New research papers involving the application of artificial intelligence (AI) to disease diagnosis and treatment come out seemingly every day. However, as G2 Intelligence has previously reported, such tools are still a ways out from actual clinical use in pathology labs. But researchers continue to make important steps toward bringing AI solutions to patients.
In a recent example in the cancer diagnostics space, researchers showed what they claim to be the first “routine clinical use of an AI-based tool for breast cancer biopsies in a pathology lab,” according to an article by Miriam Bergeret, MSc, in our February 2023 issue of Diagnostic Testing & Emerging Technologies.
As described in a paper published in the journal Breast Cancer in early December 2022, the researchers trained their quality control AI algorithm to detect various signs of invasive breast cancer using data from roughly 2,000 H&E slides from nine pathology labs that had been carefully curated from a larger pool of 115,000 slides.
During testing in an actual clinical setting, the AI tool performed well, catching missed cases of DCIS/ADH and invasive breast cancer. In particular, the algorithm:
- Showed about 98 percent sensitivity and 96 percent specificity in detecting invasive carcinoma
- Showed about 92 percent sensitivity and 92 percent specificity in detecting DCIS/ADH
Read the full article in our February 2023 issue of Diagnostic Testing & Emerging Technologies. Sign in with your DTET subscription, or sign up for a free trial to gain access.