A local research team said it successfully predicted preterm birth (PTB) risk using machine learning techniques.

Ewha Womans University Medical Center (EUMC) said on Monday that its research team, led by Professors Park Sun-wha and Kim Young-ju at the Department of Obstetrics and Gynecology, published their study which predicted the risk of PTB by analyzing bacterial risk scores in cervicovaginal fluid (CVF) with machine learning techniques.

PTB before 37 weeks of pregnancy accounts for 5 to 10 percent of the total births worldwide. Premature newborns have a high mortality rate and often require continuous rehabilitation treatment.

Many studies show that PTB, following early labor and early rupture of the amniotic membrane, was caused by ascending genital tract infection with bacteria. However, no study has found a method to detect PTB early and prevent it.

Professors Kim Young-ju (left) and Park Sun-wha at the Department of Obstetrics and Gynecology of the Ewha Womans University Medical Center said they predicted the risk of preterm birth by analyzing bacterial risk scores in cervicovaginal fluid with machine learning.
Professors Kim Young-ju (left) and Park Sun-wha at the Department of Obstetrics and Gynecology of the Ewha Womans University Medical Center said they predicted the risk of preterm birth by analyzing bacterial risk scores in cervicovaginal fluid with machine learning.

 

The research team at EUMC collected CVF from pregnant women in their second trimester and evaluated candidate bacteria cited in previous studies as predictors of PTB qualitatively and quantitively. The researchers also analyzed the difference between PTB and term birth and created a prediction model.

The results showed that Lactobacillus iners and Ureaplasma parvum were major bacteria that affected the prediction model. The researchers could predict PTB up to 72 percent. If white blood cell data were included, the prediction rate went up to 77 percent.

The research team at EUMC is also conducting a joint study with D&P Biotech, a diagnostics company that discovers new biomarkers and develops AI diagnostic algorithms.

Kim, the study's corresponding author, said combining various causes of PTB in a bacterial risk model could lead to a better prediction model.

Park, the first author, said she treated many high-risk mothers with a high risk of PBT because she was working at the EUMC, a tertiary hospital. “If doctors can take preventive measures knowing the cause of PTB, they will provide more effective treatment,” she said.

The study was published in the recent issue of the American Journal of Reproductive Immunology (AJRL).

 

Copyright © KBR Unauthorized reproduction, redistribution prohibited