An explainable machine learning based prediction model for Alzheimer's disease in China longitudinal aging study

Area of Science:

• Neurology
• Artificial Intelligence
• Gerontology

Background:

• Alzheimer's disease (AD) is the leading cause of dementia, necessitating early diagnosis for effective management.
• Mild cognitive impairment (MCI) is a prodromal stage of AD, highlighting the need for accurate prediction.
• The China Longitudinal Aging Study (CLAS) provides a valuable dataset for investigating AD and MCI prediction.

Purpose of the Study:

• To develop and validate an explainable prediction model for Alzheimer's disease (AD) and mild cognitive impairment (MCI).
• To identify key features from lifestyle and medical history that contribute to AD/MCI prediction.
• To enhance clinical decision-making through a computer-aided diagnosis system.

Main Methods:

• Utilized ensemble learning and feature selection techniques on CLAS data (n=3,514).
• Applied and compared nine machine learning classifiers and five feature selection methods.
• Employed the SHapley Additive exPlanations (SHAP) algorithm for model interpretability.

Main Results:

• Achieved high prediction accuracy for MCI (89.2%) and AD (99.2%).
• Demonstrated excellent sensitivity and specificity for both MCI and AD prediction.
• Identified significant lifestyle and medical history factors influencing cognitive function.

Conclusions:

• The developed explainable AI model accurately predicts AD and MCI, offering valuable insights into contributing factors.
• The model's interpretability aids in understanding the relationship between lifestyle, health history, and cognitive decline.
• This approach holds significant potential for clinical application in computer-aided diagnosis systems for AD and MCI.