Alzheimer's disease risk prediction using machine learning for survival analysis with a comorbidity-based approach

Area of Science:

• Neuroscience
• Gerontology
• Biostatistics

Background:

• Alzheimer's disease (AD) poses a significant global health challenge.
• Early detection and understanding disease progression are critical for effective management.
• Existing predictive models often lack comprehensive feature integration.

Purpose of the Study:

• To develop and validate a predictive model for the transition from Cognitive Normal (CN) to Mild Cognitive Impairment (MCI).
• To assess the predictive value of baseline comorbidities in Alzheimer's disease risk.
• To identify key predictors for early AD detection using survival analysis.

Main Methods:

• Utilized survival analysis techniques on data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and Australian Imaging, Biomarker & Lifestyle Flagship Study of Ageing (AIBL).
• Constructed feature sets including demographics, cognitive scores (e.g., ADAS13, RAVLT, FAQ, CDRSB), and comorbidities (Endocrine & Metabolic, Renal & Genitourinary).
• Employed various machine learning and deep learning survival analysis models, including fast random forest, for prediction and feature importance analysis.

Main Results:

• The fast random forest model achieved a high concordance index of 0.84.
• Comorbidity data was identified as a significant predictor of AD progression.
• Key predictors included age, several cognitive scores, and specific comorbidity categories.

Conclusions:

• Baseline comorbidities play a crucial role in predicting Alzheimer's disease risk.
• Comprehensive feature assessment, including comorbidities, enhances the accuracy of early AD detection.
• Findings support the integration of these factors into clinical practice for personalized treatment planning.