Artificial Intelligence: An AI Program Identifies Prostate Cancer With Almost 100% Accuracy
The AI was trained on images from more than a million parts of stained tissue slides taken from patient biopsies.
Researchers at UPMC and the University of Pittsburgh have trained an artificial intelligence (AI) algorithm to recognize and characterize prostate cancer with near-perfect accuracy. (Medical Xpress)
How they did it
The researchers labeled more than a million parts of stained tissue slides taken from patient biopsies and fed the images into the AI. The label notations trained the program to distinguish between healthy and cancerous tissue.
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 and his research colleagues then tested the algorithm on 1,600 slides taken from 100 patients under investigation at UPMC for prostate cancer.
The program showed 8% sensitivity and 97% specificity at detecting prostate cancer, readings that were much higher than previous AI studies.
The program also went a step further and reported on tumor grading, sizing, and nerve invasion.
It also identified cancer cases from six slides that escaped the attention of highly trained pathologists.
Dhir said that though humans were adept at recognizing anomalies, they suffer from their own biases and previous impressions.
“Machines are detached from the whole story,” says Dhir. “There’s definitely an element of standardizing care.”
He also said their algorithm was effective in assessing atypical lesions – the kind that would escape detection by a non-specialized person.
He also said the technology could adapt for use in other kinds of cancers such as breast cancer.
AI in brain cancer detection
In January, DailyAlts.com wrote about an AI algorithm for brain cancer developed by Daniel A. Orringer, associate professor of Neurosurgery at New York University Grossman School of Medicine, along with colleagues at the University of Michigan.
Surgeons take the brain tissue sample and send it for analysis while the patient is still on the operating table. Researchers then obtain images of the tissue using an optical imaging technique called stimulated Raman histology (SRH). An AI algorithm scans these images and delivers a diagnosis of the presence of a brain tumor within 150 seconds.
The AI-based diagnosis was 94.6% accurate versus the 93.9% accuracy from the standard lab-based procedure but took far less time.
Latest Alternative Investment News
COVID-19 is sparking a critical decline of overseas buyer interest. Foreign buyers have historically been a large part of the United States real estate market. Lately, however, the size and…
Private Equity Meets Both Privacy Concerns and Major Growth in Next-Generation Industry. The PE space is buzzing over private equity giant Blackstone’s acquisition of DNA testing company Ancestry. The firm…
Ares Leapfrogs Competitors Through Pandemic. Investment firm giant Ares has raised $5 billion for its private equity fund in the second quarter. The firm’s goal is to raise up to…
The Hoya Capital Housing ETF (NYSE: HOMZ) announced Wednesday a cut in its expense ratio from 0.45% to 0.30%, effective from August 1, 2020. It claimed that it has the…