Reference: 27 July 2020, The Lancet Digital Health.
A research study released today (July 27, 2020) in The Lancet Digital Health by UPMC and University of Pittsburgh scientists demonstrates the highest precision to date in characterizing and acknowledging prostate cancer utilizing a synthetic intelligence (AI) program.
Prostate biopsy with cancer possibility (blue is low, red is high). This case was originally identified as benign but altered to cancer upon additional review. The AI properly discovered cancer in this challenging case. Credit: Ibex Medical Analytics
” Algorithms like this are specifically helpful in sores that are irregular,” Dhir said. “A nonspecialized individual might not have the ability to make the right evaluation. Thats a major advantage of this type of system.”
AI also flagged six slides that were not kept in mind by the expert pathologists.
While these outcomes are promising, Dhir warns that brand-new algorithms will have to be trained to detect different types of cancer.
###
Prostate biopsy with cancer likelihood (blue is low, red is high). The AI properly spotted cancer in this tricky case. To train the AI to recognize prostate cancer, Dhir and his colleagues provided images from more than a million parts of stained tissue slides taken from client biopsies. The algorithm was then evaluated on a different set of 1,600 slides taken from 100 successive patients seen at UPMC for believed prostate cancer.
” Humans are excellent at acknowledging abnormalities, however they have their own biases or previous experience,” said senior author Rajiv Dhir, M.D., M.B.A., chief pathologist and vice chair of pathology at UPMC Shadyside and professor of biomedical informatics at Pitt. “Machines are separated from the entire story. Theres certainly an aspect of standardizing care.”
To train the AI to recognize prostate cancer, Dhir and his colleagues offered images from more than a million parts of stained tissue slides taken from patient biopsies. Each image was identified by specialist pathologists to teach the AI how to discriminate in between abnormal and healthy tissue. The algorithm was then tested on a different set of 1,600 slides taken from 100 successive patients seen at UPMC for suspected prostate cancer.
Also, this is the very first algorithm to extend beyond cancer detection, reporting high performance for tumor grading, sizing, and invasion of the surrounding nerves. These all are medically important features needed as part of the pathology report.
However Dhir described that this does not always indicate that the machine is superior to human beings. For instance, in the course of evaluating these cases, the pathologist might have simply seen adequate evidence of malignancy somewhere else because clients samples to suggest treatment. For less experienced pathologists, however, the algorithm could act as a failsafe to catch cases that may otherwise be missed.
Funding for this research study was provided by Ibex, which also created this commercially offered algorithm. Pantanowitz, Shalev and Albrecht-Shach report fees paid by Ibex, and Pantanowitz and Shalev serve on the medical advisory board. Bien and Linhart are authors on pending patents US 62/743,559 and US 62/981,925. Ibex had no impact over the style of the research study or the interpretation of the outcomes.
Throughout screening, the AI showed 98% sensitivity and 97% specificity at discovering prostate cancer– considerably greater than formerly reported for algorithms working from tissue slides.
Additional authors on the study consist of Liron Pantanowitz, M.B.B.Ch., of the University of Michigan; Gabriela Quiroga-Garza, M.D., of UPMC; Lilach Bien, Ronen Heled, Daphna Laifenfeld, Ph.D., Chaim Linhart, Judith Sandbank, M.D., Manuela Vecsler, of Ibex Medical Analytics; Anat Albrecht-Shach, M.D., of Shamir Medical Center; Varda Shalev, M.D., M.P.A., of Maccabbi Healthcare Services; and Pamela Michelow, M.S., and Scott Hazelhurst, Ph.D., of the University of the Witwatersrand.
While these outcomes are appealing, Dhir cautions that new algorithms will need to be trained to spot various kinds of cancer. The pathology markers arent universal throughout all tissue types. He didnt see why that could not be done to adjust this innovation to work with breast cancer.
