62 views 3 mins 0 comments

New AI System to Help Diagnose Breast Cancer

In Health
August 17, 2019

The University of California, Los Angeles (UCLA) research team has developed a new artificial intelligence (AI) system to improve the diagnosis of breast cancer using biopsies.

Designed to interpret medical images, the new system can help pathologists classify biopsies that are usually difficult for the human eye to recognise, according to researchers. The team added that the new technology demonstrated accuracy nearly similar to that of experienced pathologists.

A study conducted by UCLA researchers in 2015 showed diagnostic errors in around one out of every six women with a type of non-invasive type of breast cancer called ductal carcinoma in situ (DCIS). The study also revealed incorrect diagnoses in approximately 50% of biopsies of breast atypia.

UCLA David Geffen School of Medicine professor Dr Joann Elmore said: “Medical images of breast biopsies contain a great deal of complex data and interpreting them can be very subjective. “Distinguishing breast atypia from ductal carcinoma in situ is important clinically but very challenging for pathologists. Sometimes, doctors do not even agree with their previous diagnosis when they are shown the same case a year later.”

The new AI system is intended to provide consistent and accurate readings by identifying patterns in a large data set of cancer samples. To train the system, researchers used 240 breast biopsy images. It was trained to detect patterns related to various breast lesions, including benign, atypia, DCIS and invasive breast cancer.

A consensus among three pathologists was then used to establish a correct diagnosis for each image. The researchers also evaluated the AI programme by comparing its readings to independent diagnoses from 87 practising pathologists in the US.

It was observed that the system performed nearly similar to human doctors in distinguishing cancer from non-cancer cases. However, the technology is said to have ‘outperformed’ doctors in distinguishing DCIS from atypia. The system demonstrated a sensitivity of 0.88-0.89, while the average sensitivity with pathologists was 0.70.

Elmore added: “There is low accuracy among practising pathologists in the US when it comes to the diagnosis of atypia and ductal carcinoma in situ, and the computer-based automated approach shows great promise.” The UCLA team is currently training the AI system to diagnose melanoma.