Semi-Automated Techniques for Segmentation of Fetal Ultrasound Images

The process of segmenting fetal ultrasound images has proven to yield many informative results about fetal health, growth, and other related conditions. In specific, segmentation of the placenta is of great importance as the corresponding placental volume can help predict and quantify many of these fetal attributes. There is currently a lack of placental segmentation in the bio-medical field due to the tedious nature of this task as well as the time and monetary constraints. The purpose of this project is to develop a fully-automated segmentation algorithm using machine learning. That result would then be used to analyze and observe ultrasound data for all related medical purposes. So far, we have developed a semi-automated algorithm in MATLAB that will be used as a basis for training the machine learning algorithm. This algorithm uses advanced mathematics to predict which region of the image corresponds to the placenta and can be used to segment other fetal organs as well. This poster presentation will showcase the algorithm and supply a more in depth explanation as to how it works.