EEG classification for neurological disorders using frequency band deciles.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Electroencephalography (EEG) is crucial for monitoring brain activity and diagnosing neurological disorders.
• Traditional signal processing methods for EEG analysis often neglect important statistical properties.
• There is a need for interpretable and effective feature extraction techniques for EEG data.

Purpose of the Study:

• To introduce a novel decile-based feature extraction method for EEG signal analysis.
• To evaluate the proposed method's effectiveness in classifying neurological conditions.
• To assess the method's performance with different machine learning models and its robustness to reduced channel counts.

Main Methods:

• A decile-based feature extraction technique was developed for EEG signals.
• The method was tested on datasets for Alzheimer's disease, frontotemporal dementia, Parkinson's disease, and seizure detection.
• Classification was performed using Random Forest (RF), K-Nearest Neighbors (KNN), and LightGBM models.

Main Results:

• The decile-based features achieved competitive classification accuracy, especially when combined with RF and KNN.
• The method demonstrated robustness to a reduced number of EEG channels.
• Performance varied across different machine learning models and datasets, with LightGBM showing particular variability.

Conclusions:

• Decile-based features offer a simple, interpretable, and effective representation for diverse EEG classification tasks.
• The approach shows promise for low-cost, wearable EEG systems and ambulatory monitoring.
• Further validation in larger, diverse populations is necessary to confirm clinical applicability for early diagnosis.