AI draws on pathologists’ wisdom to improve cancer diagnoses
Researchers at the University of Waterloo are combining artificial intelligence and pathologists’ expertise for more accurate cancer diagnoses.
That’s according to a new study of nearly 11,000 patients published in Nature Digital Medicine. The approach uses AI to match new, potentially cancerous cases with previously diagnosed and confirmed tissue samples containing the disease.
And when tested using one of the largest publicly available archives in the world, the platform was up to 100% accurate at identifying certain forms of cancer, according to Hamid Tizhoosh, director of the Laboratory for Knowledge Inference in Medical Image Analysis at the Ontario, Canada-based university.
"We showed it is possible using this approach to get incredibly encouraging results if you have access to a large archive," Tizhoosh added. "It is like putting many, many pathologists in a virtual room together and having them reach consensus."
For their study, Tizhoosh et al. had their algorithm compare pathology images of tissue samples against 30,000 digitized slides from the Cancer Genome Atlas. The database covers 25 anatomic areas of the body and 32 cancer subtypes.
During the four-month study period, the AI performed remarkably well, diagnosing everything from melanoma to prostate cancer. It was 100% and 99% accurate at diagnosing thymoma and skin cutaneous melanoma, respectively.
These results will need to be carefully scrutinized and the system refined, but the investigators see great potential for the platform as a screening tool to speed up cancer diagnoses. And it may be particularly useful in resource-strapped areas with few pathologists at the ready.
"This technology could be a blessing in places where there simply aren't enough specialists," Tizhoosh said. "One could just send an image attached to an email and get a report back."