Predicting the exact timing of labor in pregnant women continues to pose a significant challenge. Currently, due dates are estimated by counting 40 weeks from the date of a woman’s last menstrual period, even though actual gestation periods in humans can range from 37 to 42 weeks. There are no clinical tools available that can precisely indicate when labor will begin, leaving expectant mothers to report symptoms themselves, a method prone to frequent false alarms. When labor begins unexpectedly, it can lead to negative outcomes such as unplanned home deliveries, missed opportunities for timely medical interventions in cases of preterm birth, or even premature labor inductions for those who live far from medical facilities. Although shifts in body temperature are known to signal labor in various mammals, this phenomenon has not yet been studied in humans. Now, researchers have created a deep learning model that uses temperature trends to predict the daily likelihood of labor onset.

A research team at University of Arizona Health Sciences (Tucson, AZ, USA) conducted a study to determine whether temperature patterns could forecast labor in humans, as has been demonstrated in animals. Temperature readings already play a role in tracking ovulation and fertility. For their study, the team collaborated with a company that makes wearable sensors in the form of a ring, which records temperature every minute instead of just once daily. With this high-frequency data, the researchers trained a deep neural network-based artificial intelligence (AI) model to analyze the information. Deep neural networks are structured to mimic brain activity, consisting of an input layer to receive data, an output layer to generate results, and multiple hidden layers in between that carry out complex computations, much like how the brain processes data.

These deep neural networks also improve over time by learning from data, comparing their predictions with real-world outcomes, and adjusting to boost accuracy. By applying this method to continuous temperature readings, the researchers achieved successful forecasts of labor timing. Their final model accurately identified the onset of spontaneous labor for 79% of participants within a 4.6-day window when assessed seven days before actual labor began, and within a 7.4-day window when assessed ten days in advance. The team plans to validate the model through a larger-scale study to assess its clinical utility further. Their ultimate aim is to incorporate this AI model into existing wearable technologies or medical devices for broader use.

“With pregnancy, there are a whole lot of different things going on in the body. It’s not as simple as figuring out if the temperature is going lower or higher. For labor prediction, daily temperature readings do not give you a cohesive pattern of when somebody might go into labor,” said Shravan Aras, PhD, assistant director of sensor analysis and smart health platforms at the University of Arizona Health Sciences Center for Biomedical Informatics and Biostatistics. “We were able to develop deep neural network-based AI models that took all of this very high frequency temperature data – one data point per minute of temperature – and come up with a predictive output of when a mother might go into labor.”

http://www.allmycbs.com/en/index.asp