AI can analyze images of breast masses to pinpoint cancer in low-resource settings

Artificial intelligence (AI) can analyze images of breast masses from low-cost portable ultrasound machines to pinpoint cancer, according to a study published in . Radiology, Journal of the Radiological Society of North America (RSNA). This can be useful for triage in low resource environments.

A breast lump is often discovered incidentally during a breast self-examination or by a medical professional.

Although cancer screening has received a lot of attention in Western countries, systematic screening programs and technology are often not available in low- and middle-income countries.

In low- and middle-income countries, breast cancer most commonly presents as a palpable lump in the breast. Ultrasound plays a key role in early detection, leading to more effective, less invasive treatments and improved outcomes.

Women in low- and middle-income countries often feel a potentially cancerous breast lump but do not have access to breast health care for months. We evaluated breast ultrasound images and focused on distinguishing suspicious breast masses requiring urgent attention from noncancerous ones. ”

Wendie A. Berg, MD, Ph.D., Professor of Radiology at the University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, and the lead author of the study

In this multicenter study, participants with at least one palpable breast lump were enrolled from December 2017 to May 2021 in Jalisco, Mexico. First, ultrasound images were obtained using portable ultrasound at the site of the mass and adjacent tissues. Women were then imaged with standard ultrasound. Breast Imaging Reporting and Data System (BI-RADS) evaluations were performed by radiologists.

After exclusion, 758 masses from 300 women (mean age 50.0 years) were analyzed by AI software as benign, likely benign, suspicious, or malignant (cancer). The average patient age was 50.0 years (range 18-92) and the average largest lesion diameter was 13 mm (range 2-54). Of the 758 masses, 360 (47.5%) were palpable and 56 (7.4%) were malignant.

AI correctly identified 96% and 98% of women with cancer on low-cost portable and standard-of-care ultrasound images, respectively. Of the benign masses, 67% could have been successfully triaged with standard-of-care ultrasound and 38% could have been successfully triaged with portable ultrasound.

Although specificity was inferior to standard medical devices, applying AI to portable breast ultrasound could reduce referrals to specialty hospitals in resource-limited areas by about half. .

Dr. Berg noted that the researchers did not train the AI ​​on images from portable ultrasound. She also said that low-cost portable ultrasound techniques have improved since the study was conducted, and with better images and her AI training, researchers hope to see even better results in the future. I’m here.

“Our results show great potential for using AI and ambulatory ultrasound in low-resource settings, such as remote and underserved areas of the United States, to improve breast health management. It shows,” said Dr. Berg. “By reducing the number of women with benign lumps who need to be seen at a central facility and who may be biopsied, we can better focus medical resources on women with cancer and reduce delays in diagnosis. This should improve access, health equity and outcomes for women.”