TITLE:
Analysis and Interpretation of EEG Signals for the Design of an Automated Diagnosis of Disorders of Consciousness and Vigilance through Artificial Neural Network
AUTHORS:
Romain Atangana, Ngassa Pelap, Amstrong Emini Me Zenanga, Daniel Gams Massi, Daniel Tchiotsop, Godpromesse Kenne
KEYWORDS:
Artificial Neural Network, Coma, Discrete Legendre Transform (DLT), Disorders of Consciousness and Vigilance, EEG Signal, Minimal State of Consciousness, Vegetative State
JOURNAL NAME:
World Journal of Neuroscience,
Vol.15 No.4,
November
28,
2025
ABSTRACT: Consciousness and vigilance disorders such as coma, vegetative state, and minimally conscious state represent severe neurological conditions that can affect individuals of all ages and sexes. The lack of early diagnosis of these disorders often leads to serious consequences, including impaired driving performance and complications in organ transplantation. This study introduces a new technique for diagnosing consciousness and vigilance disorders based on features extracted from EEG signals using orthogonal polynomial transforms, combined with machine learning paradigms such as artificial neural networks. Current diagnostic approaches suffer not only from limited accuracy but also from challenges in decision-making and delays in obtaining results. Furthermore, they typically require patients to visit medical facilities, where the diagnosis depends on the availability and expertise of specialists. In this work, we propose an automated diagnosis method that allows an individual, based on their EEG parameters, to determine whether they are affected or not. The algorithm also computes relevant metrics such as sensitivity, specificity, and accuracy. The classification task in this study is binary, aiming to distinguish between healthy individuals and patients with disorders of consciousness (DOC), including coma, unresponsive wakefulness syndrome (UWS), and minimally conscious state (MCS). We employ discrete Legendre transforms to extract discriminative features, which are then classified into two categories: healthy or affected. The results are encouraging, achieving a classification accuracy of up to 75.44% in certain cases using a 10-fold cross-validation technique. Legendre polynomials thus provide a promising tool to complement and improve the detection of consciousness and vigilance disorders. These findings may contribute to enhancing existing state-of-the-art methods for detecting such conditions.