This page includes supplemental materials for the paper, FBIRN Recommendations for Multi-Center fMRI Studies:
In order to maximize the available expertise in each domain, FBIRN was structured around Working Groups (WGs), each of which was tasked with particular facets of the study preparation and implementation.
Scanner and protocol issues: The Calibration Working Group had as its aims first to identify differences between MRI scanners across centers, in the functional and structural images, and second, to remove those differences a priori where possible (through protocol standardization), assess site compliance with the scanner protocol and quality assurance throughout the study, and minimize those remaining differences post hoc otherwise (e.g., potentially through correcting for BOLD variation using ASL or physiological measures). Many of their recommendations form the basis for this paper.
Clinical assessments and requirements: The Clinical Working Group was focused on developing the clinical assessment protocols and methods, including multi-center training for standardized neuropsychiatric ratings. The work of the Clinical WG is very similar to the clinical standardization goals of any clinical trial or cognitive study, with the exception that the Clinical WG had to keep in mind the demands of the fMRI session. This added a dimension of identifying what assessments needed to be done or exclusion criteria added, to ensure that the particular clinical population would be able to complete the session, while accounting for the differences in the clinical populations across centers (e.g., their relative occurrence of co-morbidities or education level, etc.). Their interactions with the other WGs form clinical recommendations for MC-fMRI studies.
Design and implementation of the cognitive task: The Cognitive Working Group had the goal of developing the cognitive task(s) to be used in the study, and piloting each variation across a handful of the centers. The Cognitive WG was a unique forum for rapid development, testing, and feedback on data from novel paradigms applied to subjects across centers. The multi-center aspects of cognitive task design, and MC-fMRI behavioral data collection recommendations are included here.
Design and evaluation of statistical approaches: The Statistical Working Group had the unusual demands of incorporating recommendations from the Calibration WG and the task data from the Cognitive WG, in order to determine the analysis models that maximally reduce inter-site variation for the study paradigm chosen. While a given MC-fMRI study may not have access to sufficient statistical expertise to have a full working group on the topic, the importance of evaluating and clarifying various statistical models for multi-center variability cannot be understated.
Data management and access: A MC-fMRI study may choose to have either a centralized data warehouse or a federated data management system for curation, storage, and analysis. The FBIRN, chose to developed a federated system where each center maintains data collected at their own institution in their own database node, while allowing immediate access to the whole group to all data from the study, across the internet. The data management and maintenance portion of an MC-fMRI study is equally important as the study hypothesis and design; many of the recommendations in this paper stem from the need to have complete, understandable, usable datasets available to all the collaborators within 24 hours of their being collected. The Neuroinformatics Working Group (NI WG) was the backbone of the multi-center data sharing and access capabilities. Members of the NI WG consisted of system administrators, database administrators, programmers and imaging analysts from the individual centers. Their ongoing, regular communication through teleconferences and email lists helped ensure that the responsibilities of each center for data curation and sharing were met.
Publication policies: With the effort and expertise that a MC-fMRI study requires, publication credit for the final analyses needs to be carefully considered. In a centralized study, the overall PI or lead center can state what the policy will be, and other centers either participate or not as they agree. In a federated study, a more representative system is required. In FBIRN, the chairs of the WGs formed the Steering Committee, which met on an ad hoc basis to coordinate solutions that required input from multiple groups. They drafted a publication policy which was commented on by FBIRN members, then set in place, evaluated while in practice, modified by the same draft and comment method, and this process continued until a version was set in place that seemed to fit the needs of the project. It includes basic guidelines regarding how members of the MC-fMRI study should develop ideas for papers, taking into account that many people who are not otherwise named had significant input into the study design and data collection. These policies need to be in place before data collection begins, but a process should be included for evaluating and modifying the policy as needed by the growing demands of the project.
