Integration of two-dimensional microvascular flow imaging and ultrasound scoring for prediction of placenta accreta...

This study evaluates the diagnostic performance of two-dimensional (2D) microvascular flow imaging combined with standardized ultrasound markers for predicting placenta accreta spectrum disorders in pregnancies complicated by placenta previa or low-lying placenta.

This prospective observational study included 61 women with placenta previa or low-lying placenta who underwent cesarean delivery between 32 and 38 weeks of gestation. All participants underwent standardized 2D microvascular flow imaging using the Samsung Hera W10 Elite system. The microvascular vascularization index was quantified as the percentage ratio of color-coded vascular pixels within a defined placental region of interest. Conventional Doppler indices and standardized ultrasound markers associated with placenta accreta spectrum were recorded. Placental invasion was confirmed histopathologically in 23 patients. Statistical analyses included Mann-Whitney U-testing, receiver operating characteristic (ROC) analysis, 95% confidence interval estimation, pairwise ROC comparison using the DeLong method, and exploratory internally validated machine-learning classification models. To avoid conceptual overlap and double counting, the total placenta accreta spectrum (PAS)score was excluded from the corrected machine-learning models. Intraobserver reproducibility of repeated MVI measurements was assessed using the intraclass correlation coefficient.

The vascularization index was significantly higher in PAS cases than controls (47.65 ± 9.35 vs. 26.58 ± 8.93, P < 0.001). Conventional Doppler parameters did not differ significantly between groups. ROC analysis identified the vascularization index as the strongest single quantitative predictor of PAS, with an AUC of 0.985, sensitivity of 95.7%, and specificity of 97.4% at a cut-off value of 37.11. The corrected Placental Invasion Risk Score, calculated without the total PAS score, showed excellent diagnostic performance, with an AUC of 0.998. Exploratory machine-learning models excluding the total PAS score demonstrated improved discrimination after inclusion of the vascularization index, with logistic regression achieving an AUC of 0.990. Intraobserver reproducibility of repeated MVI measurements was excellent, with an intraclass correlation coefficient (ICC) of 0.977. For prediction of cesarean hysterectomy, the vascularization index also demonstrated excellent discriminatory performance at an optimal cut-off value of ≥53.51.

The 2D MV-Flow vascularization index is a strong quantitative biomarker for predicting PAS in pregnancies complicated by placenta previa or low-lying placenta. When combined with individual ultrasound markers, but not redundantly with the total PAS score, it may improve antenatal risk stratification and preoperative planning. However, the exploratory machine-learning findings require external validation in larger multicenter cohorts.