Linda Chang

Fengling Hu

Keywords: Neuro: Brain, Image acquisition: Image processing, Image acquisition: Modelling

This session will cover techniques for combining neuroimaging data across multiple sites in order to increase statistical power and improve generalizability. The session will first distinguish between standardization of neuroimaging acquisition protocols, pre-processing techniques for mitigating specific types of technical variation, and finally post-processing image harmonization to reduce inter-site artifacts. The primary focus will be on image harmonization, including key statistical concepts and applied methodology. Attendees will gain insight into commonly used harmonization methods, as well as their assumptions and limitations, equipping them with practical knowledge to improve the quality and reproducibility of multi-site neuroimaging data analyses.

Russell Shinohara

Keywords: Transferable skills: Statistics

Multi-center studies pose opportunities for enriching study samples and promoting generalizability in scientific research. Imaging studies that leverage multi-center designs face unique challenges in study design and analysis methodology. This presentation will review the issue of inter-scanner and inter-protocol differences. We will then discuss the field of statistical image harmonization,  how this impacts design decisions, and implications for power and reproducibility.

David Mummy

Kalina Jordanova

Keywords: Transferable skills: Reproducible research, Image acquisition: Quantification, Transferable skills: Research coordination

Motivation: Multi-center imaging studies aid in the translation of MRI applications into clinical practice; however, measurement variation can confound results, especially for quantitative methods. Phantoms can be used to validate multi-center imaging results and increase study robustness to system changes.

Goal(s): To illustrate how to effectively use phantoms in multi-center imaging trials.

Approach: Discuss the design considerations to build and utilize phantoms for multi-center studies.

Results: Phantoms can help identify measurement variations due to system changes; thus, the impact of biological sources on measurement variations can be better understood.

James Goodman

Keywords: Education Committee: Clinical MRI, Transferable skills: Research coordination

Multi-center research is an essential component of widespread adoption of any new technique or widespread acceptance of significant new findings that inform disease intervention or management.  This overview of multi-site clinical trials, a sub-set of clinical research, is intended to provide a high-level introduction to various considerations and components that are necessary (though not sufficient!) for a successfully executed clinical trial.  Topics discussed will range from types of trials, endpoints, site selection, and consent, through study maintenance, analysis, closeout, and reporting.  The presentations that follow this broad introduction will treat many of these topics in greater detail.

Julien Savatovsky