J Ultrasound Med. 2021 Mar 16. doi: 10.1002/jum.15674. Online ahead of print.
ABSTRACT
OBJECTIVES: Ultrasound (US) prediction of placenta accreta spectrum (PAS) in the first trimester may be aided by postprocessing mechanisms employing color pixel quantification near the bladder-uterine serosal interface. Our objective was to create a postprocessing algorithm of color images to identify findings associated with PAS and compare quantification to sonologist impression in prospectively obtained cine US images.
METHODS: Transverse transvaginal (TV) US color cines obtained in the first trimester as part of a prospective study were reviewed. Investigators blinded to clinical outcomes reviewed anonymized cines that were archived and labeled the bladder-uterine serosal interface. Color pixels within 2 cm of the defined bladder-uterine serosal interface were ascertained using a Python-based plugin in the Horos open-source DICOM viewer. A sonologist classified the findings as suspicious for invasion, indeterminate, or normal. Statistical analysis was performed using Wilcoxon rank-sum test, Cochran-Armitage trend test, and calculation of receiver-operating characteristic (ROC) curves.
RESULTS: Fifty-four studies met inclusion criteria. Of those, six (11%) required hysterectomy with pathologic confirmation of PAS. Women requiring hysterectomy had a significantly higher color Doppler pixel area than those not requiring hysterectomy (P = .0205). A significant trend was identified in the sonologist impression of invasion (P = .0003). ROC’s comparing sonologist impression to Doppler color imaging areas were comparable (P = .054).
CONCLUSIONS: Color Doppler mapping in the first trimester showed an increase in color pixel area near the bladder-uterine serosal interface in women requiring cesarean hysterectomy with histologically confirmed PAS at time of delivery, compared to women without hysterectomy or pathologic evidence of PAS.
PMID:33724510 | DOI:10.1002/jum.15674
