RESUMO
Prenatal assessment of lung size and liver position is essential to stratify congenital diaphragmatic hernia (CDH) fetuses in risk categories, guiding counseling, and patient management. Manual segmentation on fetal MRI provides a quantitative estimation of total lung volume and liver herniation. However, it is time-consuming and operator-dependent. In this study, we utilized a publicly available deep learning (DL) segmentation system (nnU-Net) to automatically contour CDH-affected fetal lungs and liver on MRI sections. Concordance between automatic and manual segmentation was assessed by calculating the Jaccard coefficient. Pyradiomics standard features were then extracted from both manually and automatically segmented regions. The reproducibility of features between the two groups was evaluated through the Wilcoxon rank-sum test and intraclass correlation coefficients (ICCs). We finally tested the reliability of the automatic-segmentation approach by building a ML classifier system for the prediction of liver herniation based on support vector machines (SVM) and trained on shape features computed both in the manual and nnU-Net-segmented organs. We compared the area under the classifier receiver operating characteristic curve (AUC) in the two cases. Pyradiomics features calculated in the manual ROIs were partly reproducible by the same features calculated in nnU-Net segmented ROIs and, when used in the ML procedure, to predict liver herniation (both AUC around 0.85). Conclusion: Our results suggest that automatic MRI segmentation is feasible, with good reproducibility of pyradiomics features, and that a ML system for liver herniation prediction offers good reliability. Trial registration: https://clinicaltrials.gov/ct2/show/NCT04609163?term=NCT04609163&draw=2&rank=1 ; Clinical Trial Identification no. NCT04609163. What is Known: ⢠Magnetic resonance imaging (MRI) is crucial for prenatal congenital diaphragmatic hernia (CDH) assessment. It enables the quantification of the total lung volume and the extent of liver herniation, which are essential for stratifying the severity of CDH, guiding counseling, and patient management. ⢠The manual segmentation of MRI scans is a time-consuming process that is heavily reliant upon the skill set of the operator. What is New: ⢠MRI lung and liver automatic segmentation using the deep learning nnU-Net system is feasible, with good Jaccard coefficient values and satisfactory reproducibility of pyradiomics features compared to manual results. ⢠A feasible ML system for predicting liver herniation could improve prenatal assessments and CDH patient management.
Assuntos
Hérnias Diafragmáticas Congênitas , Fígado , Pulmão , Imageamento por Ressonância Magnética , Diagnóstico Pré-Natal , Humanos , Hérnias Diafragmáticas Congênitas/diagnóstico por imagem , Imageamento por Ressonância Magnética/métodos , Feminino , Reprodutibilidade dos Testes , Gravidez , Pulmão/diagnóstico por imagem , Fígado/diagnóstico por imagem , Fígado/patologia , Diagnóstico Pré-Natal/métodos , Aprendizado Profundo , Hepatopatias/diagnóstico por imagem , Aprendizado de MáquinaRESUMO
PURPOSE: To evaluate the impact of the timing of MRI on the prediction of survival and morbidity in patients with CDH, and whether serial measurements have a beneficial value. METHODS: This retrospective cohort study was conducted in two perinatal centers, in Germany and Italy. It included 354 patients with isolated CDH having at least one fetal MRI. The severity was assessed with the observed-to-expected total fetal lung volume (o/e TFLV) measured by two experienced double-blinded operators. The cohort was divided into three groups according to the gestational age (GA) at which the MRI was performed (< 27, 27-32, and > 32 weeks' gestation [WG]). The accuracy for the prediction of survival at discharge and morbidity was analyzed with receiver operating characteristic (ROC) curves. Multiple logistic regression analyses and propensity score matching examined the population for balance. The effect of repeated MRI was evaluated in ninety-seven cases. RESULTS: There were no significant differences in the prediction of survival when the o/e TFLV was measured before 27, between 27 and 32, and after 32 WG (area under the curve [AUC]: 0.77, 0.79, and 0.77, respectively). After adjustment for confounding factors, it was seen, that GA at MRI was not associated with survival at discharge, but the risk of mortality was higher with an intrathoracic liver position (adjusted odds ratio [aOR]: 0.30, 95% confidence interval [95%CI] 0.12-0.78), lower GA at birth (aOR 1.48, 95%CI 1.24-1.78) and lower o/e TFLV (aOR 1.13, 95%CI 1.06-1.20). ROC curves showed comparable prediction accuracy for the different timepoints in pregnancy for pulmonary hypertension, the need of extracorporeal membrane oxygenation, and feeding aids. Serial measurements revealed no difference in change rate of the o/e TFLV according to survival. CONCLUSION: The timing of MRI does not affect the prediction of survival rate or morbidity as the o/e TFLV does not change during pregnancy. Clinicians could choose any gestational age starting mid second trimester for the assessment of severity and counseling.