June 25, 2021
The University of Alberta is using an artificial intelligence technique developed in part by one of the Iranian students killed 17 months ago aboard a Ukrainian jet.
A team of computer scientists created a new deep-learning model a type of machine learning that uses algorithms to find patterns, then make models and predictions that helps a computer learn to identify diseases on medical scans faster and more accurately. Pouneh Gorji, who was 25 when the Ukraine International Airlines flight was shot down by Iranian missiles in January 2020, helped redesign the algorithm through her research on identifying fatty liver through AI.
Roberto Vega, the lead author of the new study, told the Edmonton Journal the model could be used to help radiologists make better diagnoses instead of spending hours reviewing scans manually.
“We were devising a new training method that allows [the computer] to make very good predictions but with only a few hundred images that is what [doctors] usually have available,” he said.
Radiologists don’t always have a lot of scans to work with for privacy reasons, he said, and it’s also a laborious process. Some hospitals particularly in rural areas with less staff, or countries with strained health-care systems may not have the resources to do a comprehensive review.
“By providing automated ways of analyzing the images, we can help them to deliver better health care,” he said.
A computer would normally need to analyze tens of thousands of images to learn what a disease looks like, but the new tool shortens that process and can identify a fatty liver, glaucoma, and hip dysplasia with only a few hundred images.
He believes their research can be used to help with other medical diagnoses in the future.
Gorji, a posthumous graduate of the department of computing science, is listed as the second author of the study because of her significant contribution.
Vega said it was important she be recognized because her work was crucial and wouldn’t have happened without her. “[Publishing] a paper is always a great accomplishment. It was especially important for me because she was my best friend…. It’s the only tribute I can give to her,” he said.
Vega thought he had finished the model during his research analyzing scans of hip dysplasia. But when he and Gorji tried it on her research into fatty liver scans it didn’t work.
They worked together and solved the problem.
“We spent a lot of time in the lab trying to figure out what was wrong, trying to generate solutions, and we ended up proposing a new mathematical model that worked well for hip dysplasia, and also for fatty liver,” he said. “And then we decided to try it in glaucoma … and then we found that it worked.”
Gorji and her husband, Arash Pourzarabi, married in Iran days before they died. They were among the 176 people who died aboard Flight PS752. Ten members of the University of Alberta community perished in the crash.