Predicting labour onset with AI
Expectant mothers and fathers to-be are understandably excited to know when their baby is likely to come, and there are important clinical implications to knowing whether a mother is likely to go into labour too early. But, currently the best we can do is estimate which day in a 5 week long window is the so-called ‘due date’, which as many parents will know is hardly ever accurate. Now, researchers at Stanford university have developed a test that uses artificial intelligence to narrow down that window. Eva Higginbotham spoke to The University of Cambridge’s Catherine Aiken, who wasn’t involved in the study...
Catherine - If we could be more accurate about when a baby was going to be born, that would be a massive game changer in obstetrics. First of all, for people having completely normal pregnancies, it would just be really nice to know when to expect their baby. But for us as obstetricians looking after high risk pregnancies, it would be amazing to be able to have a better fix on the likelihood of preterm birth, for example, because if we know about that, then we can prepare for it and we can also work on strategies to prevent it. It would also be really important for women who are at risk of pregnancy complications, like preeclampsia or sepsis. All kinds of women would benefit immensely from better and more detailed information about when the baby would be likely to be born. It's something that has always been a massive question, not only in modern obstetric science, but even before that. You know, there are ancient writings trying to predict times when babies are going to be born. And it's been one of those questions that you can find science on, right from the very beginning of scientific writing, but you never really get all that much better at predicting it, despite all the enormous amount of studies that have been conducted in this field.
Eva - So what sort of work are people doing to try and improve our estimates?
Catherine - So the kind of work that this study has done is amazing. So what the authors have done is to take blood samples from the mum over three different time points in the hundred days leading up to giving birth. Make a model of the substances, the proteins, the enzymes, the molecules that appear in mum's blood, and look at how they change up to the time of giving birth. And so what they can effectively do is produce a profile of dynamic changes in the mother's blood as you approach labor.
Eva - So is it like a fingerprint of, well, as you get closer and closer to labour, this substance goes up, but this substance goes down and they interact in that kind of way?
Catherine - Exactly. It's like finding a signature set of changes that you expect to happen in the blood of the mother leading up to the time of delivery with the idea that if you see that particular fingerprint, then you know that that's someone who's going to give birth sooner rather than later.
Eva - What sort of markers did they find?
Catherine - So the things that they find surging before delivery really weren't a huge surprise to us. They find a massive increase in the steroid hormones circulating in mum's blood, and that's something that we've known from previous studies, and also from animal models of animals going to give birth, that there's an enormous surge and a difference in the type of steroids that you produce just before delivery occurs. We also find a really interesting protein called IL-1R4, which is used in coordinating a change in immune activation and in regulating inflammatory responses. And that's really important because what we know is that pregnancy maintenance requires an immune tolerance to your baby. And then that there's an inflammatory response that seems to be dampened down by this IL-1R4, which could be really important in regulating that pre-labour phase.
Eva - With this signature then, how narrow a window did they manage to narrow down to compared to how big a window we normally have to say "the baby's going to be born in here somewhere"?
Catherine - So the five week window that we have now of a normal range for the baby to be born looks like it could be narrowed, by this kind of study, to around two weeks, which would be a hugely different clinical scenario.
Eva - How likely do you think it is that we're going to end up in a future where we can say, look, Tuesday two o'clock that's when you're going to go into labour naturally?
Catherine - Oh, wow, that would be amazing. I think we're a little away from that just yet!