DALLAS, April 6 – A newly developed computer program predicts the chances of brain cells dying as the result of a stroke and may refine the use of brain-saving stroke drugs, according to research reported in the April issue of Stroke: Journal of the American Heart Association.
The computer software, which uses artificial intelligence techniques, very rapidly combines several new types of images obtained by magnetic resonance imaging (MRI) into a map of the brain allowing physicians to assess the risk of brain damage with high specificity and sensitivity. “That is a major accomplishment because previously it took 20 to 30 minutes to pour through all the MRI data and determine what it all meant,” says A. Gregory Sorensen, M.D., senior author of the report, associate professor of radiology at Harvard Medical School and associate director of the nuclear magnetic resonance center at Massachusetts General Hospital in Boston. “In treating stroke, every minute is crucial in limiting permanent brain damage.”
The current standard – computerized tomographic (CT) scan – uses X-rays to generate an image of the brain to determine whether a stroke was caused by bleeding of leaky or ruptured blood vessels. If the CT scan is negative for a bleeding (hemorrhagic) stroke, it is likely that it is ischemic i.e., caused by an obstructed blood vessel. Blood clots can be dissolved by tissue plasminogen activator (tPA). The drug, however, is only recommended for use within three hours of stroke onset.
“Neurologists know that, although the rule is three hours, some people given tPA at four hours get better, and some do poorly,” says Sorensen. “All neurologists struggle with the fact that they have these guidelines for groups of patients, but they are faced with treating a single patient. They want to know how they can adapt general guidelines to the specific patient in front of them.”
That is precisely what the computer program is meant to do. “Instead of having people wade through five or eight different MRI images, we simplified this into a single risk image,” he says. “This is particularly helpful now that these new types of MRI give us such large amounts of information.”
The computer breaks the advanced MRI brain scan into tiny, distinct cubes about one-tenth of an inch in diameter. Two key pieces of information are measured for each cube. One tells whether blood flow through vessels in the area is blocked. The other indicates whether the brain tissue is living or dying. Both these types of MRI scans are advanced techniques developed in the past few years.
Combining this and other data, the computer provides an estimate of the likelihood that an area of the brain will die if not treated – say zero in one place, 50 percent in another and perhaps 90 percent in still a third.
Such a map could help physicians answer important questions: Does the patient still have brain tissue at risk that ought to be treated? Is it too late for treatment? Or is there some tissue that is salvageable and some that is not?
The risk map is based on actual stroke cases and their outcomes. The researchers selected imaging and other data from 14 patients who had suffered a stroke in a major brain artery – the middle cerebral artery – and did not receive thrombolytic or neuroprotective therapy. “We actually knew what happened to these 14 patients, so we could train the program to be a good predictor,” Sorensen says. “We haven’t perfected it yet so that it is a bedside tool, but we are in the process of doing that.”
Information from additional patients has now been added to the program, and the team will add several hundred more. Although the software program has not yet been used as a guide for treating patients, the researchers have made predictive maps of people who have suffered a stroke. They have found the maps do a good job of predicting outcomes.
Currently, the program focuses on middle cerebral artery strokes. The researchers plan to expand it to encompass all strokes caused by blocked arteries.
They also see the technique as useful for testing the efficacy of new stroke drugs. Once the software’s predictive powers are proven highly accurate, Sorensen suggests that it will speed testing and reduce the number of patients needed in studies. Both could reduce the cost of developing new stroke drugs.
One thing is already evident from the early work with the predictive program, Sorensen says: People need to know and heed the symptoms of stroke. “The model tells us that there is a substantial amount of tissue that can be saved if the patient seeks medical help early.”
Co-authors are Ona Wu, M.S.; Walter J. Koroshetz, M.D.; Leif Østergaard, M.D., Ph.D.; Ferdinando S. Buonanno, M.D.; William A. Copen, M.D.; R. Gilberto Gonzalez, M.D., Ph.D.; Guy Rordorf, M.D.; Bruce R. Rosen, M.D., Ph.D.; Lee H. Schwamm, M.D.; and Robert M. Weisskoff, Ph.D.
The work was funded in part by the National Institutes of Health.
Materials provided by American Heart Association. Note: Content may be edited for style and length.
Cite This Page: