Table 15 Analysis on robustness through multimodal imaging integration.
Model | Input modality | Accuracy (%) | Precision | Recall | F1-score |
|---|---|---|---|---|---|
ResNet-50 | CT only | 87.6 | 0.85 | 0.86 | 0.855 |
Capsule network (proposed) | CT only | 91.0 | 0.90 | 0.89 | 0.895 |
Hybrid (ResNet-50 + CapsNet) | CT only | 92.4 | 0.91 | 0.91 | 0.91 |
ResNet-50 | CT + PET | 90.1 | 0.89 | 0.88 | 0.885 |
Capsule network (proposed) | CT + PET | 95.1 | 0.94 | 0.95 | 0.945 |
Hybrid (ResNet-50 + CapsNet) | CT + PET | 96.3 | 0.95 | 0.96 | 0.955 |
ResNet-50 | CT + fMRI | 89.5 | 0.88 | 0.87 | 0.875 |
Capsule network (proposed) | CT + fMRI | 94.7 | 0.93 | 0.94 | 0.935 |
Hybrid (ResNet-50 + CapsNet) | CT + fMRI | 95.6 | 0.94 | 0.95 | 0.945 |
