AI in Women's Health Is Moving Faster Than Most Realize
As a physician-scientist who has worked in obstetrics and gynecology across two continents, from rural mountain hospitals in western Iran to academic medical centers in the United States, I've had a front-row seat to how slowly medicine sometimes changes, and how suddenly it can lurch forward. What I'm watching happen right now with artificial intelligence in maternal-fetal medicine and reproductive health is the lurch. And I think most people, including many of my colleagues, haven't fully absorbed what that means yet.
The shift I'm describing isn't abstract. In my own research and clinical training, I've seen AI-assisted tools begin to enter spaces that were, not long ago, entirely dependent on clinician pattern recognition and manual review. Risk stratification in high-risk pregnancies. Screening algorithms in gynecologic oncology. Diagnostic support for conditions like endometriosis, which I studied at Yale under Dr. Hugh Taylor's laboratory starting in 2013, a condition that still takes an average of years to diagnose despite affecting millions of women. The tools exist now to cut that gap. The question is how we use them.
Here's where my concern sits. The speed of adoption is outpacing the conversation about whose data these models are trained on. I've spent a significant part of my career in settings where women were underserved, under-counted, and underrepresented in research. Naghan. Shadegan. Communities near the Iraq border where I practiced as an OB/GYN attending. If the datasets powering AI diagnostics reflect the same historical gaps that clinical research has always carried, then these tools will perform well for some patients and poorly for others. That's not a technology problem. It's a justice problem.
I made a concrete decision about this in my own work. When I began reviewing manuscripts for Reproductive BioMedicine Online in 2025, I started paying close attention to how authors were reporting their study populations when AI was part of the methodology. I began asking harder questions in my reviews about demographic representation. It's a small lever, but it's one I can pull. Peer review shapes what gets published. What gets published shapes what gets built. I believe that chain of influence matters.
What gives me real confidence about where this is heading is the growing number of researchers and clinicians who share this view. The field is not sleepwalking into a biased future, at least not entirely. The conversation about algorithmic transparency and diverse training sets is getting louder. My hope is that it gets louder faster than the commercial pressure to deploy these tools at scale. That's the race I'm watching most closely right now.
There's another dimension to this shift that I don't hear discussed enough: the emotional architecture of care. AI can flag a risk. It cannot sit with a patient who has just learned she's had her third miscarriage, or recognize that the woman describing "feeling fine" at her postpartum visit is not fine at all. I know what that looks like. I've been that patient navigating a high-risk pregnancy and IVF, and I know how much the human in the room mattered. Technology and compassion are not in competition, but we have to be deliberate about protecting the latter as the former expands.
My work through the Organization of Middle Eastern Girls and Women, and the years I spent with Integrated Refugee and Immigrant Services helping displaced women access care, reinforced something I carry into every conversation about AI: the women most likely to benefit from smarter diagnostics are often the least likely to be represented in the research behind them. Closing that gap is a design choice, not an accident. Clinicians, researchers, and reviewers all have a role in making it.
I'll be writing more about these intersections on this site, and if you want the fuller picture of the research and advocacy work behind this perspective, the about page lays it out in detail. The field of reproductive medicine is at a genuinely interesting inflection point. I intend to keep pushing, from inside the exam room and inside the literature, to make sure the technology we build actually serves the patients who need it most, not just the ones who were easiest to study.