^ℛSCZ: A Riemannian schizophrenia diagnosis framework based on the multiplexity of EEG-based dynamic functional...

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Abnormal electrophysiological (EEG) activity is a known characteristic of schizophrenia (SCZ).
• Previous research focused on automatic SCZ diagnosis using EEG functional connectivity, often overlooking topological dependencies.
• Existing methods have not fully utilized the geometric properties of functional connectivity brain networks (FCBNs) or analyzed dynamic FCBNs (dFCBNs) for SCZ.

Purpose of the Study:

• To analyze dynamic functional connectivity brain networks (dFCBNs) from EEG data for schizophrenia (SCZ) using Riemannian geometry.
• To propose a novel multiplexity index for quantifying associations between multi-frequency brainwave patterns.
• To develop and validate a machine learning-based decoder for accurate SCZ detection.

Main Methods:

• Analysis of two open EEG-SCZ datasets using Riemannian geometry on symmetric positive definite (SPD) matrices for dFCBN analysis.
• Proposal of a multiplexity index to quantify multi-frequency brainwave pattern associations.
• Implementation of a leave-one-subject-out cross-validation (LOSO-CV) machine learning procedure with classifiers operating on inter-subject dFCBN distances.

Main Results:

• The proposed ^ℛSCZ decoder, utilizing Riemannian geometry and the multiplexity index, achieved 100% absolute accuracy in both analyzed datasets.
• The decoder demonstrated effectiveness in the default mode network (DMN) source space.
• Rhythm-dependent and multiplex-dependent decision-making strategies were employed by the classifiers.

Conclusions:

• The study successfully demonstrates the efficacy of applying Riemannian geometry to dynamic functional connectivity brain networks for schizophrenia detection.
• The developed multiplexity index and machine learning approach offer a highly accurate method for automatic SCZ diagnosis from EEG data.
• This research provides a significant advancement in the field of neuroimaging analysis for mental disorders.