Machine-Learning Algorithms Based on Screening Tests for Mild Cognitive Impairment

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Engineering

Background:

• The Montreal Cognitive Assessment (MoCA) has limitations in sensitivity and specificity for mild cognitive impairment (MCI).
• A novel mobile screening test system for mild cognitive impairment (mSTS-MCI) was developed to improve diagnostic accuracy.

Purpose of the Study:

• To evaluate the efficacy of machine learning algorithms utilizing the mSTS-MCI and the Korean version of MoCA.
• To compare the diagnostic performance of these algorithms against conventional screening tools.

Main Methods:

• 103 healthy individuals and 74 MCI patients were randomly assigned to training and testing datasets.
• Machine learning algorithms, including TensorFlow, were trained and validated using these datasets.
• Logistic regression was employed for cost calculation.

Main Results:

• Machine learning algorithms demonstrated superior predictive power compared to the original screening tests.
• The algorithm based on the mSTS-MCI achieved the highest positive-predictive value.
• The predictive capabilities of the machine learning models were comparable to existing clinical screening tools.

Conclusions:

• Machine learning algorithms show promise for enhancing MCI detection.
• The mSTS-MCI, when integrated with machine learning, offers a potentially more accurate method for identifying mild cognitive impairment.
• Further validation of these AI-driven tools is warranted for clinical application.