Researchers at the University of Southern California have released a database of brain scans. This open source approach will dramatically improve the global research communities access to a database that can now make use of machine learning to drive discovery.
Most notable about the endeavor is that it promotes international collaboration by providing open source data on 304 manually assessed brain scans of stroke patients. The publication of the database known as Anatomical Tracings of Lesion After Stroke (ATLAS) has opened a new path to understanding where and how these lesions occur might and for providing better treatment to those who suffer from the damage. Globally, strokes are notorious for being the leading cause of adult disability. The symptoms of a stroke are severe to the point that a person can have total loss of memory and the loss of motor skills. Normally manual assessment of brain scans must identify where lesions occur post stroke. Lesions are the areas of the brain that have lost brain cells from the sustained lack of oxygen caused by strokes.
When it comes to neuroimaging and identifying robust biomarkers, the more data that is collected the more accurate medical professionals become in identifying biomarkers. Special algorithms aid in automating the process; however, they rely highly on manually created data done by anatomical experts. The machine learning implemented in this technology requires large data sets to train and optimize its performance.
The standardization required for compiling the data has made data more accessible. This is quickly changing from the use of different custom softwares developed for different hospitals or research centers. Tyler Ard, assistant professor of research at the university has created their own custom software. The data was then used to create visual high resolution images and videos. The collaborative work of seventeen other co-authors from the university included collecting and storing the data.
Sook-Lei Liew, the paper’s lead author, shared his view surrounding the motivations and impact of this sharing technology.
“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.”
According to university administrators, the data set has already been downloaded more than 30 times by separate research teams around the world. The projects data is stored between the Child Mind Institute and the University of Michigan. However it has been accessed by researchers in countries such as Finland, Iran and Australia.
The project was revealed at the annual meeting of the American Society for Neurorehabilitation in November and then had the paper submitted for review. In order to obtain the data anyone can complete a Google form. The form will allow for applicants to submit their information and have a encryption code emailed to them giving access to the database.