Ductal carcinoma in situ (DCIS) is a low-risk breast cancer form that largely presents as small calcifications on mammograms. For women with DCIS, Duke researchers are working to use “machine learning” to reduce unnecessary follow-up breast imaging and offer some women less-invasive treatment options.
Assistant Professor of Radiology Lars Grimm, MD, MHS, and colleagues, led by Professor of Surgery Shelley Hwang, MD, are using artificial intelligence to train computers to screen images to identify which DCIS incidences can be monitored over time without treatment and which ones are suspicious and require more immediate intervention. Other collaborators include Associate Professor of Surgery Jeffrey Marks, PhD, and Professor of Radiology Joseph Lo, PhD.
The goal, Grimm says, isn’t to replace radiologists reading and diagnosing images, but to extract more information from the images. Greater knowledge about DCIS findings can also benefit surgeons and patients.
“We can guide surgeons and give patients more information up front without relying on tissue sampling and biopsy,” Grimm says. “We can tell women and their surgeons if we think they have DCIS or more invasive cancer so they can make more informed decisions about care.”
To date, with patients from Duke, the group has demonstrated the strategy is successful. The goal, Grimm says, is to test the strategy with larger patient populations at multiple institutions.
“Right now, the trained computer is as good as a radiologist in describing DCIS,” he says. “With more refinement of our algorithms, we want to replicate the same level of performance on a larger scale.”
Hwang says, “We are only starting to unlock the immense potential of digital imaging data, and the kind of work that Drs. Grimm and Lo are doing could be tremendously important for the over 40 million women who undergo mammography annually in the United States.”
Gauging RiskBy Whitney J. Palmer
Led by breast surgeon and Assistant Professor of Surgery Jenniffer Plichta, MD, MS, Duke’s Breast Risk Assessment Clinic helps women learn about their individual likelihood for developing the disease. Read
CIRCLE PHOTO (TOP): Shelley Hwang, MD