A lightweight machine learning tool for Alzheimer's disease prediction.

Area of Science:

• Neuroscience
• Computational Biology
• Medical Informatics

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder requiring improved prediction methods.
• Current diagnostic approaches can be invasive or costly, highlighting the need for accessible predictive tools.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting Alzheimer's disease (AD).
• To create a practical, lightweight clinical tool for AD risk assessment using routinely collected data.

Main Methods:

• Utilized a large dataset (52,537 individuals) from the National Alzheimer's Coordinating Center (NACC) Uniform Data Set.
• Employed advanced ML techniques including LightGBM, genetic algorithms, and iterative backward feature elimination (IBFE) for model development and feature selection.
• Applied SHAP and permutation importance for model interpretability.

Main Results:

• The refined LightGBM model achieved high predictive performance with an ROC-AUC of 0.91 and 82.0% accuracy.
• A simplified 19-feature model maintained strong performance (ROC-AUC 0.90, accuracy 81.2%), identifying key predictors like arthritis, age, BMI, and heart rate.
• SHAP analysis elucidated feature contributions, enhancing model transparency.

Conclusions:

• A lightweight, 19-feature ML tool effectively predicts Alzheimer's disease using common variables.
• The developed tool is accessible via an interactive web app and GitHub, facilitating clinical and research applications.
• Further external validation is recommended due to the study's cross-sectional nature.