EP01.11: Application of fetal lung maturity grading model in predicting neonatal respiratory morbidity
Year: 2024
Type: conference-abstract
Abstract: This study aimed to apply artificial intelligence-based ultrasound radiomics methods to establish a fetal lung maturity grading model (FLM-GM) and apply the model to predicting neonatal respiratory morbidity (NRM). Ultrasound images of fetal lungs of singleton pregnancies at 2140 weeks of normal pregnancy were retrospectively collected. Each gestational week collected 500-1000 standard four-chamber cardiac images. The images were divided into 7 classes based on the gestational week: class I: 20 to 24 + 6 weeks, class II: 25 to 28+6 weeks, class III: 29 to 30+6 weeks, class IV: 31 to 32+ 6 weeks, class V: 33 to 34+ 6 weeks, class VI: 35 to 36+ 6 weeks, class VII: 37 to 40+6 weeks. The ratio of the training set to the validation set is 4:1. The independent validation set consisted of 100 ultrasound images of each grade. Deep learning convolutional neural network was used to extract the characteristics related to the maturity of fetal lung ultrasound images at different gestational weeks, and established FLM-GM. Then prospectively collected ultrasound images of fetal lungs at 24 to 40+6 weeks of gestation within 72 hours of delivery before, and evaluated with the FLM-GM. At delivery, perinatal outcomes and the occurrence of NRM were followed up. A total of 14,973 ultrasound images of fetal lungs were collected from 4840 cases of normal singleton pregnancies at 21~40 weeks of gestation and established the FLM-GM. The accuracy of the model evaluation independent validation set was 0.72, the sensitivity was 0.70, and the specificity was 0.95. The cutoff value of the FLM-GM for predicting NRM was class VI, sensitivity was 0.80, specificity was 0.79, AUC: 0.85, P < 0.001. The cutoff value of the FLM-GM for predicting RDS was class V, sensitivity was 0.98, specificity was 0.86, AUC: 0.94, P < 0.001. The fetal lung maturity grading model had good accuracy in predicting NRM. It had great clinical value for prenatal non-invasive assessment of fetal lung maturity.
DOI: 10.1002/uog.28026
Author: Q. Wu
Language: English
Subfield: Pulmonary and Respiratory Medicine
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W4402358109