Poincaré feature-based classification of electroencephalography signals for multiple sclerosis diagnosis

As one of the most widespread neurodegenerative disorders, Multiple Sclerosis (MS) is a progressive neuroinflammatory disorder affecting millions of individuals worldwide. Although magnetic resonance imaging (MRI) techniques are commonly employed for diagnosis, there is an increasing emphasis on electroencephalogram (EEG) signal processing to diagnose neurodegenerative disorders.

This study explores the use of Poincaré plot-derived features from EEG signals to differentiate MS patients from healthy individuals. EEG recordings from 50 subjects (25 MS, 25 Healthy controls) have been analyzed. EEG data were segmented into epochs, and four quantitative Poincaré features were extracted from each segment. These features were used as inputs to various classifiers, including traditional machine learning methods, k-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), and ensemble models, and deep learning architectures such as Multilayer Perceptron (MLP), Convolutional Neural Network combined with Long Short-Term Memory (CNN+LSTM), and LSTM combined with Gated Recurrent Units (LSTM+GRU). Additionally, sub-frequency bands of EEG signals were analyzed separately to evaluate the discriminative potential of each band.

The results demonstrated that Poincaré-based features effectively distinguished MS patients from healthy individuals. This yielded accuracies, sensitivities, and specificities of 99.8%, 100%, and 99.7%, respectively. Moreover, classification based on the Beta band consistently exhibited the highest discriminative power in differentiating MS patients. However, given the limited sample size, these results should be interpreted as preliminary and warrant further validation.

The results indicate that EEG-based Poincaré feature analysis offers a promising, low-cost, and noninvasive approach for assisting in MS diagnosis. While the observed classification performance is encouraging, larger and more diverse datasets, along with rigorous validation strategies, are required to confirm the robustness and clinical applicability of the proposed approach. Integrating this method with clinical workflows may enhance diagnostic accuracy and provide new insights into MS-related alterations in brain dynamics.