New AI Provides More Accurate Diagnosis of Breast Cancer Than Human Doctors



[ad_1]

Although pathologists usually detect breast cancer, assistance is always helpful. As such, UCLA scientists have developed a new artificial intelligence system that facilitates the reading of biopsies.

"It is essential to obtain a correct diagnosis from the beginning in order to guide patients to the most effective treatments," said Dr. Joann Elmore, lead author of the study and professor of medicine at the David Geffen School of Medicine at UCLA.

Diagnostic errors

Why would such a study be necessary? Well, because, according to a 2015 study by Elmore, pathologists often disagree about the outcome of breast biopsies. In addition, further previous research has shown that misdiagnosis has occurred in approximately one in six women with ductal carcinoma in situ and incorrect diagnoses in approximately half of the atypical biopsy cases. breast.

These are pretty important mistakes. The reason for these misinterpretations is due to the fact that it is notoriously difficult to read breast biopsies accurately.

"Medical images of breast biopsies contain a large amount of complex data and their interpretation can be very subjective," said Elmore, also a researcher at the UCLA's Jonsson Comprehensive Cancer Center. "The distinction between breast atypia and ductal carcinoma in situ is clinically important, but very difficult for pathologists, and sometimes doctors do not even agree with their previous diagnosis when they are diagnosed with it. shows the same case a year later. "

In order to find a more consistent method of diagnosing readings, the researchers pointed out that an AI could help by leveraging a large body of data. As such, they introduced 240 breast biopsy images into a computer system and trained them to recognize patterns associated with several types of breast lesions.

They then compared his findings to independent diagnoses made by 87 American forensic scientists. The program is almost as well behaved as human physicians to differentiate cancer from non-cancer cases.

Differentiate the DCIS from the atypical

However, he has eclipsed human physicians in a particularly delicate area; differentiate DCIS from atypia. This area is considered the biggest challenge in breast cancer diagnosis. The system had a sensitivity between 0.88 and 0.89, while the average sensitivity of pathologists was 0.70.

"These results are very encouraging," said Elmore. "The accuracy of US pathologists is low for the in situ diagnosis of atypia and ductal carcinoma, and the computerized approach is very promising."

The study is published in JAMA network open.

[ad_2]
Source link