Abstract:
|
Alzheimer's disease (AD), the most common forms of dementia, is an irreversible age-related condition resulting in an increase in dependency on care providers for basic functioning. A proper diagnosis of sporadic AD suffers from the lack of diagnostic tools that can accurately distinguish AD from other forms of dementia at an early stage of the disease. If validated, non-invasive biomarkers such as Magnetic Resonance Imaging (MRI) variables hold the most premise for timely diagnosis and management of the disease. We propose a two-stage model for validation of MRI parameters as biomarkers for a specific histology staining in selected brain regions of interest. In the first stage, a bivariate joint distribution for histology and MRI endpoints is specified to estimate the disease progression effect on both endpoint. Subsequently, a linear regression model of the disease effects is used to estimate the disease-level surrogacy using the model coefficient of determination. Results indicate that the surrogacy of MRI variables for a given histology staining depends highly on the MRI variables being evaluated and the brain region of interest.
|