One of the largest open-source data sets of brain MRIs from stroke patients is now available for public download via Scientific Data, a team of University of Southern California scientists reported this week.

The data set—the Anatomical Tracings of Lesion After Stroke, or ATLAS—consists of 304 manually segmented MRI scans from stroke patients, according to the study. Sook-Lei Liew, the paper’s lead author and an assistant professor at USC, said in a university release that the aim of her team’s work was to identify biomarkers to help guide personalized treatment recommendations and reduce manual labor in the lab.

“One of our goals is to meta-analyze thousands of stroke MRIs from around the world to understand how the lesions impact recovery,” Liew said. “We can’t do it by hand at the scale of thousands, so we are really interested in helping find better automated ways, using machine learning and computer vision, to identify the lesions and have machines draw those boundaries.”

Normally, neuroanatomy experts manually draw boundaries around lesions on imaging scans, which can be time-consuming. Clinicians then study those segments to design, test and implement recovery programs for stroke patients.

With ATLAS, a neuroimaging analysis pipeline helps standardize images in the data set before the information is run through custom software designed for advanced visualization. Scans are rendered into high-resolution videos and images, and segmentation is automated, allowing researchers to expedite the review process. This way, the authors wrote, radiologists can focus on studying the lesions, rather than spending time finding and highlighting them individually.

Liew and the research team have already started implementing the data set, using it to test existing predictive algorithms, but they aren’t alone. According to USC, the set has already been downloaded by 33 separate research teams from places as far-reaching as Finland, Iran and Australia, though the project’s data is stored between the Child Mind Institute and University of Michigan in the U.S.

The authors said their long-term goal is for clinicians to be able to use a system like this to inform their decisions about stroke patients’ treatment and recovery plans.  

“Ultimately, we would run their data through an automated pipeline that would give us some measures of their likelihood of recovery, or, more importantly, their likelihood of responding to different types of therapies,” Liew said. “We could then personalize their rehabilitation therapy based on their MRI results and, hopefully, improve their recovery.”