Enhancing Alzheimer's Disease Detection Through Electroencephalography Channel and Feature Optimization

Alzheimer's Disease (AD), the leading type of dementia, a progressive neurodegenerative disorder impacting mil-lions worldwide, prevails itself with cognitive dysfunction, starting from mild memory impairment and potentially progressing to the loss of conversational ability and responsiveness to the environment while affecting the areas of the brain involved in cognition, memory, and language. Early detection of AD, particu-larly in its preclinical stages, is crucial for timely intervention and effective management. The promoted method stands in contrast to conventional diagnostic approaches, such as neurological ex-aminations, cognitive and functional assessments, neuroimaging techniques and cerebrospinal fluid test some of which are known for their high cost and time requirements and unreliability due to subjectivity and behavioral variability. This paper proposes a novel framework that leverages electroencephalogram (EEG) signals -electrical signal measurements of brain activity- as a more reliable and cost-time effective diagnostic tool. To achieve this, we reduce the necessary number of contact regions into two, while utilising distinct pre-processing tools such as noise removal and filtering, onto 30-second signal segments simultaneously distinguishing the prominent features belonging to different frequency bands which yields higher diagnosis accuracy. Overall, employing widely used machine learning models for AD vs. non-AD binary classification task achieves accuracy rates of 85.5 %.