Hybrid Segmentation Approach to Segment Fetal Cardiac Chambers of Ultrasound images

Abstract

Fetal echocardiography uses ultrasound technique to view heart of the baby even when the baby is in the mother’s womb. Obstetricians refer the patients to undergo this procedure if they find any suspicious conditions of having defects in the fetal heart. This test is done usually during 2nd trimester. Fetal ultrasound measurements are one of the most important factors for high quality obstetrics health care in order to estimate the Gestational age, exact delivery date and growth of the fetus.The standard 2D ultrasound imaging technique examines the fetal heart in different views. The information related to the cardiac size, structure, rhythm and movement can be obtained in four chamber view. Various congenital defects can be visualized by the examination of fetal cardiac chambers. It is very difficult to locate the cardiac chambers and do the relevant measurement, hence it is challenging work to researchers of biomedical community.A semi automated method of segmentation has been proposed in this experimental study to segment the fetal heart chambers using Possibilistic c-means clustering technique. The Ultrasonic images of fetal heart having four – chambers in apical view was used in order to do the study with simulation. The input fetal cardiac frame was denoised and enhanced. The enhancement of the input image was carried out by converting it in to gray image. This was then converted to binarized image. On the selected chambers the region growing technique followed by PCM was used to segment the cardiac chambers. Finally the region properties were used to measure the ventricles that is the left ventricle and right ventricle of the fetal heart and LV/RV estimation is done based on standard values.The ratio of the width of left ventricle and right ventricle were calculated which closely matched with the theoretical width considered as the standard for different gestation week. The proposed semi – automated technique need to be validated before being used for assessment as the clinical tool by large datasets.