Depression Detection on MODMA (5-fold cross-val)
86.4Macro F1TRI-DEP (Weighted Averaging)
Evaluation Results
| Method | Links | |
|---|---|---|
| TRI-DEP (Weighted Averaging)Modalities=EEG + Speech + Text, Fusion Weights=0.05 : 0.35 : 0.602025.10 | 86.4 | |
| TRI-DEP (Bayesian Fusion)Modalities=EEG + Speech + Text, Fusion Weights=0.05 : 0.35 : 0.652025.10 | 86.4 | |
| Bayesian FusionModalities=Speech + Text, Fusion Weights=0.20 : 0.802025.10 | 86.3 | |
| Weighted AveragingModalities=Speech + Text, Fusion Weights=0.45 : 0.552025.10 | 83.9 | |
| Weighted AveragingModalities=EEG + Speech, Fusion Weights=0.05 : 0.952025.10 | 80.9 | |
| Bayesian FusionModalities=EEG + Speech, Fusion Weights=0.05 : 0.952025.10 | 80.9 | |
| Majority VotingModalities=Speech + Text2025.10 | 80.9 | |
| Majority VotingModalities=EEG + Speech + Text2025.10 | 80 | |
| Weighted AveragingModalities=EEG + Text, Fusion Weights=0.15 : 0.852025.10 | 78.6 | |
| Bayesian FusionModalities=EEG + Text, Fusion Weights=0.05 : 0.952025.10 | 78.4 | |
| DenseNet-121Fusion Strategy=Baselines2025.10 | 58.6 | |
| ViTFusion Strategy=Baselines2025.10 | 56 | |
| Majority VotingModalities=EEG + Speech2025.10 | 55.7 | |
| Majority VotingModalities=EEG + Text2025.10 | 50 |