXR)PARTIAL, Fine-tuning=uses optimal number of uni-modal samples2022.07 | 0.768 | 0.429 | |
| MedFuseTraining set=(EHR + CXR)PARTIAL, Handling of missing chest X-ray images=substituted with a learnable vector2022.07 | 0.758 | 0.418 | |
| JointTraining set=(EHR + CXR)PARTIAL, Handling of missing chest X-ray images=substituted with a learnable vector2022.07 | 0.754 | 0.41 | |
| EarlyTraining set=(EHR + CXR)PARTIAL, Handling of missing chest X-ray images=substituted with a learnable vector2022.07 | 0.748 | 0.394 |