AI Detects Stroke, Dementia from Brain Scans

Software developed at Imperial College London (ICL) and the University of Edinburgh has helped enhance stroke and dementia diagnosis by quantifying the severity of small vessel disease (SVD), according to a study published in Radiology . "This is the first time that machine learning methods have been able to accurately measure a marker of small vessel disease in people presenting with stroke or memory impairment who undergo computed tomography (CT) scanning," says ICL's Paul Bentley. The researchers used the historical data of 1,082 CTs of people who had strokes across 70 U.K. hospitals over 14 years. The software identified and measured an SVD marker, producing a severity score. The team compared the results to a committee of expert physicians, whose estimated SVD severity matched the software's. In addition, the application of magnetic resonance imaging to estimate SVD in 60 subjects demonstrated that the software is 85 percent accurate at predicting severity.