Inter-Hospital Advanced and Mild Alzheimer's Disease Classification Based on Electroencephalogram Measurements via...

Area of Science:

• Neuroscience
• Computational Biology
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) diagnosis requires tools to differentiate disease stages.
• Electroencephalography (EEG) and machine learning (ML) offer a potential non-invasive approach.

Purpose of the Study:

• To evaluate the efficacy of six classical ML algorithms in classifying AD stages.
• To assess the diagnostic accuracy of EEG-ML for mild-to-moderate AD (ADM) and advanced AD (ADA).

Main Methods:

• Utilized a dataset of 668 patients from multiple hospitals (2011-2022).
• Classified patients into control, ADM, and ADA groups.
• Employed six ML algorithms: SVM, BLDA, Decision Tree, GNB, KNN, and Random Forest (RF).

Main Results:

• The Random Forest (RF) algorithm demonstrated superior performance.
• RF achieved 93.55% balanced accuracy for ADA classification.
• RF achieved 93.25% balanced accuracy for ADM classification.

Conclusions:

• EEG combined with ML provides a robust method for AD classification.
• The study highlights the feasibility of using straightforward ML models for reliable AD diagnosis.
• This approach shows significant potential for developing advanced diagnostic tools for Alzheimer's disease.