Table 6. Group Model Performance.
| Noise (Not Noise) | Sensitivity[95% CI] | Specificity[95% CI] | Best CVA[95% CI] | AUC | |
| All Noise | 21 (23) | 0.910.83 1.0 | 0.810.75 0.86 | 0.870.77 0.97 | 0.82 |
| All Selected Features | Weights | ||||
| Percent total activation in GM | −0.7905 | ||||
| Dynamic count diff low | −0.5534 | ||||
| Frontal Inf Tri L | −0.5432 | ||||
| Temporal Pole Mid L | 0.4395 | ||||
| Kurtosis measure how outlier-prone | 0.3655 | ||||
| Cerebellum Crus1 L | −0.3348 | ||||
| Insula L | −0.2233 | ||||
| Angular R | −0.1752 | ||||
| Cuneus R | −0.1367 | ||||
| Skewness of IC distribution | 0.1189 | ||||
| Cingulum Mid L | −0.0743 | ||||
| Occipital Sup R | −0.0314 | ||||
| Four lag auto correlation | 0.0131 | ||||
| Frontal Med Orb R | −0.0129 | ||||
| SupraMarginal L | −0.0077 |
“All Noise types” Group ICA Classifier (M5) (built with combined group ICA decompositions of Data A and Data B) performance, selected features, and weights.