OP19.01: Magnetic resonance imaging of placenta accreta spectrum disorders: radiomics analysis correlates to surgical and pathological outcome
Year: 2019
Type: conference-abstract
Abstract: To investigate textural analyses as the radiomics in magnetic resonance (MR) imaging of the placenta in predicting of placenta accreta spectrum disorders (PAS) requiring Caesarean hysterectomy in a high-risk population. We performed a retrospective review of 51 pregnancies with prior Caesarean delivery referred for MRI from 2014-2018 because of sonographic suspicion for PAS. Placental regions of interest (ROIs) were generated on T2-weighted, diffusion-weighted images (DWI) and the DWI-derived apparent diffusion coefficient (ADC), based on definitions of areas of placenta in proximity to and remote from previous surgical incision sites. T-tests and chi-square analyses between the primary outcome (hysterectomy vs. no hysterectomy) were performed. 13 Haralick texture features calculated from gray-level co-occurrence matrixes were extracted from manually drawn placental ROIs within each of three MR acquisitions. Univariate logistic regression models tested the association between the primary outcome and the texture features. Of 51 pregnancies at risk for PAS, 38 required Caesarean hysterectomy (75%), with excellent correlation between need for hysterectomy determined surgically and pathological confirmation of PAS in the hysterectomy specimen [κ=0.81(0.62, 1)]. Of the 13 Haralick texture features within each of 3 acquisition groups, significant differences were seen in 21 of 39 parameters comparing placental ROIs in proximity to incision scar(s) to those ROI's remote from scar(s). Stepwise selection algorithm indicated that the combination of several texture features (T2WFcm.joint.entr, T2WFcm.info.corr.2, T2WFcm.joint.var, ADCFcm.sum.avg and ADCFcm.info.corr.2) gave the highest leave-one-out-AUC of 0.87 (0.80,1). We identified textural features on MR images of the placenta in the region of the prior uterine scar that differentiated pregnancies with invasion severe enough to require Caesarean hysterectomy from those that did not, suggesting predictive capabilities of these radiomics features.
DOI: 10.1002/uog.20835
Institution: The University of Texas Southwestern Medical Center
Language: English
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2978706392