Table 4.
Performance of models in predicting DMFS
| Models | Training cohort |
Validation cohort |
||||
|---|---|---|---|---|---|---|
| C-index | 95% CI | p value | C-index | 95% CI | p value | |
| Radiomics combined model | 0.66 | 0.59–0.73 | Ref | 0.71 | 0.62–0.80 | Ref |
| Clinical model | 0.77 | 0.71–0.83 | 0.005 | 0.73 | 0.64–0.82 | 0.97 |
| DL combined model | 0.85 | 0.78–0.92 | <0.001 | 0.84 | 0.78–0.90 | 0.02 |
| RC model | 0.78 | 0.71–0.84 | <0.001 | 0.78 | 0.70–0.86 | 0.43 |
| DC model | 0.89 | 0.84–0.94 | <0.001 | 0.87 | 0.80–0.93 | 0.009 |
| DC model vs. DL combined model | – | – | 0.02 | – | – | 0.76 |
| RC model vs. Clinical model | – | – | 0.65 | – | – | 0.28 |
| AUC | 95% CI | p value | AUC | 95% CI | p value | |
|---|---|---|---|---|---|---|
| Radiomics combined model | 0.67 | 0.59–0.74 | Ref | 0.70 | 0.60–0.80 | Ref |
| Clinical model | 0.78 | 0.73–0.85 | 0.01 | 0.72 | 0.62–0.81 | 0.78 |
| DL combined model | 0.85 | 0.78–0.93 | <0.001 | 0.84 | 0.77–0.91 | 0.02 |
| RC model | 0.80 | 0.73–0.87 | <0.001 | 0.74 | 0.65–0.84 | 0.36 |
| DC model | 0.90 | 0.85–0.96 | <0.001 | 0.85 | 0.78–0.93 | 0.01 |
| DC model vs. DL combined model | – | – | 0.007 | – | – | 0.75 |
| RC model vs. Clinical model | – | – | 0.53 | – | – | 0.65 |
Note that, DL combined model and Radiomics combined model were conducted based on the three MR sequences (T1W, T2W, and CET1W). DC model, a model combining deep learning and clinical variables. RC model, a model combining radiomics and clinical variables.
Abbreviations: NPC, nasopharyngeal carcinoma; C-index, Harrell’s concordance index; CI: confidence interval; DL, deep learning; Ref, reference; AUC, area under curve.