Performance of multiclass classification across all models and experts in test set
| Method | Micro-Averaged AUC | Accuracy (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) |
|---|---|---|---|---|
| Radiomics (by TPOT) | 0.91 | 0.83 (0.72–0.90) | MB: 0.87 (0.67–0.96) | MB: 0.91 (0.78–0.97) |
| EP: 0.67 (0.46–0.83) | EP: 0.98 (0.88–1.00) | |||
| PA: 0.95 (0.76–1.00) | PA: 0.86 (0.72–0.94) | |||
| Radiomics (by CHSQ and GLM) | 0.92 | 0.74 (0.62–0.83) | MB: 0.96 (0.77–1.00) | MB: 0.84 (0.70–0.92) |
| EP: 0.33 (0.17–0.55) | EP: 0.93 (0.81–0.98) | |||
| PA: 0.91 (0.71–0.99) | PA: 0.84 (0.70–0.92) | |||
| Expert 1 | NA | 0.58 (0.46–0.69) | MB: 0.65 (0.45–0.81) | MB: 0.67 (0.52–0.79) |
| EP: 0.57 (0.36–0.75) | EP: 0.82 (0.68–0.91) | |||
| PA: 0.50 (0.31–0.69) | PA: 0.86 (0.72–0.94) | |||
| Expert 2 | NA | 0.50 (0.38–0.62) | MB: 0.57 (0.37–0.75) | MB: 0.66 (0.51–0.77) |
| EP: 0.43 (0.25–0.64) | EP: 0.80 (0.66–0.89) | |||
| PA: 0.50 (0.31–0.69) | PA: 0.77 (0.63–0.87) |
Note:—NA indicates not applicable.