Multi-Class Detection of Neurodegenerative Diseases from EEG Signals Using Lightweight LSTM Neural Networks.

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Neurodegenerative diseases pose a growing global health challenge, impacting millions.
• Electroencephalography (EEG) is a valuable diagnostic tool but generates large datasets.
• Machine learning (ML) offers a solution for analyzing complex EEG data and early disease detection.

Purpose of the Study:

• To develop and evaluate a machine learning model for multi-class detection of neurodegenerative diseases using EEG.
• To assess the efficacy of Long Short-Term Memory (LSTM) neural networks in classifying EEG data from patients with specific neurodegenerative conditions.

Main Methods:

• Acquisition of a custom dataset of EEG recordings from subjects with five neurodegenerative diseases and a control group.
• Design and training of Long Short-Term Memory (LSTM) neural networks with varying numbers of units.
• Rigorous data pre-processing to prepare EEG signals for analysis.

Main Results:

• Achieved up to 98% accuracy in multi-class classification of EEG data.
• Successfully distinguished between different neurodegenerative disease classes, including a control group.
• The developed model demonstrated high performance without requiring excessive computational resources.

Conclusions:

• LSTM-based neural networks are effective tools for the accurate, automated classification of neurodegenerative diseases from EEG data.
• This approach shows significant potential for early diagnosis and improved patient outcomes.
• The method offers a computationally efficient solution for analyzing complex neurological data.