Table 2.

External test performance results in scoliosis and unreliability detection using the KUH dataset. Values shown are averages with minimum and maximum value over the 10 models that produced the median AUC value in repeated 10-fold cross-validation using the OSTPRE dataset.

Experiment	Sensitivity (%)	Specificity (%)	Accuracy (%)	AUC
Scoliosis	74.5 (65.0–80.5)	93.7 (90.9–96.9)	89.6 (88.3–90.2)	0.93 (0.92–0.95)
Unreliability	92.7 (89.1–97.9)	72.9 (56.7–81.0)	82.8 (77.2–85.0)	0.94 (0.94–0.95)