Supervised Functional Principal Component Analysis Under the Mixture Cure Rate Model: An Application to Alzheimer'S...

Area of Science:

• Neuroimaging
• Biostatistics
• Medical data analysis

Background:

• Brain imaging is crucial for Alzheimer's disease (AD) risk assessment.
• Existing models often fail to account for patients with very low AD risk (cure fraction).
• Right-censored time-to-event data is common in AD progression studies.

Purpose of the Study:

• To extract image-based features for predicting time-to-event outcomes in AD.
• To develop a statistical model that incorporates a cure fraction for AD risk.
• To enhance predictive performance using novel feature extraction techniques.

Main Methods:

• Introduction of a functional mixture cure rate model extending the proportional hazards model.
• Proposal of a supervised functional principal component analysis (sFPCA) for image feature extraction.
• Accommodation of irregular brain image boundaries using bivariate splines over triangulations.

Main Results:

• The proposed sFPCA method effectively extracts AD risk-associated image features.
• The functional mixture cure rate model improves prediction accuracy by considering a cure fraction.
• The method demonstrates robustness and advantages in simulation studies.

Conclusions:

• The novel sFPCA and functional mixture cure rate model offer a powerful approach for AD risk prediction using brain imaging.
• Accounting for a cure fraction and handling censored data are critical for accurate AD prognostication.
• The method shows promise for application in real-world datasets like the Alzheimer's Disease Neuroimaging Initiative (ADNI).