ANALYSIS OF CHALLENGES IN FETAL ULTRASOUND SEGMENTATION

In this research paper the results of the methods submitted for the Challenge in Fetal Ultrasound Segmentation will be analysed. The contest set out to compare and analyse the image segmentation algorithms for ultrasound images this was an automated process involving the process of segmentation of foetal anatomical structures from 3D fetal ultrasound images at various gestational ages to represent evidence encountered in direct medical environments. Four different sub-challenges were suggested, as a feature of the object of concern assessed in clinical practise. Five teams took part in the head and leg challenges, while two teams also participated in the thigh competition. Nobody had attempted pelvic and whole-fetal dissection. Respondents in the challenge were tasked with two main categories and had been asked to apply the segmented object measurements as well as their segmentation data. Several kinds of studies were conducted with a wide sample and the findings were compared. Many experts with differing subdisciplines (three for the main subchallenge and two for the subchallenge fibula), manually delineated the objects of interest to identify the basic truth used within the assessment process. For the head sub-challenge, some groups provided similar results to manual delineations, which could probably be used in clinical settings. The femur sub-challenge has a less complicated segmentation issue and relied more on the shape of the femur's surface.