Can EEG patterns distinguish psychiatric-service adults with recorded violent offences from non-clinical controls, and how reliable is that distinction?

Details

This retrospective study examined whether rhythm-resolved EEG features could separate two groups and whether those features could be traced to explicit rhythms and sensors. The groups were psychiatric-service adults with recorded violent offences and non-clinical controls; this was not a comparison of offenders with non-offenders.

The researchers analysed 2554 non-overlapping 15-second EEG epochs from 95 adults: 51 controls and 44 recorded-offence participants. RhythmFluxPat generated 9216 rank-transition features. CWImRMR selected 352 features using full-cohort labels, while classifier selection was nested within 95 participant-held-out leave-one-subject-out (LOSO) folds.

Classifier-nested tkNN correctly classified 2392/2554 epochs (93.66%) and 93/95 participants (97.89%). An independent check reproduced both confusion matrices. A nested support-vector machine (SVM) produced similar values: 94.71% of epochs and 93/95 participants. In a comparative evaluation of four feature selectors, CWImRMR remained the best branch at 93.85%, and a 44-feature subset reached 91.11%.

DRLobish linked the selected features to delta and beta-to-gamma routes in the recorded-offence-assigned panel, and to theta and alpha-to-theta routes in the control-assigned panel. The authors concluded that RhythmFluxPat provides traceable rank-persistence and rank-exchange coordinates.

The reported accuracies are optimistic internal estimates. Fully nested replication with matched psychiatric controls is required before any claim of generalisation or biomarker validity.