A smart healthcare-based system for classification of dementia using deep learning

Area of Science:

• Gerontology
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Dementia poses a significant health challenge for the elderly population.
• Early detection is crucial for timely intervention and management of dementia.
• Current diagnostic methods can be invasive or require specialized clinical settings.

Purpose of the Study:

• To develop a deep learning-based classification model for early dementia detection.
• To utilize data from a non-invasive wearable device measuring skin conductivity, temperature, and movement.
• To integrate wearable data with a smart healthcare application for real-time analysis.

Main Methods:

• Recruited 18 elderly participants (aged 65+) categorized into high- or low-dementia risk groups based on the Korean Mini-Mental State Examination (K-MMSE).
• Collected physiological data (skin conductivity, temperature, movement) using wearable devices.
• Employed a deep neural network with scaled principal component analysis for classification and analyzed correlations between K-MMSE scores and wearable data.

Main Results:

• The deep learning model achieved up to 99% accuracy in classifying high-risk dementia individuals.
• The proposed system demonstrated superior performance compared to conventional classification algorithms.
• Skin electrical conductivity showed the strongest correlation with K-MMSE scores among the measured variables.

Conclusions:

• The developed system offers an effective, non-invasive method for early dementia detection in the elderly.
• The combination of wearable technology and a simple cognitive test facilitates accessible dementia risk assessment.
• Future research will involve larger cohorts, additional wearable variables, and long-term effectiveness analysis.