AI holds ‘transformative potential’ for gynecology, obstetrics and menopause care

Key takeaways:

• AI-assisted technology can help guide care for reproductive-aged women.
• There are few AI programs aimed at menopausal women due to a lack of good data.

CHICAGO — The benefits of AI-assisted technology in medicine are rapidly extending to obstetrics and gynecology, with programs in development or already deployed to predict everything from mode of delivery during labor to cervical cancer risk.

AI offers the ability to one day achieve the “holy grail” of personalized medicine via sophisticated machine learning algorithms, helpful chatbots, and fast and accurate image interpretation, all supporting the day-to-day work of human physicians, Melissa Wong, MD, MHDS, director of informatics and AI strategies and assistant professor of maternal-fetal medicine at Cedars-Sinai Medical Center, said during the opening symposium at the Annual Meeting of The Menopause Society. AI holds “transformative potential” for both gynecology and obstetrics care; however, there is also an urgent need for improved data on midlife women to fully harness these same technologies for menopause care, Wong said.

“AI and its integration into clinical medicine is very much here to stay,” Wong told Healio. “As clinicians, we need to remain engaged and at the forefront of AI so that it achieves what I think it should, which is ultimately allowing us to do what we got into medicine for — to focus on the physician-patient relationship. Otherwise, AI will be essentially placed upon us and we will prioritize the wrong things.”

For diagnostic applications, AI technology is already widely deployed, whether clinicians realize it or not, Wong told Healio.

“When a patient receives a CT scan of their chest, at least at Cedars-Sinai, AI is pre-triaging which scans are pulled up for the radiologist to see first,” Wong told Healio. “Certain conditions are more acute — say, a blood clot in your lungs — and the AI looks through the images and moves priority cases to the front of the line. All of that is almost invisible to most clinicians, but there is a reason your patient’s scan came back sooner.”

Wong said there is a growing role for AI in obstetrics and gynecology to potentially improve clinical outcomes, though current data limitations have slowed the uptake of any AI technology in menopause care.

AI in obstetrics, gynecology

AI is already advancing ultrasound diagnosis in obstetrics and, within gynecology, clinicians are employing AI across the field, including in endometriosis, infertility, and even for cervical cytology, Wong said.

At Cedars-Sinai, Wong and colleagues developed an AI application that performs a contextual analysis of cervical cytology results, produces a recommendation for next steps, if needed, and generates a letter to the patient about the results.

“We are using generative AI in a different way to try to guide clinical decision making and contextualize Pap smear results,” Wong said during the presentation. “I do not need to tell anyone here what a pain it is to interpret that single Pap smear. You need the context of their previous data; you have to have your phone next to you so you can pull up the app. Despite how ubiquitous these Pap smears are, interpretation has been convoluted.”

Using machine learning, Wong and colleagues developed an algorithm that can help predict mode of delivery, using real-time, ongoing intrapartum data in laboring patients.

“We have spent the last 7 years developing our partometer, a machine learning-based guidance for the likelihood of a patient to have a vaginal delivery,” Wong said. “At this point, our model is almost 90% accurate and, perhaps most exciting is it is actually running. For every patient who walks into our labor and delivery, we are generating this ongoing personalized recommendation, in real time, every 5 minutes.”

AI for menopause health? More data needed

Despite the profound impact of menopause on women’s health, it remains underrepresented in medical research and AI model development, primarily due to difficulties in data acquisition and interpretation, Wong said.

As for specific AI-assisted devices and technology created to inform menopause health, Wong said there are “practically none.”

“The only things out there are chat bot guidance [programs] regarding menopause symptoms,” Wong told Healio. “That is it. These programs are just repackaging existing data. To do anything novel, we need better data.”

Wong said electronic health records need to be redesigned to better assess menses and menopause status.

“We will not get anywhere on menopause and being able to study it with AI without better data around menopause,” Wong told Healio. “Most EHRs, if they even ask about menopause, simply ask: ‘Are you menstruating?’ We do not know anything about symptoms surrounding menses. We have never taken a deep dive in a data-driven way about what it means to experience menopause. We need more granularity around menopause.”

If such data problems can be resolved, future AI applications might better predict and even help manage menopause symptoms, personalizing hormone therapy and individualizing care, Wong said. For these tasks, clinician involvement in AI development is crucial, so that AI technologies enhance, rather than replace, human judgment, Wong said.

“What do we need to be cognizant about? Three principles: How was the target selected, what does implementation look like, and algorithmic bias,” Wong hold Healio. “If we do not do it right, the mistakes of our past will become the recommendations of our future.”

For more information:

Melissa Wong, MD, MHDS, can be reached at X (Twitter): @MelissaWongMD.