AI shows promise for better outcomes at Michigan Medicine

Researchers use human body comparative data to identify potentially fatal conditions and predict serious events. The approach eventually could be more widely used.


High-resolution computerized tomography is being used to diagnose a patient’s risk for a heart attack or other cardiac concern before a dangerous incident occurs.

Researchers at Michigan Medicine are using the imaging technology to measure growth in the largest artery in the human body, seeking to use that comparative information to identify potentially fatal conditions.

The technique is called vascular deformation mapping, and it uses CT scans to measure changes in the thoracic aorta, which carries blood from the heart to the rest of the body. The technique uses high-resolution CT, which combines a series of images from different angles around the body to create a three-dimensional representation of the body.

Using the high-resolution CT imaging is believed to significantly outperform standard manual rating methods of cardiac-connected arteries, according to results published in professional journals, Radiology and Medical Physics.

Michigan Medicine researchers have developed an algorithm that can visualize the three-dimensional growth of an aneurysm in the thoracic aorta, said Nicholas Burris, MD, assistant professor of radiology and director of aortic imaging at the wholly owned academic medical center of the University of Michigan in Ann Arbor.

Burris suggests that the combination of the algorithm and the imaging data vastly improves clinicians’ ability to measure changes and offer an accurate picture of a patient’s condition.

Aneurysms in the thoracic aorta occur when the largest part of the aorta becomes weakened and dilates increasing the risk of a rupture. About 3 percent of adults older than 50 have enlarged aortas and are recommended to undergo regular testing, often with CT scans, to assess aortic growth. But it’s hard to measure changes in the aorta because they are relatively small, and standard measurement methods can’t accurately detect it.

The Michigan Medicine technique uses an image analysis technique known as image registration, which aligns pixels in multiple CT scans to develop a three-dimensional “map” that accurately shows small variations of growth in potential aneurysms.

Tests of the algorithm assessing the registered CT images were able to achieve accuracy of less than one millimeter, compared with manual, human-based error rates of 3 millimeters.

The new technique is still being tested, but can be facilitated by routine CT scans of the aorta.

Researchers say this application of quantitative imaging and artificial intelligence can improve clinical care “by empowering clinicians rather than attempting to replace them,” says Charles Hatt, adjunct research assistant professor of radiology at Michigan Medicine.

The approach also could be used to measure variations in other diseases, and the possibility of using three-dimensional measurements to gain accuracy over current use of measuring off of one-dimensional images.

This story is based on information from a story on the University of Michigan Health Lab blog.

More for you

Loading data for hdm_tax_topic #better-outcomes...