Table 1.
Summary of patients’ characteristics.
Fig 1.
Manual segmentation of the entire oral cavity on CT images from a representative patient.
Fig 2.
Four-Step Core Workflow Diagram.
Table 2.
Performance Comparison of Nine Machine Learning Models (5-Fold Cross-Validation, Mean ± SD).
Fig 3.
Performance comparison of nine conventional machine learning models.
Bars represent mean performance (accuracy, F1 score, and AUC) across five-fold cross-validation. Error bars indicate standard deviation (SD).
Fig 4.
Comparison of 1D-CNN and 3D-CNN Performance Metrics.
Fig 5.
Comparison of ROC Curves Between Traditional Machine Learning Models.
Fig 6.
Comparison of ROC Curves Between Deep Learning Models.
Fig 7.
Confusion matrices of top-performing machine learning algorithms and representative deep learning models.
Fig 8.
Model interpretability and modality contribution analysis.
(A) SHAP summary plot displaying the top 20 features and their impact on model predictions. (B) Relative contribution of Dosiomics, Radiomics, and Clinical modalities to the multimodal model.
Fig 9.
(A) Calibration curve with Brier score = 0.006. (B) Decision curve analysis (DCA) showing net clinical benefit.