TL;DR:
- Researchers at Washington University have developed an AI model to predict preterm births as early as 31 weeks of gestation.
- The model combines electrohysterogram (EHG) measurements with clinical information for accurate predictions.
- Deep learning algorithms outperformed other methods in predicting preterm births, offering a promising approach to prenatal care.
- The team’s method utilizes EHG recordings and statistical signal processing techniques for improved accuracy.
- Higher frequency components of EHG measurements were found to be more predictive of preterm births.
- The model’s effectiveness with shorter EHG recordings makes it potentially cost-effective and usable in clinical and home settings.
- The research represents a significant advancement in the field of maternal healthcare, with implications for early interventions and personalized care.
Main AI News:
The alarming rise in preterm births, affecting almost 10% of pregnancies worldwide, has spurred researchers at the prestigious McKelvey School of Engineering, Washington University in St. Louis, to devise a cutting-edge solution. By harnessing the power of deep learning and analyzing electrical activity during pregnancy, they have successfully developed an innovative model capable of predicting preterm births as early as 31 weeks of gestation. These groundbreaking findings were recently published in the esteemed scientific journal PLoS One.
Under the guidance of Arye Nehorai, the distinguished Eugene & Martha Lohman Professor of Electrical Engineering, and Uri Goldsztejn, a renowned biomedical engineering expert who obtained his master’s and doctorate degrees from Washington University, the research team has paved the way for a revolutionary approach to prenatal care. Their method combines electrohysterogram (EHG) measurements with essential clinical information obtained around the 31st week of gestation, achieving a level of performance comparable to the clinical standards utilized to detect imminent labor in women exhibiting symptoms of preterm labor.
To develop their groundbreaking methodology, Nehorai and Goldsztejn harnessed the power of deep learning, training a sophisticated neural network on data collected from 159 pregnant women who were at least 26 weeks into their pregnancy. The dataset comprised 30-minute EHG recordings obtained during regular check-ups, as well as recordings from expectant mothers hospitalized with signs of preterm labor. Impressively, the deep learning algorithm outperformed other existing methods, effectively combining EHG data with clinical information to deliver highly accurate predictions.
The team’s deep recurrent neural network was trained using data samples indicative of specific pregnancy outcomes, enabling the network to learn the most informative features from the data and predict preterm births with remarkable accuracy. This groundbreaking work represents the first-ever method capable of predicting preterm births as early as 31 weeks using EHG measurements, while maintaining a clinically useful level of accuracy. Building upon prior research conducted in Nehorai’s lab and published in PLoS One, this latest study represents a significant advancement in the field of maternal healthcare.
In a previous study, Nehorai and his collaborators pioneered a method to estimate electrical current in the uterus during contractions using magnetomyography. By leveraging this noninvasive technique, which maps muscle activity by recording abdominal magnetic fields generated by electrical currents in muscles, they established a foundation for further advancements in preterm birth prediction.
Additionally, the team’s recent research, published in Biomedical Signal Processing and Control, introduced a statistical signal processing method capable of isolating uterine electrical activity from baseline electrical activity, such as the woman’s heart activity, in multidimensional EHG measurements. This breakthrough allows for more precise identification of uterine contractions.
Nehorai and Goldsztejn’s extensive research unveiled that different components of the EHG measurements played crucial roles in the predictions made by their model. Notably, higher frequency components within the EHG measurements demonstrated greater predictive power for preterm births. Furthermore, the team observed that their model yielded effective predictions even with shorter EHG recordings. This remarkable finding could significantly enhance the usability and cost-effectiveness of the model in both clinical and home settings.
According to Nehorai, preterm birth should be perceived as an abnormal physiological condition rather than merely an early end to a pregnancy. Consequently, physiological measurements, such as EHG recordings, have the potential to highlight a more pronounced distinction between pregnancies that conclude prematurely and those that reach full term, surpassing the information provided by continuous characteristics correlated with gestational age at delivery.
Looking ahead, Nehorai and Goldsztejn have ambitious plans to develop a device capable of recording EHG measurements. By collecting data from a larger cohort of pregnant women, they aim to refine their methodology and validate the results further. This significant step forward in predicting preterm births has the potential to revolutionize maternal healthcare, ensuring early interventions and personalized care for expectant mothers, ultimately improving outcomes for both mother and baby.
Conclusion:
The development of an AI model capable of accurately predicting preterm births at an early stage has immense implications for the maternal healthcare market. This breakthrough offers healthcare providers the opportunity to intervene and provide personalized care to expectant mothers at risk, potentially reducing the rates of preterm births and improving outcomes for both mother and baby.
The integration of deep learning algorithms and noninvasive EHG measurements has the potential to revolutionize prenatal care, leading to a demand for advanced medical devices and technologies to support this transformative approach in the market. Healthcare companies and investors should keep a close eye on the advancements in this field and consider the potential market opportunities that arise from this groundbreaking research.
