Developing an explainable machine learning and fog computing-based visual rating scale for the prediction of dementia...

Area of Science:

• Gerontology and Cognitive Science
• Health Informatics and Machine Learning

Background:

• Current dementia research often relies on Magnetic Resonance Imaging (MRI) for machine learning models, but these methods may not detect early-stage affected brain regions.
• Cognitive assessments like the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) provide valuable indicators for dementia likelihood and cognitive impairment.

Purpose of the Study:

• To develop and evaluate an innovative, dependable, context-aware health monitoring system using fog computing for real-time mental impairment measurement in the elderly.
• To leverage MMSE and MoCA screening tests within the fog computing framework for early detection and timely treatment of cognitive disorders.

Main Methods:

• A dataset of 450 subjects (150 Mild Cognitive Impairment (MCI), 150 Parkinson's Disease (PD), 150 Alzheimer's Disease (AD)) was utilized.
• An ensemble AdaBoost Machine Learning (ML) model was employed for accurate health risk determination and classification.
• The ML model's effectiveness was validated on an additional 18 subjects (6 from each class) using the proposed scoring test.

Main Results:

• The proposed ML model achieved a high accuracy of 0.93 in diagnosing cognitive impairment.
• The system's performance surpassed traditional scores: MoCA (0.90) and MMSE (0.83).
• The AdaBoost model demonstrated superior performance across accuracy, precision, recall, F-score, and Area Under the Curve (AUC).

Conclusions:

• The context-aware fog computing approach significantly enhances early dementia diagnosis by effectively utilizing mental test scores.
• This system offers a promising, reliable method for real-time cognitive function monitoring and early intervention in elderly populations.