Application of ultrasound-based radiomics technology in fetal lung texture analysis in pregnancies complicated by gestational diabetes or pre-eclampsia.

OBJECTIVES: To analyze and compare the fetal lung texture of gestational diabetes mellitus (GDM), pre-eclampsia (PE) and normal pregnancy at different gestational age using ultrasound-based radiomics technology. METHODS: 430 high-throughput features per fetal lung image were extracted from 548 fetal lung ultrasound images (4-cardiac-chamber view) of 548 pregnant women. Images were obtained during routine ultrasound examinations between 28 and 41 weeks of gestation before delivery. A standard machine-learning model based on ultrasound-based radiomics technology was composed of feature extraction, and a regression model was used to evaluate the relationship between texture features, GDM/PE, and the gestational age. RESULTS: Fetal lung ultrasound images were divided into 4 groups: GDM group (108 cases), GDM + PE group (25 cases), PE group (71 cases) and normal group (344 cases). The overall performance of the GDM and PE prediction model by fetal lung image texture analysis was superior to that of the gestational age prediction model, with the average AUC 0.95-0.99, sensitivity 79.2-97.1% in the validation set and 74.5-91.3% in the independent test set, specificity 79.8-94.3% in the validation set and 75.7-88.4% in the independent test set, accuracy 81.0-95.3% in the validation set and 80.6-86.4% in the independent test set. CONCLUSIONS: The fetal lungs associated with GDM/PE and gestational age could be distinguished by ultrasound-based radiomics technology. This supports further research to explore the use of this technology to predict neonatal respiratory complications in women with PE, GDM or their combination. This article is protected by copyright. All rights reserved.