Radiologists are like detectives, says Linda Moy, MD, a radiologist at NYU Langone Health’s Perlmutter Cancer Center, using imaging technology to search for clues inside the body and figure out why someone is not feeling well. A breast imaging specialist, Dr. Moy’s latest research focuses on using artificial intelligence (AI) and machine learning to make MRI for breast cancer more accurate by decreasing the number of false positives to reduce the need for unnecessary biopsies or follow-up exams.

“What is unique about MRI is that, in general, it is done for patients who are at high risk for breast cancer because of family or personal history,” says Dr. Moy, who is also a professor in the Department of Radiology at NYU Grossman School of Medicine. “This is a group of people who are very worried and who have very intensive imaging surveillance, including mammograms and ultrasounds. If we can determine what is the optimal surveillance that they need, I think it will make a big improvement in their quality of life.”

In a paper published September 28, 2022, in Science Translational Medicine, Dr. Moy and colleagues at NYU Grossman School of Medicine, including co-corresponding authors Krzysztof J. Geras, PhD, an assistant professor in the Department of Radiology and a member of Perlmutter Cancer Center, and Jan Witowski, MD, PhD, a postdoctoral fellow working with Dr. Geras, described an AI system for detecting breast cancer in dynamic contrast-enhanced MRI. The researchers found that their machine learning model, if used to assist radiologists, has the potential to significantly reduce unnecessary biopsy referrals.

Radiologists are trained in pattern recognition, reviewing imaging scans to determine a mass’s likelihood of being benign or malignant based on its size and shape, and how it interacts with adjacent normal tissue or how it picks up contrast dye. AI systems, Dr. Moy says, can “see” features that are imperceptible to the human eye, and because of that, they can then see many more features than what the human eye can appreciate.

Before AI systems can be used widely to help radiologists decide whether a suspicious spot on an MRI exam needs further investigation, however, both radiologists and patients need to trust the technology, Dr. Moy says.

“As radiologists, we need to test the system across a wide variety of patients and evaluate mammograms, ultrasound, and MRIs done with equipment by a variety of manufacturers to feel confident that the false-negative, or miss, rate is very low,” Dr. Moy says. “In order to trust an AI system, we have to understand how it made its decision, which is called explainable AI, and also ensure that we have done our due diligence in minimizing bias in the system by making sure that it works with people of all races.”

Building trust with the people who come to see Dr. Moy for imaging is central to her work. People who are sent to see a radiologist for a suspected cancer are understandably anxious, but cancers detected early on a mammogram, ultrasound, or MRI are at a stage when they are the most treatable and the type of treatment needed is minimal. Dr. Moy estimates that more than 90 percent of the time, a mammogram is going to be normal. For the remaining 10 percent, three-quarters of those will be found to be normal after additional imaging.

For the people for whom cancer has been detected, Dr. Moy assures them that there is a plan for their treatment.

“People have to leave my office with peace of mind. And if they don’t have that, then I don’t think I’ve done my job,” Dr. Moy says. “One of the benefits of being treated for cancer at Perlmutter Cancer Center is that we develop a network of healthcare providers—radiologists, oncologists, nurses, and advanced practice providers—who are all talking to each other and working with each other so people with cancer feel like they’re not lost in the system. We care about our patients, and they trust us to guide them as they are deciding how to proceed in their cancer treatment.”