New AI developed to identify prostate cancer with ‘near-perfect accuracy’

PITTSBURGH – Human error can be fascinating as an attractive type, but no one appreciates mistakes on a serious subject like cancer. On that note, researchers at the University of Pittsburgh developed a new artificial intelligence program with the most accurate record to date when it came to identifying prostate cancer.

Senior writer Drs. “Humans are good at identifying anomalies, but have their own experience or past experience,” says Rajeev Dhir, pathologist and vice chairman of pathology and vice-president of UPMC Shadyside and professor of biomedical informatics. “Machines are cut off from the whole story. Standardization is certainly an element of care. “

What makes this AI different from the rest of the robot pack? Dr. Dhir and his team “fed” images of their program from over one million pieces of tissue slides extracted from a prostate cancer patient biopsy. Then, the AI ​​program was tested on 1,600 different slide images collected from 100 suspected prostate cancer patients.

AI performed incredibly well on that test. Results show 98% sensitivity and 97% specificity in finding and detecting prostate cancer. Those figures are much higher than those recorded by previous cancer detection algorithms.

Furthermore, it is the first algorithm that does much more than detect cancer. The program also scores well in categories including tumor grading and sizing and assessment of surrounding nerve invasion by cancer cells.

Click here to SUBSCRIBE our newspaper and get the latest study from STUDYFINDS.ORG by EMAIL!

Even AI detects cancer in six slides that were tripped by a human pathologist.

While all of this is very promising, the study’s authors warn that the AI ​​program is not yet quite ready to completely replace human doctors. For example, about the six slides that went unnoticed by a human doctor, that pathologist must have seen enough evidence to diagnose cancer. before this Coming on those special slides.

Nevertheless, Drs. Dhir states that the algorithm can, at the very least, serve as a great failure.

“Such algorithms are particularly useful in lesions that are atypical,” comments Dr. Dhir. “An autocratic person may not be able to make the correct assessment. This is a major benefit of such a system. “

Of course, this project focuses only on prostate cancer. Although a completely new algorithm must be trained for each type of cancer, the research team is optimistic that their results can be tailored to other cancer variations.

The study is published in Lancet Digital Health.

Like studying? follow us on Facebook!

Leave a Reply

Your email address will not be published.