Automated Schizophrenia Detection Using EEG Signal Transformation via Markov Transition Fields and Deep Learning Techniques

Abstract

Diagnosing schizophrenia using Electroencephalograph (EEG) signals is a
challenging task due to subtle and overlapping patterns between patients and healthy individuals.
To address this, this paper proposes a hybrid deep learning framework that transforms onedimensional
EEG signals into two-dimensional representations using the Markov Transition
Field (MTF), effectively capturing temporal dynamics and statistical dependencies. The
generated images are processed using a pre-trained VGG-16 model for robust feature extraction.
The extracted features are then evaluated through two classification pipelines: a Support Vector
Machine (SVM) for traditional machine learning and a deep learning approach combining an
auto encoder for feature selection with a neural network classifier. Experimental evaluation on
the publicly available Schizophrenia EEG dataset from Lomonosov Moscow State University
demonstrates the superior performance of the proposed framework. The deep learning pipeline
achieves a maximum accuracy of 98.51% with 100% recall, while the SVM-based model attains
96.28% accuracy and 97.89% recall, thereby validating the effectiveness of the approach.
Furthermore, the framework incorporates a biomimetic paradigm for enhanced pattern
recognition and decision-making