OP10.09: A novel semi‐automated (SA) technique for 3D ultrasound measurement of placental volume
Year: 2010
Type: conference-abstract
Abstract: To develop a new SA volumetric technique for segmentation of the placenta that performs concordantly or better than manual delineation or the current state-of-the-art SA method, VOCAL (GE Medical Systems, Milwaukee, WI, USA). 3D US scans of the placenta were performed on 10 women with anterior placental implantation and a maternal BMI < 25. Two observers (GS, SC) performed measurement of placenta volume manually, using VOCAL and using the new SA volumetric method on the same scan. Measurement was repeated three times for each technique. Statistical analysis was performed on the volume results. Measurement difference was illustrated by Bland-Altman plot between both automated measures and manual. Evaluation of inter/intra observer variability was performed using intra-class correlation (ICC), having tested for normality of the data using the Lillefors test. For Observer 1, mean difference (MD) between new method and manual was − 0.776 ml (LOA 14.03, − 15.58) and between VOCAL and manual − 1.5329 ml (LOA 9.29, − 12.36). For Observer 2, MD between new method and manual was − 6.20 ml (LOA 10.64, − 23.05) and between VOCAL and manual − 9.59 ml (LOA 11.17, − 30.34). Table 1 shows the ICC values for inter/intra-observer variability for placental volume. Results show that the new method is concordant with manual segmentation based on ICC values for intra/inter-observer variation and mean difference. The new method would appear to produce concordant results to manual segmentation and outperforms VOCAL in aspects of observer repeatability/agreement. This accurate, time-efficient method to calculate placental volume has the potential to become a promising research tool.
DOI: 10.1002/uog.8020
Authors: G. N. Stevenson, S. L. Collins, L. Impey, A. Noble
Institutions: University of Oxford, John Radcliffe Hospital
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W1972277027
