Metastatic tumor, which spreads from its site of origin to other parts of the body, can be difficult to detect. This is partly the reason why 90 percent of breast cancer deaths are the result of metastasis.
Google’s researchers, however, have been working on an artificial intelligence tool that could make detection of metastatic breast cancer easier for pathologists.
The deep learning tool known as LYmph Node Assistant (LYNA) can spot metastatic, or advanced breast cancer with impressive accuracy.
The researchers trained LYNA to recognize the characteristics of tumors using two sets of pathological slides. The trained AI system eventually became adept at telling the difference between cancer and non-cancer slides.
Researchers found that the algorithm can correctly identify a slide with metastatic cancer from one without cancer 99 percent of the time. The AI can also detect extremely small metastases that human pathologists may miss.
“LYNA was able to correctly distinguish a slide with metastatic cancer from a slide without cancer 99% of the time. Further, LYNA was able to accurately pinpoint the location of both cancers and other suspicious regions within each slide,” researchers Martin Stumpe and Craig Mermel wrote in the Google AI blog.
LYNA, however, is not perfect. It occasionally misidentified germinal cancers, giant cells, and bone marrow-derived white blood cells called histiocytes. Nonetheless, it still performed better compared with a practicing pathologist who was tasked to evaluate the same slides.
LYNA is particularly effective when used as a companion tool for pathologists. The researchers found that the pathologists who were given the deep learning tool performed better than those who did not have the tool. They also performed better compared with the tool used on its own to pick up cancerous cells.
Pathologists who performed simulated diagnoses likewise found that the AI made their work easier. LYNA did not only reduce the rate of missed micro-metastases by a factor of two, but it also slashed the inspection time in half to a single minute.
The tool essentially works like a spell-check on the computer, users can miss a misspelled word but the algorithm can catch that error.
“This represents a demonstration that people can work really well with AI algorithms than either one alone,” study researcher Yun Liu, a member of the Google AI team, told Business Insider.
The researchers described their work in two papers published in The American Journal of Surgical Pathology and Archives of Pathology and Laboratory Medicine.