Utility of MemTrax and Machine Learning Modeling in Classification of Mild Cognitive Impairment

Area of Science:

• Neurology
• Cognitive Science
• Machine Learning

Background:

• Alzheimer's disease and mild cognitive impairment (MCI) are prevalent, necessitating early detection methods.
• Current cognitive screening and assessment tools require validation for early-stage detection.

Purpose of the Study:

• To determine if MemTrax performance metrics, demographics, and health data can predict cognitive health using machine learning.
• To classify cognitive health (normal vs. MCI) based on MemTrax data and patient characteristics, validated against the Montreal Cognitive Assessment (MoCA).

Main Methods:

• A cross-sectional study of 259 adult patients from China.
• Patients completed the Chinese MoCA and the MemTrax episodic memory test.
• Machine learning models with 10-fold cross-validation were used to predict cognitive status based on MemTrax metrics and demographic features.

Main Results:

• Naïve Bayes models achieved high classification performance (0.9093).
• MemTrax-based classification using the top four features (0.9119) outperformed models using all ten features (0.8999).

Conclusions:

• MemTrax performance metrics are effective for machine learning-based predictive models.
• This approach shows promise for screening applications in detecting early-stage cognitive impairment.