Towards Precision Forensic Psychiatry: An Advanced Machine Learning EEG Model for High-Accuracy Borderline Personality Disorder Diagnosis

AbstractBorderline Personality Disorder (BPD) is characterized by emotional instability, impulsivity, and turbulent interpersonal relationships. Despite its profound clinical and forensic implications, diagnosis largely relies on subjective assessments. Recent studies suggest that electroencephalography (EEG) can reveal neurophysiological biomarkers associated with BPD, such as altered spectral power, abnormal event-related potentials (ERPs), disrupted functional connectivity, and modified signal complexity. This article presents a comprehensive machine learning framework that integrates a wide range of EEG features to classify individuals with BPD versus healthy controls. Our approach employs traditional classifiers (e.g., Support Vector Machines, Random Forests) and deep learning models (e.g., Convolutional Neural Networks, Long Short-Term Memory networks, and transformer-based architectures) as well as ensemble strategies. Six graphs illustrate key findings: (1) power spectrum differences, (2) ERP differences (focusing on an emotional Late Positive Potential), (3) connectivity alterations, (4) complexity analysis via sample entropy, (5) performance comparison across models, and (6) a confusion matrix for the best model. Our results underscore the potential of EEG-based machine learning to contribute to a more objective and precise diagnosis of BPD.