نهمین کنگره نقشه برداری مغز

نمایش مقاله

Title:

Multidimensional Analysis of Personality-Scale Prediction from EEG Representations (MAPPER)

Subject:

Clinical Applications of Brain Mappin

Presenter Author:
Name:	Seyed Mahan Mohajerani

Author:
Name 1:	Seyed Mahan Mohajerani - Corresponding - Presenter
Affiliation 1:	College of Humanities and Social Sciences, Islamic Azad University, Science and Research Branch, Tehran, Iran

Name 2:	Moein Esghaei
Affiliation 2:	Faculty of Computer Science and Engineering, Shahid Beheshti University, Tehran, Iran

Name 3:
Affiliation 3:

Name 4:
Affiliation 4:

Name 5:
Affiliation 5:

Name 6:
Affiliation 6:

Name 7:
Affiliation 7:

Name 8:
Affiliation 8:

Name 9:
Affiliation 9:

Name 10:
Affiliation 10:

Abstract

Original Research

introduction :

Personality is a multidimensional phenotype expressed through interacting affective, motivational, cognitive, and regulatory systems. Resting-state electroencephalography (EEG) studies have yielded mixed evidence for personality prediction, and clinically oriented personality dimensions remain comparatively understudied. It is therefore unclear whether EEG contains reproducible information about different Millon Clinical Multiaxial Inventory–IV (MCMI-IV) clinically oriented personality-pattern scales and whether prediction depends on how personality scores and EEG signals are represented. We evaluated this prediction landscape across 15 MCMI-IV clinically oriented personality-pattern scales, four target formulations, and multiple EEG feature families.

Methods:

We analyzed 254 adults presenting to routine clinical practice in Tehran, with complete MCMI-IV and eyes-closed 19-channel resting-EEG data. Each scale was modeled as a continuous base-rate (BR) score, BR60 classification (<60 vs. ≥60), BR75 classification (<75 vs. ≥75), and ordinal severity. Because the prespecified four-class ordinal formulation contained sparse upper categories, a three-class formulation (<60, 60–74, ≥75) was analyzed post hoc and used for primary ordinal interpretation. Eight feature families were extracted: band power, asymmetry, connectivity and graph measures, entropy/complexity, Hjorth parameters, kurtosis, microstates, and periodic–aperiodic spectral parameters. Across these families, 1,215 candidate configurations were evaluated using a common five-fold outer participant-level cross-validation scheme and repeated five-fold inner cross-validation for configuration selection. Performance was summarized in parallel across the Top-5 and Top-20 selected configurations using Area Under the Receiver Operating Characteristic Curve (ROC AUC), Pearson correlation, or quadratic-weighted kappa (QWK), as appropriate. Bootstrap confidence intervals, fixed-candidate permutation tests, and within-formulation multiplicity correction were used.

Results:

As expected, most of the 60 scale-by-formulation combinations showed weak or near-null held-out performance. Convergent rank-mean findings across Top-5 and Top-20 were observed for BR60 Masochistic prediction (AUC = .630 and .594; unadjusted p = .002 and .003), continuous Turbulent prediction (r = .149 and .145; p = .008 and .005), and BR75 Negativistic prediction (AUC = .599 and .585; p = .009 and .013). Masochistic prediction survived within-formulation BH-FDR correction at both widths (q = .030 and .045). Three-class ordinal Histrionic prediction also showed corrected evidence (QWK = .171 and .150; within-formulation max-statistic p = .020 and .030). Additional nominal findings included continuous Avoidant and Compulsive prediction at both widths and BR60 Sadistic prediction at both widths, although its performance was variable due to class imbalance. Some other single width signals involved BR60 Turbulent, BR75 Compulsive, ordinal Borderline and Masochistic, and BR60 Dependent and Avoidant; six scales showed no evidence. Feature-family recurrence differed across leading findings: Masochistic emphasized asymmetry and kurtosis; Turbulent, entropy, spectral parameterization, and microstates; Negativistic, entropy, asymmetry, and kurtosis; and Histrionic, asymmetry and Hjorth features. Exact configurations were unstable across folds, while 3–4-second epochs generally ranked highest.

Conclusion:

Resting EEG contained modest, selective, and formulation-dependent information about clinically oriented personality variation, rather than providing a general decoder of personality pathology. The results suggest that predictability depends jointly on the personality dimension, score formulation, and EEG representation. The identified relationships provide targets for independent, multisite, and repeated-session validation, but should not yet be interpreted as stable biomarkers or used clinically.

Keywords:

electroencephalography; Personality Disorders; Personality Inventory; prediction; individual differences