Many of those researching equity in healthcare can see profound differences in care by gender, and the advancement of women’s health, as it compares to men’s health. That’s in addition to other disparities that seem to crop up somewhat consistently, leading to challenges in building a healthcare world that’s egalitarian in nature.

“Artificial intelligence is widely heralded as a transformative force in medicine, promising to accelerate drug discovery and enhance clinical research to combat diseases that have plagued humans for millennia,” writes Michelle R. Kaufman , PhD, MA, assisted by Ai Yajuan. “For everyone to benefit from the predicted future breakthroughs, AI developers should work to rectify a long-standing inequity in medical research: the systemic bias against women.”

Part of this relates to the practice of radiology, through which so many types of serious conditions are diagnosed. AI has really had a big impact in radiology, in helping clinicians to read scans. What does that mean for women’s health?

At our Imagination in Action event April 9 and 10, Boston Globe reporter Aaron Pressman interviewed Connie Lehman of Clarity, Inc. about her journey to figure out the impact of AI on women’s health.

“I was so interested in women's health and global health and access around the world to better health care,” Lehman said, setting the stage for a discussion about advances in this area. “My friends were surprised that I got really excited about radiology. They just didn't think it was a fit for me. But the power of the image, I thought it was so incredible and so untapped in healthcare, just the power of imaging the body, and the technology by which we can take images outside and inside the body. So I started in that domain.”

She talked about the limitations of the traditional mammogram.

“The mammogram isn't enough,” Lehman said. “Some women really need contrast enhanced imaging where you can see blood flow and vascular flow inside the body, such as in an MRI or contrast enhanced mammogram.”

Approaching analysis on a case by case basis, she said, is important.

“Every hammer isn't the right hammer for every single problem, and we were using mammography as a very blunt tool, and we didn't have ways to identify the right test in the right patient at the right time, so we're really stuck in a very crude age, based on a one size fits all screening paradigm,” Lehman noted.

Describing early research on this type of technology in the 1970s, Lehman explained that women’s healthcare also got a shot in the arm when an MIT professor named Regina Barzilay found herself diagnosed, and when broadcast professional Katie Couric also battled breast cancer.

“She was shocked by how little science was being used to determine exactly what she should do next,” Lehman said, of Barzilay’s story. “And so she felt that this was a domain where her lab could really help change the field.”

Couric’s case, she said, exemplifies the sadly common scenario where someone gets breast cancer despite an absence of family medical history.

She also pointed to challenges with the data, and the process.

“I think we are always battling, maybe in humanity, but definitely in healthcare,” she said. “We are uncomfortable with change, so we have to look at our best tools for change management, and how to have buy-in, and how to have interest, and how to have science lead our decision making.”

There’s really no lack of data here, Lehman conceded.

“We have a huge mass of data around the globe, and within the U.S., very, very large numbers of women being screened with mammography every year,” she said. “So much of what we do in breast cancer and breast imaging is about these very large databases.”

Lehman pointed to the use of something called a “clarity score” as a more targeted rating of numerous mammography scans. This type of profile, she suggested, can boost the power of diagnosis in clinical environments.

“The input to the model is the four basic views of a woman's breast tissue that's obtained with every screening mammogram,” she explained. “So two views of the right breast, two views of the left breast, those four images go into the model. The model extracts the predictive data and provides a percent score of the likelihood of that woman to develop breast cancer in the next five years. So it's a five year AI-mammogram-based risk score.”

Diversity in Patient Audiences

“Just because you're at increased risk doesn't mean you'll stay there,” she said, noting that with the new tools, doctors can figure out a woman’s changing risk rating, provide the right information to support ongoing risk management, and make better care decisions.

She also talked about diversity as a major goal for healthcare, and a global consortium hard at work on these issues, while acknowledging that, in this sense, the medical world still has a long way to go.

“Throughout the history of medicine, we have really fallen down in the area of ensuring that our studies include the full diversity of patients at risk of that disease. And we're not there yet. We think we are. Still, at NIH, we have studies in mice that don't include female mice because their cycles really mess up the data. Think about it, all the doses that we're getting in our medications as women are based on men, and it is a real challenge.”

There’s a lot more in the interview, about patient equity, the FDA process, and other happenings around this part of a vastly important sector. It illustrates how using AI can actually even out patient care in key areas, and how to enhance diagnosis when it counts. Stay tuned for more from our spring conference.