A new method that combines electrohysterogram measurements and clinical data may help predict preterm delivery as early as 31 weeks of gestation.
Preterm birth, when a baby is born before 37 weeks of pregnancy, affects one in 10 pregnancies globally. The United States is among the 10 countries with the highest number of preterm births, with African-American women being disproportionately burdened.
Current diagnostic tools, such as measuring cervical length, can predict premature birth within one week before delivery in women with preterm birth symptoms.
In an effort to improve prediction and birth outcomes, the researchers in the McKelvey School of Engineering at Washington University in St. Louis used measurements from electrohysterography (EHG), a noninvasive technique that detects uterine electrical activity through electrodes placed on the abdomen.
They also used clinical information from two public databases, such as age, gestational age, weight, and bleeding in the first or second trimester.
First, the researchers attempted to predict preterm births by using only clinical information. Then they trained a deep-learning model on data from 30-minute EHGs performed on a total of 159 pregnant women who were at least 26 weeks gestation. Among these mothers, nearly 19% delivered preterm.
In the study published in the journal PLOS One, the prediction models trained on both clinical information and EHG measurements slightly outperformed the models trained on clinical information and EHG measurements alone. Moreover, the new method outperformed the gold standard biomarkers of preterm birth, such as cervical length and fibronectin alpha.
"We predicted the pregnancies’ outcomes from EHG recordings using a deep neural network because neural networks automatically learn the most informative features from the data," the study author Uri Goldsztejn said in a press release. "The deep learning algorithm achieved a better performance than other methods and provided a good way to combine EHG data with clinical information."
Although the new prediction method shows promise in predicting preterm birth, additional examinations are necessary to determine which therapies are more likely to reduce the risk of early delivery and improve its outcomes. Moreover, machine learning algorithms in healthcare settings need to be developed in a way that protects data privacy and prevents social biases from driving the predictions.
Leading cause of death
Preterm birth complications are the leading cause of death in children younger than 5 years. The final weeks of pregnancy are crucial for a baby’s brain, lungs, and liver to fully develop. Those born earlier than 37 weeks are vulnerable to serious health complications, such as:
- Respiratory distress syndrome
- Chronic lung disease
- Injury to the intestines
- A compromised immune system
- Cardiovascular disorders
- Hearing and vision problems
- Neurological insult
Scientists still don’t know all the reasons why preterm births occur. There are factors that may increase the risk of preterm delivery, such as a history of early labor, being pregnant with two or more babies, using tobacco or abusing substances, and getting pregnant in less than 18 months after giving birth. However, some women experience premature births without having known risk factors.
Although the new method that involves clinical data and EHG measurements may help to predict preterm birth at 31 weeks of pregnancy, additional examinations are necessary to choose suitable therapies to reduce the risk of early delivery.
- PLOS ONE. Predicting preterm births from electrohysterogram recordings via deep learning.
- Washington University in St. Louis. Preterm births could be predicted at around 31 weeks’ gestation, new model shows.
- National Library of Medicine. Mortality and Acute Complications in Preterm Infants.
- BioMed Central. 15 million preterm births annually: what has changed this year?
- CDC. Preterm Birth.
- JAMA Network. Global, Regional, and National Incidence and Mortality of Neonatal Preterm Birth, 1990-2019.