Table 3.

Predictive performance of the various random forest radiomics classifiers (model A and model B) on the external validation set.

Feature set	AUROC (95% CI)	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)
Model A (n = 33)	79% (62%–96%)	50% (16%–84%) 4/8	93% (81%–99%) 39/42	57% (18%–90%) 4/7	91% (78%–97%) 39/43
Model B (n = 91)	83% (70%–97%)	38% (9%–76%) 3/8	90% (77%–97%) 38/42	43% (10%–82%) 3/7	88% (75%–96%) 38/43

Abbreviations: AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value; pCR, pathological complete response; pPR, pathological partial response