Researchers plan to use artificial intelligence to improve health care

UI researchers work to develop a tool to enhance the practices of health-care providers and help give fast and accurate diagnoses.

Elena Alvarez

UI Health Care researcher Zhihui Guo poses for a portrait at the Pappajohn Biomedical Institute on Tuesday, February 26, 2019. Guo has been researching the uses of AI for healthcare purposes for the past four years.

Alexandra Skores, News Reporter

Researchers at the University of Iowa are working on developing algorithms in artificial intelligence to learn from sample data and ultimately attempt to predict disease outcomes.

Graduate research assistant Zhihui Guo, who has worked with the team for two years, is very interested in the potential it has for health care.

“By building such a model, it will tell you the result,” Guo said. “This is a software where you place it into a machine, and we can teach it to do what we want it to. It’s basically like robotics.”

Traditionally, health-care professionals use medical imaging techniques such as CT and MR scanning to produce multi-dimensional images that enable radiologists to look into the human body noninvasively, Guo said.

However, the researchers hope to alleviate the tedious and time-consuming manual quantitative analysis of medical imaging techniques such as CT and MR scans, in which radiologists must look through the images slice by slice, she said.

The team has worked on the project for around six years.

RELATED: AI-based cancer treatment platform helps UI doctors

The researchers have placed a focus on the research of glaucoma patients, Guo said. They hope to expand the technology to be able to research other disease outcomes.

“We look at each layer of the retina,” she said. “We look at the thickness of the layers through these scans. Right now, we have pretty good accuracy, but we are working to improve it.”

UI Professor of ophthalmology Young Kwon said in an email to The Daily Iowan that the project has explored ways to predict visual functional assessment of glaucoma patients directly from optic-nerve scans.

“The biggest benefit is for the patient, who does not have to undergo long, difficult visual function tests and instead gets a quick and easy optic-nerve scan,” said Kwon’s email. “The biggest benefit for the doctor is that the visual function information is derived from objective optic-nerve scans rather than subjective visual-function tests which are often noisy and unreliable.”

Kwon also noted that the AI technology can be used in the future to diagnose glaucoma earlier and detect disease progression more accurately.

Milan Sonka, a co-director of the Iowa Institute for Biomedical Imaging and the associate dean for graduate programs in the College of Engineering, said the current testing used on patients is not well-liked.

“Current visual-function tests are done by faint lights blinking,” Sonka said. “The test takes about 20 minutes and requires mental concentration on the task. This is especially difficult for older patients with impaired sight. OCT imaging takes about a minute and does not require any special attention of the imaged subject.”

Another area of focus, Guo said, is on the detection of pancreatic tumors.

“We want to detect tumors with the data early on,” she said. “When doctors look at images, they may miss them altogether. We want to help them develop a tool to detect these tumors. It is quite hard to develop such a tool. We face a lot of challenges within our research, but our accuracy is pretty high.”