The San Diego-based biotechnology company CureMetrix has a product coming to market that may shake up the computer-aided detection software field.
The cmAssist software leverages artificial intelligence to help radiologists distinguish cancer from benign breast lesions.
"We don't want to dictate what the doctor does," Kevin Harris, CEO of CureMetrix, told HCB News. "All we are doing is providing a means by which they can sort the images [so they] might be more patient-friendly and clinic-friendly."
The software anonymizes a facility's mammogram images and processes them on its cloud-based platform. It then returns an overlay file to the radiologists that helps them understand what they are seeing, and also spots things they might have missed.
"There is an opportunity to learn, the same way that a doctor learns," said Harris. "The first time a doctor looks at their very first mammogram, they don't know what they are looking at, but by the time a doctor has read hundreds of thousands of mammograms, they're an expert. Machine learning works in a very similar fashion."
The CureMetrix staff used examples of cancer and benign biopsies to teach the cmAssist software. Over time, the algorithm learned to recognize what it's seeing and can now categorize the mammograms into three buckets — really serious, really normal or somewhere in the middle.
"Our software is quantitative in nature — we provide a score for every anomaly," said Harris. "We can use those scores to work with the doctors to adjust their workflow."
cmAssist is commercializing globally in places such as Mexico, while working on FDA approval in the U.S.
