Machines can be taught to recognize the tell-tale and tiny changes in mammograms that later turn into tumors.

Dr. Regina Barzilay is an MIT professor and a recipient of the MacArthur Fellowship in 2017. She survived a 2014 diagnosis of breast cancer. She was astounded by the seeming uncertainty accompanying the treatment of her disease at the Massachusetts General Hospital. Dr. Barzilay, therefore, decided to use artificial intelligence to collect and analyze the available information on breast cancer.

Harnessing artificial intelligence to treat breast cancer

Dr. Barzilay is a holder of a doctorate in computer science. She, therefore, went on to create a database from the pathological reports stretching back three decades at the Mass General Hospital. For this project, she developed an algorithm to parse the data, which covered more than 100,000 breast cancer patients.

Dr. Barzilay also taught computers to analyze the medical images of breast cancer using mammogram images from 60,000 patients.

“Machines work more effectively than human eyes,” Dr. Barzilay explains. “They can register subtle changes in tissue — influenced by genetics, hormones, lactation, weight changes — that we cannot see.”

The computer model’s results were superior to those from current clinical standards. The artificial intelligence breast cancer algorithm identified 31% of future patients with breast cancer as “high risk.” In contrast, the current clinical standard identified only 18%.

Dr. Regina Barzilay collaborated on the project with Dr. Constance Lehman, the chief of breast imaging at Mass General; Dr. Kevin S. Hughes, also at Mass General; and Adam Yala, her graduate student at M.I.T.

[Related Story: Artificial Intelligence May Provide Advance Warning of a Heart Attack ]