MR brain imaging is a non-invasive tool with high spatial and temporal resolution that can be used to investigate brain function, and subtle structural changes related to learning, development, treatment, or as a result of a pathophysiologic condition. Standard (commonly used) MR imaging methods include high resolution structural scanning (T1 or T2 contrast) for brain morphometry changes, functional changes (BOLD) to study the dynamic changes in the brain related to a stimulus or in the resting condition, diffusion tensor imaging (DTI) and derived fractional anisotropy (FA) maps for monitoring white matter integrity and connectivity.
Increasingly clinical trials are adopting multiple sites/centers to increase their subject pool and to expedite the studies. Data sharing is another effective approach to increase the subject pool in populations that are difficult to recruit or require large numbers, eg. The Alzheimer’s Disease Neuroimaging Initiative (ADNI), Multidisciplinary Advances in Pelvic Pain (MAPP), and 1000 Functional Connectomes Project (fcon 1000). Longitudinal studies following the progression of disease have to consider the variability induced by software and hardware upgrades to the MRI scanner. Additional complications arise when incorporating scanners from different vendors, field strengths, software levels, and coil architecture. Careful investigation of the test-retest reliability and image quality of inter- or intra- scanner neuroimaging measurements are critical in the statistical analysis and interpretation of results. Previously, such studies have focused on a single imaging methodology specific metric: e.g. Cortical thickness based on T1 (Wonderlick et al., 2009), temporal Signal-to-Noise Ratio (Parrish et al., 2000) and FA values based on DTI (Vollmar et al, 2010).
In this paper, we propose a framework for quantifying the reproducibility and image quality for multimodal neuroimaging data (structural, BOLD and DTI) collected across identical scanners and following a major hardware repair (gradient coil replacement). The uniqueness of our study is (1) we examined the three standard brain imaging methods (T1, BOLD, DTI); (2) we examined the impact of a major hardware repair (gradient coil); and (3) we tested a battery of metrics for image quality and reliability assessment applicable to local and multi-center imaging trials.