How One Woman’s Breast Cancer Experience May Revolutionize Cancer Care

Unknown-23
mammograms for early detection

I love to write about good news. I especially enjoy elaborating on advances in the world of science during these times when science is too often attacked. This story shares some qualities with my recent post about the extraordinary Nobel Prize Winners in Physiology or Medicine. 

Like the Nobel discovery, this one seems destined to save lives and dramatically reduce suffering. It’s the result of one brilliant woman’s using her own status as a breast cancer survivor to create potentially dramatic changes in the detection and treatment of the disease.

My new hero is Regina Barzilay, PhD. She isn’t a physician, yet she seems to be upending medical practice for the better through the use of artificial intelligence (AI).

Barzilay is a professor of computer science at the Massachusetts Institute of Technology (MIT) and a certified genius: in 2017, she was the recipient of a MacArthur Fellowship “genius grant.” 

She and her team, which now includes experts from both MIT and Massachusetts General Hospital (MGH), have created computer algorithms that predict the likelihood of a patient’s developing breast cancer in the next five years. 

The model they designed began with a database containing pathology reports of more than 100,000 women treated at MGH over 30 years. Barzilay and her team then “taught” the computers to provide specific information from mammograms of more than 60,000 patients. 

According to an article in MIT News, 

“Trained on mammograms and known outcomes from over 60,000 MGH patients, the model learned the subtle patterns in breast tissue that are precursors to malignant tumors.”

Barzilay told Susan Gubar, who wrote about this remarkable work in Science Times, the special Tuesday section of The New York Times, that

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

Barzilay showed Gubar the results of her own mammograms from 2012, 2013, and 2014. The cancer that was diagnosed in 2014 was, in fact, evident in the two previous views.

I found myself deeply touched by that information, imagining what it must have been like for her to learn her cancer could have been caught and treated two years earlier, and seeing how heroically she turned her personal knowledge into this bold campaign to prevent other women from experiencing similar anguish—or worse.

Gubar reports that

“The enthusiasm Dr. Barzilay brings to this undertaking is fueled by her dismay at current approaches to cancer care. While being treated at MGH, she was struck by the high degree of uncertainty surrounding treatment of her disease.

“Why did her questions go unanswered about how other patients at the same hospital with similar tumors fared with this or that drug or with this or that surgery? Why was there so little information?”

The apparent explanation was that oncologists rely on the results of clinical trials in determining treatment regimens. That’s not surprising; they seek evidence-based medicine.

The problem Barzilay saw was that the trials enrolled just about 3 percent of eligible women, meaning 97 percent weren’t part of the picture. 

Barzilay termed this approach a “primitive practice” that was a “travesty,” Gubar reports, “especially because large volumes of information about patients accumulate in every hospital.” (Emphasis mine)

But a stumbling block to the work she proposed was that the data are written in “free-text” English, rather than in a form a computer could process. That’s when she and her colleagues began building the databases.

In one study, the Barzilay team’s model identified 31 percent of patients as high risk for future breast cancer, in contrast with the existing clinical standard, which identified 18 percent. That difference encompasses a great many women.

Once this work is more fully implemented, the result, Gubar writes, will be that

“New patients will be empowered by learning how tumors with particular characteristics responded to specific treatments. Machines accessing subsets of the population will also make it faster and cheaper for clinicians to identify patients with particular disease characteristics and to enroll them in clinical trials.”

One particularly valuable aspect is that the cancers are detected regardless of the patient’s race—an important consideration in view of the much higher breast cancer mortality rate among African-American women.

According to Gubar, similar efforts are occurring at Google, where AI specialists are examining scans for lung cancer. It seems reasonable to me, as a nonscientist, that this approach is potentially replicable with all sorts of cancers. (I’d welcome hearing from anyone with expertise in AI, cancer, or the intersection of the two fields.)

Barzilay knows buy-in from oncologists is critical to this effort. She sought to learn whether oncologists were reaching out to AI researchers; when she found that they weren’t, she also made one of her aims to enlighten them about these new possibilities.

Writes Gubar:

“Dr. Barzilay and her collaborators want to usher in the day when no woman is surprised by a late-stage diagnosis and when all breast cancers are curable.

“They also hope to solve the problems of over- and under-testing. Instead of a one-size-fits-all practice, the frequency of screenings and biopsies could be customized with sufficient data.”

That could be a huge benefit to patients. For example, at present, the MIT article notes, there is a discrepancy in screening guidelines, with the American Cancer Society recommending yearly screening beginning at age 45, while the U.S. Preventative Task Force says screening should be every two years, beginning at age 50.

And for implications for individual patients, Gubar points to the young women she knows who are aware that they have an inherited BRCA genetic mutation, which can substantially increase their risk for breast cancer (as well as for ovarian cancer).

With great anxiety, they are contemplating prophylactic double mastectomies—although there’s no assurance that such drastic surgery is necessary for them. The numbers of such women are increasing now that genetic testing is so readily available. 

Barzilay’s work can help women better face this difficult decision. In responding to Gubar’s query about such affected women, she stressed:

 “With a CD of their scan, we would be able to tell them their personal risk.”

I wish Dr. Regina Barzilay a long and productive life as she continually refines and expands her invaluable work.

Annie