A newly developed AI-based system can quickly assess MRI exams of women with dense breasts and eliminate those without cancer, thus allowing radiologists to focus on more difficult cases.

Developed by scientists from the Image Sciences Institute at the University Medical Center Utrecht in the Netherlands, the system used breast MRI data from earlier research to develop and train the deep learning model to distinguish between breasts with and without lesions. The model was trained on data from seven hospitals and tested on data from an eighth hospital.

Tested on more than 9,000 dense breasts, the AI triaged about 90% for radiologists’ review due to abnormal lesions. Nearly 40%, meanwhile, did not have lesions and were later confirmed cancer-free.

“We showed that it is possible to safely use artificial intelligence to dismiss breast screening MRIs without missing any malignant disease,” team leader Erik Verburg, MSc, noted in an announcement. “The results were better than expected. Forty percent is a good start. However, we have still 60% to improve.”

According to experts, women with extremely dense breasts have a three- to six-times higher risk of developing breast cancer than women with almost entirely fatty breasts and a twofold higher risk than the average woman.

To create their tool, Verburg and team used 4,500 MRI datasets from an earlier Dutch study underscoring the benefits of supplemental MRI screening for women with dense breasts. 

The AI-based triaging system has the potential to significantly reduce radiologist workload, Verburg said. In the Netherlands alone, nearly 82,000 women may be eligible for biennial MRI breast screening based on breast density.

In their report, published at Radiology, the team said the aim of the research was to determine the feasibility of automated triaging using deep learning (DL) based on screening breast MRI to reduce the workload and to prioritize the work of breast MRI radiologists by dismissing the largest number of MRI examinations without lesions while still identifying all scans with malignant disease.

“The approach can first be used to assist radiologists to reduce overall reading time,” Verburg said. “Consequently, more time could become available to focus on the really complex breast MRI examinations.”

The researchers plan to validate the model using other datasets.

Photo by Zapp2Photo/Getty Images