Episodic-Memory Performance in Machine Learning Modeling for Predicting Cognitive Health Status Classification

Area of Science:

• Neuroscience
• Artificial Intelligence
• Gerontology

Background:

• Memory dysfunction is a hallmark of aging and a key indicator of Alzheimer's disease (AD).
• Early detection of cognitive impairment is crucial for effective patient management.
• The MemTrax Continuous Recognition Tasks (M-CRT) test offers a potential tool for preliminary memory assessment.

Purpose of the Study:

• To apply machine learning for developing predictive models using M-CRT data.
• To validate the efficacy of the M-CRT test in screening for cognitive impairment and early AD detection.
• To explore the utility of M-CRT in conjunction with demographic and health data.

Main Methods:

• Utilized a dataset of 18,395 participants including demographic information, health screening questions, and M-CRT test results.
• Employed machine learning algorithms, including logistic regression, to predict health status and cognitive function.
• Analyzed M-CRT performance metrics (e.g., response time, accuracy) and participant features.

Main Results:

• Logistic regression demonstrated moderate predictive performance (AUC 0.648-0.769) for cognitive impairment and general health status.
• Significant differences in M-CRT performance were observed across different health score groups.
• The M-CRT test showed utility in assessing episodic memory and cognitive status.

Conclusions:

• Supervised machine learning and predictive modeling validate the cross-sectional utility of MemTrax for cognitive screening.
• The M-CRT test, enhanced by machine learning, shows promise for early-stage cognitive impairment and Alzheimer's disease detection.
• This approach offers a complementary tool for assessing memory function in aging populations.