Your Pregnancy Matters

3 ways Ob/Gyns use AI in pregnancy care

Your Pregnancy Matters

Image Here
Ob/Gyns use artificial intelligence every day to improve pregnancy care.

Artificial intelligence (AI) is becoming a major part of nearly every industry, and Ob/Gyn is no exception. Though some doctors worry AI will become so sophisticated it will one day take over their jobs, research suggests that AI is no substitute for instincts, expertise, and the human connection. 

In April 2019, NPJ Digital Medicine published a paper on doctors’ impressions about AI that reflected an interesting cross-section of opinions. Approximately half of the respondents were tenured doctors and half were younger trainees – digital natives who likely grew up with technology.

The survey results showed a significant divide between fear and excitement. Fewer than 20% of respondents said AI would one day replace human doctors, while more than 42% anticipate that AI will actually help create new jobs in health care. 

Ob/Gyn teams regularly use AI – even though they may not recognize it. I consider it a clinical support tool that can lead to more accurate diagnoses, better training, and improved patient outcomes. Let’s discuss the three main ways we use AI and what the future might hold for artificial intelligence in our field.

"Pattern recognition helps us more accurately diagnose and treat patients through validation of our diagnoses. Patients benefit because their doctor isn’t making decisions in a vacuum, cross-checking not only with colleagues but also a wealth of anonymized data."

Robyn Horsager-Boehrer, M.D.

3 ways we use AI today

The foundation of AI is pattern recognition. Machines are trained by data scientists to “learn” what results are considered normal and abnormal based on a library of previously entered data. 

Pattern recognition helps us more accurately diagnose and treat patients through validation of our diagnoses. Patients benefit because their doctor isn’t making decisions in a vacuum, cross-checking not only with colleagues but also a wealth of anonymized data.

1. Electronic medical records 

EMR is one of our most basic applications of AI. As we enter data into the EMR system, we receive notifications to recommend subsequent testing for certain results. These alerts can help us, for example, identify which mothers might be at risk for preterm birth based on personal risk factors and the relationship between certain test results and patterns the system has learned.  

2. Pap smear 

We use computer-assisted technology in screening women for cervical cancer. The cells we collect during a Pap smear are turned into slides by a machine, ensuring higher quality information to review. Then, an image processor helps identify areas on the slide to review under the microscope, reducing screening error by technicians. AI may also be able to perform cervical cancer screening without a Pap smear. Researchers are evaluating computerized image analysis of digital photographs of the cervix to advance cervical cancer screening in parts of the world where resources are limited and deaths from cervical cancer high. 

3. Ultrasound

Ultrasound machines that incorporate AI can recognize structures that a doctor must measure. For example, if I want to measure the size of the fetus’ head, our machine uses AI to recognize the image I’ve taken and place the calipers in the position it believes is most accurate. Of course, I can tweak the endpoints if I don’t agree, but it’s a big step forward that doctors don’t have to rely solely on their expertise to provide optimal measurements. 

For training, we use AI-backed ultrasound simulation platforms that assign virtual image-gathering tasks. A trainee might be asked to obtain certain views of a fetus’ heart in a simulated setting. The platform then grades the quality of the images on accuracy and other criteria, offering valuable feedback and experience without involving patients. What’s more, it is very efficient as it doesn’t require a teacher “at the elbow” of each trainee.

Related reading: A head-to-toe tour of your baby’s ultrasound

Overcoming challenges of AI 

Despite its advantages, artificial intelligence discussions make some patients and providers nervous. Two concerns often arise: bias and “cheating.”


Bias – conscious or subconscious differences in care based on personal characteristics, such as ethnicity or gender – can occur at any point of care regardless of whether AI is used. The concern is that physician bias will skew the data that an AI algorithm recognizes as good/normal or bad/abnormal.

I would argue that in Ob/Gyn, the majority of data we analyze with AI today is not dependent on patient context. Take the analysis of Pap smear samples – there is a general consensus among doctors of what is normal and what isn’t, regardless of the age or nationality of the patient. For ultrasound images, there are standard requirements of how the images should appear.


In the physician survey, 15.6% of respondents thought their colleagues would view them negatively if they used AI. However, I’d argue that whether clinical support data are compiled in a book or available through an app or machine, using available resources shouldn’t be considered “cheating.” 

Imagine how many thousands of images a doctor must examine to become an expert in recognizing just one normal vs. abnormal pattern, let alone many. And not every organization has the luxury of specialized doctors for every condition like academic medical centers do. It would be ridiculous to expect doctors to remember every known and new pattern, and we’d be doing our patients a disservice if we didn’t use AI as a support tool.  

Related reading: Why pregnant women need ‘pit crew’ Ob/Gyns, not ‘cowboys’

The future of AI for Ob/Gyn

We are excited about potential future applications for fetal heart rate monitoring. While there have been studies in this area, researchers have not yet successfully used AI to recognize and reduce instances of low Apgar scores and neonatal intensive care unit admissions. 

I hope AI eventually will help us recognize patterns in fetal heart rate issues. However, further confirming that AI will not “take our jobs,” fetal heart rate data must be overlaid with contextual information about the mother and baby to be considered useful, including:

●     Fetal age

●     Objective data, like the mother’s blood pressure or temperature

●     Whether the mother or fetus has other conditions that impact the in-utero environment 

Interestingly, many survey respondents suggested that the time they might save using advanced AI could be spent on education and research that can translate into further efficiencies and improvements in patient care. 

In 10 to 20 years, we’ll likely use AI to solve problems and answer women’s health and pregnancy questions we can’t answer today. Rather than fearing AI, Ob/Gyns should embrace it for the current benefits and how it might shape our jobs, careers, and ability to care for patients in the future.

Stay on top of health care news. Subscribe to our blog today.