Document Type : Research Paper
Authors
1
Department of Mechanical Engineering, University of Michigan, Michigan, USA.
2
Department of Computer Engineering, Ghiaseddin Jamshid Kashani, Qazvin, Iran.
3
Faculty of Engineering, Departmant of Metallurgical and Materials Engineering, Karadeniz Technical University, 61040 Trabzon, Turkey.
4
Department of Computer Engineering, Bushehr Branch, Islamic Azad University, Bushehr, Iran.
5
ITMO University, Saint Petersburg, Russia.
Abstract
Accurate analysis of electroencephalogram (EEG) signals is essential for the early detection and diagnosis of neurological disorders such as Alzheimer’s, epilepsy, and Parkinson’s disease.
The high-dimensional, noisy, and complex nature of EEG data poses significant challenges to traditional machine learning approaches, which often struggle with interpretability, adaptability, and computational efficiency in clinical applications. Additionally, variability in EEG recordings across individuals further complicates the classification process, necessitating more robust and adaptive methods. To address these challenges, this study introduces the Fuzzy Growing Map (FGM), a novel neuro-fuzzy method that integrates the dynamic properties of the Growing Self-Organizing Map (GSOM) with the uncertainty modeling capabilities of fuzzy logic.
The proposed FGM leverages if-then fuzzy rules to dynamically generate and refine its structure during the learning process. Preprocessing steps extract meaningful features from EEG signals, including Delta, Theta, Alpha, Beta, and Gamma frequency bands, which play crucial roles in neurological assessments. FGM is employed for classification tasks, providing both high accuracy and interpretable outputs, which are critical for clinical decision-making. Experimental results demonstrate that the FGM achieves a classification accuracy of approximately 92% on benchmark EEG datasets, outperforming traditional classification approaches such as Support Vector Machines (85–88%), k-Nearest Neighbors (80–83%), and Multilayer Perceptrons (MLPs) (87%).
By enabling real-time, adaptive, and accurate analysis of EEG signals, the proposed method bridges the gap between theoretical innovations and practical clinical applications. This work underscores the potential of FGM in advancing personalized diagnostics and treatment strategies for patients with neurological conditions.
Future research may focus on extending this approach to multi-channel EEG analysis and real-time braincomputer interface applications, further enhancing its clinical utility.
Keywords
Main Subjects