Optimizing EEG Channel Selection for Alzheimer’s Disease Classification: A Regional Brain Analysis

Progressive neurodegenerative disorders like Alzheimer’s disease (AD) have been difficult to diagnose earlier using traditional techniques like MRI and CT scans since they don’t work on a real-time approach. Electroencephalography (EEG) signals are now considered effective in mitigating this late detection issue. This is more popular due to its non-invasiveness and instantaneous work function. However, it also has some adversities due to its lack of cost-effectiveness since many electrode channels are used to detect the dysfunctions from the different lobes of the brain, making it look more complex. So, in this study, we have aimed to enhance EEG-based diagnostics for Alzheimer's through comprehensive feature extraction incorporating machine learning as well as electrode channel reduction analysis. Three major brain regions: frontal, parietal, and temporal, have been considered for effective AD detection. 14 electrodes from these brain regions are analyzed for significant feature extraction. Then, we figured out the necessary statistical and frequency features from each EEG segment. The features went through different tests to check the most important features. Different combinations of features were applied for various machine learning algorithms, and the highest accuracy of 88.5% was achieved in the parietal region using 26 features. In the future, the parietal region can be further analyzed with more features and deep learning methods which will help in medical diagnosis and treatment.