Machine learning tool aids radiologists in IDing cancerous breast lesions

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - breast cancer CAD
Dynamic contrast-enhanced MR imaging in a 47-year-old woman with DCIS appearing as a nonmass lesion in the left breast. Inset is a magnified view of segmented lesion.
Source: Radiology (DOI: http://dx.doi.org/10.1148/radiol.2015150241)

Researchers at the Massachusetts Institute of Technology (MIT)'s Computer Science and Artificial Intelligence Laboratory and breast imaging experts at Massachusetts General Hospital (MGH) have developed a machine learning tool capable of identifying high-risk breast lesions that are likely to become cancerous. The tool aims to reduce the amount of unnecessary surgeries in patients with low-risk lesions.

High-risk breast cancer lesions are usually surgically removed to prevent the progression of cancerous tissue. But many lesions do not pose an immediate threat to the patient and unnecessary surgery could negatively affect the patient. In this study, explained in Radiology, researchers utilized machine learning to identify lesions that do not require surgery to save costs and surgery complications.

"There are different types of high-risk lesions," said study author and radiologist Manisha Bahl, MD, MPH, from MGH and Harvard Medical School, both in Boston. "Most institutions recommend surgical excision for high-risk lesions such as atypical ductal hyperplasia, for which the risk of upgrade to cancer is about 20 percent. For other types of high-risk lesions, the risk of upgrade varies quite a bit in the literature, and patient management, including the decision about whether to remove or survey the lesion, varies across practices."

Researchers trained the machine learning tool with traditional risk factors, including patient age and lesion history, as well as with unique factors like words appearing in biopsy pathology reports. The tool was trained with 1,006 high-risk lesion cases, 115 or which progressed to cancer. In testing the accuracy of the tool after training, researchers evaluated the tool on 335 lesions—and the tool was able to identify 37 or the 38 cancerous lesions.

"Our study provides 'proof of concept' that machine learning can not only decrease unnecessary surgery by nearly one-third in this specific patient population, but also can support more targeted, personalized approaches to patient care," said the paper's senior author, Constance Lehman, MD, PhD, professor at Harvard Medical School and Director of Breast Imaging at MGH.