A new machine learning system may lead to more accurate early detection of breast cancer and reduce the amount of unnecessary surgeries performed.
Researchers—from the Massachusetts Institute of Technology’s (MIT) Computer Science and Artificial Intelligence Laboratory (CSAIL), the Massachusetts General Hospital and the Harvard Medical School—used machine learning to help predict if a so-called “high-risk” lesion, identified via a needle biopsy and a mammogram, will turn out be benign at surgery.
Early detection is crucial for cancer sufferers, with mammograms being the best test available for detection.
However, one common cause of false positives are so-called “high-risk” lesions that appear suspicious on mammograms and have abnormal cells when tested by needle biopsy.
In these cases, the patient typically undergoes an expensive, painful and scar-inducing surgery to have the lesion removed, which ends up being benign about 90 percent of the time.
Using machine learning, researchers tested 335 high-risk lesions and correctly diagnosed 97 percent of the breast cancers as malignant, reducing the number of benign surgeries by more than 30 percent.
“Because diagnostic tools are so inexact, there is an understandable tendency for doctors to over-screen for breast cancer,” Regina Barzilay, Ph.D., MIT’s Delta Electronics Professor of Electrical Engineering and Computer Science and a breast cancer survivor herself, said in a statement. “When there’s this much uncertainty in data, machine learning is exactly the tool that we need to improve detection and prevent over-treatment.”
The system is trained on information about more than 600 existing high-risk lesions and looks for patterns among many different data elements, including demographics, family history, past biopsies and pathology reports.
Using a method known as a “random-forest classifier,” the team’s model resulted in fewer unnecessary surgeries compared to the strategy of always doing surgery, while also being able to diagnose more cancerous lesions than the strategy of only doing surgery on traditional “high-risk lesions.”
“This work highlights an example of using cutting-edge machine learning technology to avoid unnecessary surgery,” Dr. Marc Kohli, director of clinical informatics in the Department of Radiology and Biomedical Imaging at the University of California at San Francisco, said in a statement. “This is the first step toward the medical community embracing machine learning as a way to identify patterns and trends that are otherwise invisible to humans.”
The next step will be to incorporate the actual images from the mammograms and images of the pathology slides into the model, as well as more extensive patient information from medical records.
The model could also be tweaked to be applied to other kinds of cancers and diseases.
Breast cancer accounts for 40,000 deaths annually in the U.S.
