EEG Signal-Based Machine Learning Approaches for Alzheimer's Disease: A Review of Methodological Analysis

One of the most common dementia forms is Alzheimer's disease (AD). While there may not be a solution available right now, there is promising progress in treatment management and preventative medicine studies. As a result, there is a lot of effort put into creating clinical algorithms that diagnose or identify biomarkers. The superior temporal resolution and low cost of EEG data have made it a top priority to develop automated systems for AD diagnosis, monitoring, and prediction. Many factors, including the dynamic nature of EEG data, changes in subject demographics and clinical aspects, EEG recording technology, and individual variability, limit cutting-edge solutions. This article reviews EEG-based studies that sought to create electrophysiological biomarkers for AD or offered an automated system for doing so. A variety of methods were highlighted in the study, starting with the collection of raw EEG data until the final diagnosis of AD. Several methods of data collection, classification, feature extraction, analysis, and inference are required for this purpose. Each stage's benefits, downsides, and overall effects on the field have been analyzed. Last but not least, a roadmap for future research on cognitive impairment detection systems has been laid forth.