Placental T2-weighted MRI-based radiomics-clinical nomogram to predict postpartum hemorrhage of placenta previa.

Abstract

Background: Placenta previa is an obstetric complication related to severe maternal morbidity and mortality. Magnetic resonance imaging (MRI) can be used for the preoperative evaluation of postpartum hemorrhage.

Purpose: To investigate the value of MRI-based radiomics analysis in predicting postpartum hemorrhage among pregnant women with placenta previa.

Material and methods: Preoperative T2-weighted MRI and related clinical data of 371 patients were retrospectively collected, and these patients were randomly allocated into two subsets: the training dataset (n = 260) and the validation dataset (n = 111). The logistic regression (LR) classifier was used for the development of the radiomics model and the calculation of the radiomics score (Radscore).

Results: A total of eight radiomics features and five clinical features were selected for model development. The area under the receiver operating characteristic curve (AUC) of the radiomics model in the training and validation datasets were 0.929 (95% confidence interval [CI] = 0.891-0.957) and 0.914 (95% CI = 0.846-0.959), respectively. Combined with clinical factors, nomograms demonstrated improved diagnostic efficacy, with an AUC of 0.968 (95% CI = 0.939-0.986) in the training dataset and 0.947 (95% CI = 0.888-0.981) in the validation dataset.

Conclusion: The MRI-based model has certain value in predicting postpartum hemorrhage in pregnant women with placenta previa.