Abstract:
|
In recent years, multi-center studies have become more common in order to increase subject recruitment, statistical power and generalizability. A major challenge in such studies is that there is considerable between-center variability in multi-center brain images due to the use of different scanners and processing protocols. Assessing the comparability across multi-center images provides useful information to guide calibration and quality control across centers, which is key to meaningful analysis when combining data across different sites. In this work, we present statistical methods to assess reproducibility of brain images in multi-center studies. We propose a two-stage network-based method and functional agreement indices to effectively and efficiently quantify the exchangeability of the same subject's images acquired at different centers. We develop nonparametric estimation methods and establish the statistical properties for the estimators. The proposed methods are applied to the fBIRN Traveling Subject study to investigate center effect in multi-center imaging studies.
|