AI model could help predict dangers in unborn babies

SALT LAKE CITY — Researchers have developed a new artificial intelligence tool that may one day be able to help doctors assess the risk a fetus has for serious complications like stillbirth.

Dr. Nathan Blue with the University of Utah said that anytime a fetus is in the bottom 10% for growth, doctors become concerned about the possibility of stillbirth.

“This could be a very serious thing that might require a lot of extra monitoring, concern, and early delivery to prevent stillbirth,” he said.

On the other hand, sometimes, there’s nothing to worry about.

“Most of those babies are actually totally healthy. [It’s] a small percentage of those babies, maybe 15 or 20% [who] are at actually at a meaningfully increased risk of more serious complications like stillbirth,” Blue said.

The current tools used to assess fetal risks are “somewhat crude,” Blue said. But now, there is AI.

Blue said a new AI model helps identify patterns in the baby and its mother that can help doctors give a much more accurate diagnosis.  The tool also saves on the added expense and stress of extra monitoring and doctor’s visits. 

“The goal is to, rather than me having to say ‘this could be either very high risk or pretty low risk,’ to say, ‘well, actually, we can be a little less concerned because given your specific scenario, the risk of a serious complication, for example, a stay in a NICU, is very low.’”

He said this artificial intelligence tool looks at patterns that wouldn’t be obvious to most doctors. 

“Obviously, the advantage to a doctor and the patient then would be not instilling worry where there doesn’t need to be as much worry,” Blue said. “But then also allowing us to actually target all of our resources, and all of that extra monitoring and attention and care to the people who really are in the highest risk scenarios.”

Blue said there will be another benefit to doctors using this new tool.

“The idea here is that we could customize [a] plan, or personalize it in a way that meant the high-risk people get the care they need and the low-risk people don’t get exposed to unnecessary interventions and cost and stress,” Blue said.

He said it’s still going to be a few years before this model is ready for distribution, as experts want to be rigorous and cautious.

“Now, that doesn’t mean that we can’t test it or introduce it or whatnot,” Blue said. “But what we don’t want to do is implement something before we have assessed its effect on care.” 

Follow @KSLdbrinkerhoff