The research team trained the algorithm by gathering images from more than a million stained tissue slides from patient biopsies. Pathologists then labeled each image so the algorithm could differentiate healthy and malignant tissue.
The algorithm then assessed 1,600 slides from 100 UPMC patients who had been suspected of having prostate cancer. According to study results published in The Lancet Digital Health, the algorithm exhibited 98 percent sensitivity and 97 percent specificity when detecting prostate cancer. It even spotted six potentially malignant slides that expert pathologists failed to flag initially.
The algorithm is the first of its kind to take its analysis beyond cancer identification, as it can also accurately categorize tumors by grade.
“Algorithms like this are especially useful in lesions that are atypical,” one of the study’s authors, Rajiv Dhir, MD, said in a news release. “A nonspecialized person may not be able to make the correct assessment. That’s a major advantage of this kind of system.”
More articles on artificial intelligence:
Physicians, data scientists propose all hospitals implement clinical departments for AI
Oxford deploys AI screening tool for rapid COVID-19 diagnosis
Healthcare leaders depend on 7 AI and analytics vendors most amid COVID-19
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.