Computer taught to help doctor detect stroke lesions in Magnetic Resonance Images

Amherst 25 June 1999 A group of researchers from the Computer Vision Laboratory and the Mathematics Department at the University of Massachusetts is working together with a team of physicians, attached to the Neurology Division in the Radiology Department at Baystate Medical Center, Springfield in the Stroke Project. The common aim is to design intelligent computer vision techniques to support the clinical study of ischemic strokes treatment. The project staff is trying to teach computers to interpret magnetic resonance images (MRIs) as to determine in which way stroke patients are responding to therapy.

Advertisement

A group of researchers from the Computer Vision Laboratory and the Mathematics Department at the University of Massachusetts is working together with a team of physicians, attached to the Neurology Division in the Radiology Department at Baystate Medical Center, Springfield in the Stroke Project. The common aim is to design intelligent computer vision techniques to support the clinical study of ischemic strokes treatment. The project staff is trying to teach computers to interpret magnetic resonance images (MRIs) as to determine in which way stroke patients are responding to therapy.

Annually over 500.000 Americans suffer strokes, as stated by Dr. A. Bernard Pleet, chairman of the Neurology Division at Baystate. A stroke occurs whenever an artery delivering blood to the brain, becomes clogged or bursts. The blood supply is disrupted, causing the brain cells to die through lack of oxygen and nutrients. The effects of a stroke may be slight or severe, temporary or permanent. Depending on what portion of the brain is affected by the stroke, the damage can result in loss of important functions relating to speech, vision, or memory. The patient might even get paralysed, fall into a coma or die.

Victims suffering a stroke are submitted to Magnetic Resonance Imaging in order to show to which extent brain tissue has died. The MR images display the lesions as well as the healthy regions to the doctor, who is able to detect which injured parts probably will recover. The area in which the stroke has occurred, comprising dead and injured tissue, shows up as a bright region in the image, whereas the remaining healthy tissue is displayed as a pattern of various shades of grey along with some white. The team from the University of Massachusetts currently develops a training programme for the computer to perform spatial localization and volumetric measurement of stroke-related lesions and of surrounding penumbra, if possible.

In the end the computer has to be able to compare images taken previous to, during, and after treatment to quantitatively define the response of a patient to the therapy. This method, elaborated to perfection, would be quicker and more sensitive than the techniques, which are applied at present, according to the researchers. The team consists of six scientists: Joseph Horowitz and Donald Geman from the Department of Mathematics and Statistics; Edward Riseman, who heads the Computer Vision Lab in the University's Computer Science Department; computer science professor Gary Whitten; postdoctoral research associate Yasmina Chitti; and graduate student Ben Stein.

The distinction between dark from light in the images cannot be easily made by a computer. The visual data in fact displays various shades of grey, with blurry patches. Anything involving prior knowledge or interpretation is hard for a physician, let alone for a computer, as Dr. Horowitz explains. Indeed, the information inserted into a computer must be extremely precise in order for the machine to be able to interpret it without any ambiguity. The brain as well as the lesions actually constitute three-dimensional objects whereas the MRI is capturing "slices". Therefore, a graphical user interface is developed, allowing the doctor to circle the lesion on the slice and having the computer detect it on the slice where the physician has circled it.

Mathematics is complementary to computer imaging for several reasons, as stated by Dr. Horowitz. First, in order to give a computer instructions, they have to be expressed in mathematical terms. Also, MRI formation relies on quantum physics, according to which the protons in the body act like tiny magnets. The MRI functions through the ability to encode the behaviour of the protons mathematically, as to create images based on the mathematical statement. Statistics are required because there is an intrinsic element of randomness at the quantum level and also in the formation of the image, as Dr. Horowitz explains.

The Stroke Project eventually has to bring about that physicians can view a computer image of the brain and circle an area containing a suspected lesion with a computerized "pen". This allows the machine to focus on the region of particular concern, and break it up into lesion and non-lesion parts, without creating a massive computational burden. Dr. Pleet, who is working on the project along with Dr. Richard Hicks, chief of the division of neuroradiology at Baystate, hopes that neurologists in the future can watch the volume of damaged tissue while treating the patient with experimental drugs, to investigate which of them can promote a decrease in the damaged area.

This method is bound to be more efficient than older techniques consisting of performing periodic neurological examinations in order to try and compare them. The procedure will equally be much quicker than outcome studies in which the functional status of stroke victims can only be determined months after the event to compare it with that of other patients who did not have the treatment, according to Dr. Pleet's conviction. The applied mathematics and statistics are paying tremendous services to a problem of the highest clinical importance in order to succeed in achieving results that would not have been possible otherwise. For more technical details, you can visit the home page of the Stroke Project.


Leslie Versweyveld

[Medical IT News][Calendar][Virtual Medical Worlds Community][News on Advanced IT]