A Better Way to Sort fMRI Data

Text-mining software makes fMRI analysis more quantitative

2 min read
A Better Way to Sort fMRI Data

Individual research findings mean little without context. This is especially apparent in the neuroimaging field, where sample sizes are typically in the low double digits. Researchers using functional magnetic resonance imaging (fMRI) rely particularly heavily on literature reviews to interpret data and even to decide what experiments to conduct.

What many people may not realize is that most of these references to past studies are the result of manual searches on research databases such as PubMed. Tal Yarkoni, an fMRI researcher at the University of Colorado, is trying to automate this process with a computer program called NeuroSynth, which uses text-mining to find correlations between studies. The site went live earlier this summer, and Yarkoni and his colleagues at the University of Texas, Austin; the University of Warwick, Coventry, UK; and Washington University in St. Louis published a proof-of-concept study of the program in this month's Nature Methods.

Their study showed that NeuroSynth can sort through thousands of fMRI studies by the subject being studied—for example, "pain"—and the brain area associated with it—for example, "insula" (there were 1,000 fMRI studies published in 2010 alone). The program also allows researchers to search in either direction, by subject or brain region, and can calculate statistical significance, making meta-analysis easier.

The goal is to make literature review in the fMRI field more quantitative and unbiased. For example, a NeuroSynth search for "working memory" turned up connections with a frontal brain area called the dorsolateral prefrontal cortex, but that association disappeared when the researchers did the search in the other direction, starting with the brain region.

The next step for Yarkoni and his collaborators is to make the NeuroSynth software more specific and more user-friendly. "We provide the software on the site, but it's not exactly easy to use," Yarkoni says. But with nearly 5,000 studies in the NeuroSynth database so far, researchers don't have to master the software to benefit from the site.

IMAGE: Tal Yarkoni, University of Colorado

The Conversation (0)