A-I May Improve Breast Cancer Detection

As with nearly every disease or condition, early detection equals early treatment equals few deaths, and longer life. Up till now the most popular method of early detection is by Mammogram (x-ray of the breast). However, the rate of false positives (errors indicating cancer) is still way too high, resulting in tens of thousands of un-necessary biopsies and surgeries. Part of that is overcautiousness on the part of the physicians in charge, partly because of patient fear, and partly because of peer-pressure. Women’s magazines are full of accounts of famous women receiving the procedures.

Press Contact

Abby Abazorius
Email: abbya@mit.edu
Phone: 617-253-2709
MIT News Office

Media Resources

1 images for download

Access Media

Media can only be downloaded from the desktop version of this website.

Share

Comment

High risk lesions on the x-ray appear to be a possible suspicion and many of those demonstrate abnormal cells on biopsy. A biopsy is where a small amount of breast tissue is taken through a large needle and analyzed by microscope. The suspicious lesion can then be removed through minor surgery. However, 90% of the time the lesion is not cancer (or “benign”) rendering the somewhat painful and scarring surgery completely unnecessary.

Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), Massachusetts General Hospital, and Harvard Medical School have turned to artificial intelligence (AI) to improve the detection and analysis of suspicious lesions. With the computerized enhancement, the incidence of unnecessary biopsies and surgeries should be reduced.

Regina Barzilay, MIT’s Delta Electronics Professor of Electrical Engineering and Computer Science and a breast cancer survivor herself explained that many doctors over-test with mammograms due to the known false positives, and to err on the side of knowing instead of guessing.

Using a “random-forest classifier” the new computer model looks at several factors including pathology reports, biopsies, and family history. When it was used on 335 patients it correctly found 97% of the lesions as being cancer, and thereby reducing the benign surgeries by 30% compared to normal systems. The new model diagnosed 97 percent of cancers compared to 79 percent.

Collaborator Constance Lehman, professor at Harvard Medical School and chief of the Breast Imaging Division at MGH’s Department of Radiology believes this is the first study of its kind and predicts a great deal more research and utilization in the near future, for better prediction and informed decisions for breast cancer victims. She states they will be employing this new and future improved systems within the next year.

Marc Kohli, director of clinical informatics in the Department of Radiology and Biomedical Imaging at the University of California at San Francisco feels the technology is cutting edge. He hopes it will lead to physicians accepting computer/machine diagnostics and patterns that might go otherwise undetected and lead to improved outcomes.

Dr. Ronald Klatz, President of the A4M, on Oct 20, 2017 says, “It’s wonderful that world-wide research is getting closer to eliminating this form of cancer. With the increased knowledge from this study it can then be applied to other forms of cancer. While mammograms have been an extremely useful tool in the past few decades, new technologies like this will undoubtedly increase reliability and decrease false positives.”