Although pathologists generally do a good job of looking at breast cancer, there is no doubt that help is always helpful. In this way, UCLA scientists .The scientists have developed an innovative artificial intelligence system that facilitates biopsy readings.
Senior author of the study and professor of medicine at the David Geffen School of Medicine at UCLA. "It is critical to get the right diagnosis right from the beginning so that we can guide patients to the most effective treatment," said Jonan Elmore.
Why would such a study be needed? Well, because, according to a 2015 study led by Elmore, pathologists often disagree on the outcome of breast biopsy. Moreover, previous research has shown that diagnosis errors occurred in about half of every six women who had ductal carcinoma in situ and misdiagnosed in approximately half of cases of biopsy of breast women.
These are just some of the notable mistakes. The reason for this misinterpretation is that breast biopsies are notoriously difficult to read accurately.
"The medical images of breast biopsies contain very complex data and can be very subjective to interpret," said Elmore, who is also a researcher at the UCLA Johnson Comprehensive Cancer Center. "Separating breast atypia from ductal carcinoma in situ is clinically important but very challenging for pathologists. Sometimes, doctors don't even agree with a previous diagnosis when they are shown a similar case after one year."
To find a more consistent method of diagnosing readings, the researchers decided that drawing from a larger data set could help AI. As such, they fed 240 breast biopsy images to a computer system and trained them to identify patterns associated with many types of breast lesions.
Then they compared its results to the US. The program introduced human doctors to differentiate cancer from non-cancerous cases.
Isolation of DCIS from Ethiopia
However, it overtook human doctors in a particularly difficult area; Isolation of DCIS from Ethiopia. This area is considered one of the biggest challenges in breast cancer diagnosis. The system showed sensitivity between 0.88 and 0.89, while the pathologist had an average sensitivity of 0.70.
"These results are very encouraging," Elmore said. "When it comes to the diagnosis of atypia and ductal carcinoma in the US, pathologists in the US have less accuracy in practice, and a computer-based automated approach shows great promise."
The study is published in JAMA Network open.