Multi-modal machine learning for predicting amyloid positivity using on-ramp driving.

Area of Science:

• Neurology
• Data Science
• Gerontology

Background:

• Early detection of amyloid positivity is crucial for identifying Alzheimer's disease (AD) risk.
• This study explores non-invasive methods for cognitive health monitoring.

Purpose of the Study:

• To classify older adults with and without amyloid positivity using multi-modal data.
• To evaluate the effectiveness of machine learning models in predicting amyloid status.

Main Methods:

• Collected driving and physiological data from 53 cognitively normal older drivers.
• Utilized random forest and XGBoost classifiers trained on statistically significant features (P ≤ 0.05).
• Analyzed multi-modal attributes including vehicular, physiological, and demographic data.

Main Results:

• Integrating multiple data modalities improved classification performance for amyloid status.
• The XGBoost model achieved the highest accuracy (85.1%) using all significant features.
• Vehicular data, particularly on-ramp driving behavior, demonstrated the most predictive power.

Conclusions:

• Multi-modal data analysis during on-ramp driving aids in early cognitive decline detection.
• Challenging traffic environments can be leveraged for non-invasive cognitive health monitoring.
• Results highlight the importance of on-ramp driving insights for predicting amyloid status and early AD detection.