Automated Alzheimer’s Disease Diagnosis Using Norm Features Extracted from EEG Signals

Alzheimer's disease (AD) is a neurodegenerative disorder that progresses over time and affects cognitive abilities. It is marked by symptoms such as memory loss, language and communication difficulties, and impaired judgement and decision-making. These symptoms are severe enough to affect daily life of mostly middle-aged and elderly adults. Although conventional AD diagnosis approaches, such as cognitive assessment tests, are inexpensive but time-consuming, biased, and less accurate. Hence, it is essential to develop computer-assisted automated diagnosis of AD at an early stage to minimize the symptomatic effects. In the early stages, neuroimaging (magnetic resonance imaging (MRI) or computed tomography (CT)) scans may not reveal any abnormalities, but in later stages, diffuse cortical atrophy may be observed in the frontal and temporal regions of the brain. In light of this problem, the electroencephalogram (EEG) is opted for early diagnosis of AD because of its ability to show cortical neuronal functionalities in high temporal resolution at low-cost and non-invasively. In this study, an automated AD diagnosis framework has been designed based on machine learning (ML) algorithms using EEG signals. The signals are pre-processed and decomposed into five popular EEG rhythms or bands. The norm features are extracted from each bands and trails of EEG data. Later, the extracted features are supplied as input to the popular classifiers and their classification performance are evaluated. A publicly accessible EEG dataset featuring 16-channels EEG obtained from 23 adults (11 healthy and 12 AD) is used to demonstrate the proposed methodology. An extensive comparison of different ML algorithms is done on selected EEG data, and it is found that the ensemble tree classifier shows the best results at delta band. The experimental results suggest that the proposed methodology could be a potential solution and provide a robust biomarker for effective diagnosis of AD.