You are here

CARE Consortium: MRI stability

Researchers examine how advanced imaging techniques compare at different sites

At its inception, the Concussion Assessment, Research and Education Consortium study included four Advanced Research Core testing sites — UCLA, North Carolina, Virginia Tech and Wisconsin — where magnetic resonance imaging is used for cutting-edge imaging techniques that examine the brains of college athletes before and after injury. Given that the data is being gathered on different MRI scanners at different sites, are comparisons across different sites viable? Can this massive data set that is being collected at imaging sites across the country be considered reliable for further research, or does the potential for increased variance across the sites outweigh any benefits of casting a wider net? Andrew Nencka, associate director of the Medical College of Wisconsin’s Center for Imaging Research, led a team that sought to find out in a study published in October 2017 in the academic journal Brain Imaging and Behavior

What did the study find?

The study used the same subjects at each of the Advanced Research Core imaging sites and found that there was little variance in the results across test sites. Even though the machines are in different locations and provided by different vendors, the data produced by each machine was comparable with that yielded by its counterparts at other sites. Results for individuals did not vary significantly from site to site.

What are the implications?

These results open the door for researchers — those involved in the CARE study or other significant projects — to feel confident about relying on multiple imaging testing sites, which will enable them to compile more robust data sets that can far expand a study’s reach and impact.

“For the larger community of neuroimaging, this study adds to the growing consensus that carefully designed multisite imaging studies, even including different imaging system vendors, are well positioned to greatly increase subject recruitment with negligible negative impact to the statistical power of the study,” Nencka said. “As the field continues its push into big data and artificial intelligence, this insight offers a great opportunity to design further studies probing the nuances of neurocognitive science with massive imaging data sets.”

What’s next?

The study also gave Nencka and his team a better sense of which imaging modalities are most effective in evaluating different sections of the brain. Multiple modalities, each targeting a different effect of concussion, provide the most comprehensive indication of changes in brain structure and function after concussive injury.

“We have a baseline for evaluating potential changes in these metrics in athletes exposed to head impacts in contact sports, as well as in athletes recovering from concussion,” Nencka said. “The next steps will be to develop multidimensional models, which should offer greater statistical power for diagnosis and prognosis of brain injury on a more individual level.”