A Residual Bootstrap Method for Evaluating Preprocessing Pipelines in Functional Magnetic Resonance Imaging (fMRI)

• Activity Number: 405
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Xiangxiang Meng*+ and Scott K. Holland and Xiaodong Lin
• Companies: University of Cincinnati and Cincinnati Children's Hospital Medical Center and Rutgers University
• Keywords: functional magnetic resonance imaging ; bootstrap ; motion correciton ; general linear model ; image reliability

Abstract

Motion is the major issue that affects the result of the statistical analysis in functional magnetic resonance imaging (fMRI). In this paper, we develop a GLM-based image reliability measure using residual resampling, and evaluate a variety of fMRI preprocessing pipelines for motion correction. Unlike usual cross-validation techniques for imaging reliability, this approach is based on bootstrapping the residuals in the GLM analysis of single-subject fMRI data which does not require homogeneity and independence assumption across subjects. Thus, our approach can be applied to a variety of experimental setting such as longitudinal and case-control. We demonstrate the performance of the proposed image reliability measure using fMRI data corrupted with different levels of motion artifact, and then apply it to evaluate motion correction schemes such as realignment to the first fMRI scan, realignment to the best fMRI scan, and the use of motion parameters as regressors in the GLM analysis.