28 Oct WisBusiness: Medical College of Wisconsin is using AI to improve prostate cancer analysis
Radiology researchers at the Medical College of Wisconsin are using AI to improve the process of prostate cancer analysis.
Peter LaViolette, an associate professor of radiology and biomedical engineering at MCW, says specially trained artificial intelligence software can help determine the risk of cancer recurrence in patients who’ve undergone a prostatectomy. He spoke yesterday during a webinar hosted as part of Milwaukee Tech Week.
He explained that prostate cancer is typically graded after surgery or through biopsies using the Gleason pattern scale, which predicts how likely the patient is to have aggressive recurring cancer that spreads to other parts of the body. By applying an AI system to digitized samples collected after surgery, LaViolette said his team has been able to increase the accuracy of that analysis.
“What we found in a large cohort of 100 patients is that these features, when we combine them with AI classifiers, it allows us to predict better which patients recur after prostatectomy surgery,” he said, adding that the system “predicts at over 90 percent accuracy, which is really exciting.”
Since 2014, he said his team has recruited over 270 patients to provide their medical imaging information that forms the basis of its data set. By training the AI algorithms on the patient data, the software can be used to annotate imaging slides automatically, which he said can “ease the amount of burden” on pathologists.
“It does a really good job of picking out those areas of high-grade cancer,” he said.
By combining this information with MRI data acquired from patients prior to surgery, he said additional AI models can be trained that “we can then use to predictively map prostate cancer presence on the MRI alone.”
LaViolette is the director of MCW’s Quantitative Imaging Laboratory, and his research is supported by grant funding from the National Institutes of Health’s National Cancer Institute.
See more on the work conducted by his lab: https://www.mcw.edu/departments/radiology-medical-imaging/people/peter-laviolette-phd-ms