Abstract:
|
There have been a lot of interests in using neuroimaging approaches to help guide clinicians in selecting treatment for patients of major depressive disorders and understand placebo response. In this talk, we discuss a unified Bayesian framework for modeling treatment outcome while exploiting image-based features as predictors. Traditional methods usually take two separate steps. First, a dimension reduction procedure is conducted to reduce high-dimensional images to low dimensional features. Second, a regression analysis is carried out to investigate the relationship between a treatment response of interest and the extracted low dimensional imaging features such that the effect of treatment depends on these imaging features. In contrast, our method performs these tasks simultaneously to ultimately take into account uncertainty in both steps. We also illustrate the application of the method on a large placebo-controlled depression clinical trial using baseline EEG measurements.
|