The new 3D-based fetal segmentation and visualisation method

초록

Objectives The use of three-dimensional (3D) ultrasound imaging tools has increased the accuracy of congenital malformation diagnosis and provided additional information. However, the ability to visualise a particular structure using 3D ultrasound is entirely dependent on the quality of the initial image, fetal position and surrounding obstacles. Therefore, the purpose of this study was to develop a new 3D segmentation technique to differentiate fetus and intrauterine structures. Methods We developed a new 3D-based volumetric method for visualisation of intrauterine structures using automatic segmentation and multi-volume rendering algorithm. Data set consisted of fetal images tagged at Korea University Guro Hospital for model development and internal validation. First, volumetric data were obtained and segmented within hundreds of milliseconds using a algorithm to differentiate the fetus, amniotic fluid, placenta, uterus, and umbilical cord. Finally, a 3D rendering technique was applied to visualise each part with different colours and levels of transparency. Results In the first trimester, each structure was easily identified and separated, enabling confirmation of fetal anatomy. Especially in multifetal pregnancies, parts that were obscured by the uterus or other fetuses were separated, allowing for differentiation of fetal position and relationships with the placenta and umbilical cord. Beyond the second trimester, it was possible to obtain distinct images of facial features, extremities, and external abnormalities even when they were obscured by other structures. Conclusions The newly developed 3D segmentation algorithm overcomes the limitations of previous 3D ultrasound techniques that made it difficult to obtain clear images due to fetal position and surrounding structures.