OBJECTIVE: To test an objective ultrasound marker for diagnosing the presence and severity of abnormally invasive placenta. METHODS: Women at risk of abnormally invasive placenta underwent a three-dimensional power Doppler ultrasound scan. The volumes were examined offline by a blinded observer. The largest area of confluent three-dimensional power Doppler signal (Area of Confluence [Acon], cm2) at the uteroplacental interface was measured and compared in women subsequently diagnosed with abnormally invasive placenta and women in a control group who did not have abnormally invasive placenta. Receiver operating characteristic curves were plotted for prediction of abnormally invasive placenta and abnormally invasive placenta requiring cesarean hysterectomy. RESULTS: Ninety-three women were recruited. Results were available for 89. Abnormally invasive placenta was clinically diagnosed in 42 women; 36 required hysterectomy and had abnormally invasive placenta confirmed histopathologically. Median and interquartile range for Acon was greater for abnormally invasive placenta (44.2 [31.4–61.7] cm2) compared with women in the control group (4.5 cm2 [2.9–6.6], P<.001) and even greater in the 36 requiring hysterectomy (46.6 cm2 [37.2–72.6], P<.001). Acon rose with histopathologic diagnosis: focal accreta (32.2 cm2 [17.2–57.3]), accreta (59.6 cm2 [40.1–89.9]), and percreta (46.6 cm2 [37.5–71.5]; P<.001 analysis of variance for linear trend). Receiver operating characteristic analysis for prediction of abnormally invasive placenta revealed that with an Acon of 12.4 cm2 or greater, 100% sensitivity (95% confidence interval [CI] 91.6–100) could be obtained with 92% specificity (95% CI 79.6–97.6); area under the curve is 0.99 (95% CI 0.94–1.0). For prediction of abnormally invasive placenta requiring hysterectomy, 100% sensitivity (95% CI 90.3–100) can be obtained with an Acon of 17.4 cm2 or greater with 87% specificity (95% CI 74.7–94.5; area under the curve 0.98 [0.93–1.0]). CONCLUSION: The marker Acon provides a quantitative means for diagnosing abnormally invasive placenta and assessing severity. If further validated, subjectivity could be eliminated from the diagnosis of abnormally invasive placenta. LEVEL OF EVIDENCE: II
21 November 2011 in Ultrasound in Obstetrics & Gynecology
Volumetric placental measurement using 3-D ultrasound has proven clinical utility in predicting adverse pregnancy outcomes. However, this metric cannot currently be employed as part of a screening test due to a lack of robust and real-time segmentation tools. We present a multiclass (MC) convolutional neural network (CNN) developed to segment the placenta, amniotic fluid, and fetus. The ground-truth data set consisted of 2093 labeled placental volumes augmented by 300 volumes with placenta, amniotic fluid, and fetus annotated. A two-pathway, hybrid (HB) model using transfer learning, a modified loss function, and exponential average weighting was developed and demonstrated the best performance for placental segmentation (PS), achieving a Dice similarity coefficient (DSC) of 0.84- and 0.38-mm average Hausdorff distances (HDAV). The use of a dual-pathway architecture improved the PS by 0.03 DSC and reduced HDAV by 0.27 mm compared with a naïve MC model. The incorporation of exponential weighting produced a further small improvement in DSC by 0.01 and a reduction of HDAV by 0.44 mm. Per volume inference using the FCNN took 7-8 s. This method should enable clinically relevant morphometric measurements (such as volume and total surface area) to be automatically generated for the placenta, amniotic fluid, and fetus. The ready availability of such metrics makes a population-based screening test for adverse pregnancy outcomes possible.
1 December 2015 in Ultrasound in Medicine & Biology