We propose to develop statistical methods for the analysis of longitudinal positron emission tomography (PET)
data for patients with Alzheimer’s disease (AD). Disease biomarkers identified from PET data are necessary for
integrating these imaging modalities in observational studies and in clinical trials for drug development for AD.
We propose statistical methods for the analysis of PET images including dimension reduction and biomarker
extraction from voxel-wise intensity analysis that incorporate disease progression in two observational studies
including data on early-onset and late-onset AD participants. Our methods allow for integration of MRI atrophy
measures in the analyses to obtain multimodal predictors of disease severity and progression. The proposed
methods will utilize the complex data and noise structure for developing powerful tools for biomarker identification
that can be used for finding differences of disease progression between populations of interest. Particularly, our
proposed biomarkers incorporate information on shape and texture of radiological images for prediction. The
proposed methods can be incorporated to analyze data in future studies in most AD data collection centers as
well as to apply in radiomics of imaging data in general.