EEG Signal Analysis for Alzheimer's Disease Diagnosis

Background and Aim: Neurodegenerative disorders are a group of diseases that affect the brain. They are fundamentally related to changes in the brain's structure, including the death of some neurons. Alzheimer's Disease (AD) causes structural changes in the brain, affecting behavior, cognition, emotions, and memory. It is difficult to diagnose early, and the symptoms of the disease may be confused with the effects of natural aging. Given this, electroencephalography (EEG) is considered a useful method for the early diagnosis of AD. EEG is one of the imaging modalities to check brain activity. The economic cost of EEG and its ease of use compared to other methods have made it a suitable choice for hospitals and research centers. Methods: In this study, we had ۳۰ subjects with an average of ۶۲ years old, ۱۵ healthy subjects, and ۱۵ subjects with pre-diagnose of AD as a result of the psychiatrist examination and mini- mental state examination (MMSE) test. EEG recording from each subject was in resting state, eyes closed, and for ۱۰ minutes. The recording system was a ۱۹-channel with a ۱۰-۲۰ standard. We extracted signal sub-bands (Delta (۰.۱-۴)Hz, Theta(۴-۸)Hz, Alpha(۸-۱۲) Hz, Beta(۱۳-۳۰)Hz, and Gamma(۳۰-۱۰۰)Hz) using wavelet transform. We analyzed various features as absolute and relative power spectrums and coherences of these sub-bands. In the next step, we used a support vector machine as our classifier. Results: There are many multivariate analysis algorithms to classify AD and distinguish them from healthy cases. Early classification of AD from a healthy population is essential because preventive care can reduce risk factors and costs for the patient. There is currently no definitive cure for AD, but there are ways that it can delay the symptoms if presented early. EEG recordings in patients with AD show some of the modifications that can are used as pathological biomarkers, and classifying them using specific features and machine learning methods can improve our accuracy. Conclusion: It is noticed that EEG modality may be particularly useful in distinguishing between subjects with AD and healthy controls. Results show that there are meaningful relations between changes in relative powers of Theta and Beta sub-bands and AD patients. The accuracy for classification was also around ۹۲% between AD and healthy subjects.

کلیدواژه ها:

Alzheimer's Disease ، Electroencephalography ، Mini-Mental State Examination ، Wavelet Transform ، Support Vector Machine