Learning Generative Factors of EEG Data with Variational auto-encoders

TL;DR

This paper introduces a variational auto-encoder framework for analyzing EEG data that classifies pathologies and uncovers underlying neurological mechanisms, demonstrating improved interpretability and performance over benchmarks.

Contribution

The study presents a novel variational auto-encoder approach to learn and decode generative factors of EEG data related to neurological pathologies, enhancing interpretability and classification.

Findings

Successfully classified schizophrenia with and without hallucinations

Learned disease mechanisms consistent with domain knowledge

Outperformed benchmark methods in classification and interpretability

Abstract

Electroencephalography produces high-dimensional, stochastic data from which it might be challenging to extract high-level knowledge about the phenomena of interest. We address this challenge by applying the framework of variational auto-encoders to 1) classify multiple pathologies and 2) recover the neurological mechanisms of those pathologies in a data-driven manner. Our framework learns generative factors of data related to pathologies. We provide an algorithm to decode those factors further and discover how different pathologies affect observed data. We illustrate the applicability of the proposed approach to identifying schizophrenia, either followed or not by auditory verbal hallucinations. We further demonstrate the ability of the framework to learn disease-related mechanisms consistent with current domain knowledge. We also compare the proposed framework with several benchmark approaches and indicate its classification performance and interpretability advantages.