A machine-learning approach for differentiating borderline personality disorder from community participants with brain-wide functional connectivity

• DOI: 10.1016/j.jad.2024.05.125
• Publication Date: 2024-05-26T21:39:36Z
• Authors: Juha M. Lahnakoski; Tobias Nolte; Alec Solway; Iris Vilares; Andreas Hula; Janet Feigenbaum; Terry Lohrenz; Brooks King-Casas; Peter Fonagy; P. Read Montague; Leonhard Schilbach
• Keywords: Functional connectivity; Borderline personality disorder; fMRI; 610; BPD; Classification; Multivariate; info:eu-repo/classification/ddc/610

Functional connectivity has garnered interest as a potential biomarker of psychiatric disorders including borderline personality disorder (BPD). However, small sample sizes and lack within-study replications have led to divergent findings with no clear spatial foci. Evaluate discriminative performance generalizability functional markers for BPD. Whole-brain fMRI resting state in matched subsamples 116 BPD 72 control individuals defined by three grouping strategies. We predicted status using classifiers repeated cross-validation based on multiscale within between regions (ROIs) covering the whole brain—global ROI-based network, seed-based ROI-connectivity, consistency, voxel-to-voxel connectivity—and evaluated classification left-out portion non-matched data. Full-brain allowed (~70 %) patients vs. controls inner cross-validation. The remained significant when applied unmatched out-of-sample data (~61–70 %). Highest accuracies were similar range global (~70–75 %), but spatially more specific. most seed included midline, temporal somatomotor regions. Univariate values not predictive after multiple comparison corrections, weak local effects coincided seed-ROIs. achieved full clinical interview while self-report results at chance level. vary considerably random sub-samples population, signal covariates limiting practical applicability. Spatially distributed patterns are moderately despite heterogeneity patient population.