Table 2 Comparative model performance.
From: Multimodal deep learning for midpalatal suture assessment in maxillary expansion
Category | Model | Accuracy | F1-Score | Assessment Basis |
|---|---|---|---|---|
Manual assessment | Resident-1 | 40.11% | 36.38% | Clinical experience |
Resident-2 | 30.36% | 29.00% | Clinical experience | |
Single Modality | ResNet50-SG | 71.25% | 72.00% | CBCT only |
EfficientNet-SG | 70.00% | 68.14% | CBCT only | |
ResNet18-SG | 56.25% | 55.87% | CBCT only | |
MLP-SG | 47.50% | 44.89% | Tabular only | |
Semi-multimodal | ResNet50-SM | 81.25% | 80.77% | CBCT + LCR |
EfficientNet-SM | 75.00% | 74.87% | CBCT + LCR | |
ResNet18-SM | 73.75% | 72.43% | CBCT + LCR | |
Full Multimodal | DeepMSM | 93.75% | 93.81% | CBCT + LCR + Tabular |
