Endocardium Segmentation Based On 3D Ultrasound Image

In modern times, more and more people are suffering from heart disease, including, cardiomegaly, cardiac insufficiency, etc. Lack of blood supply is the main cause of these symptoms. The common appearance of heart disease is the abnormal systole of ventricle. Ultrasonic examination is always the first choice in physical examination. The reason is that it is cheaper than MRI without harm and loss, The cardiogram is easier to find the pathological changes. So the research of endocardium segmentation becomes important.Traditional endocardium segmentation algorithms use geometrical model. In these algorithms chambers are divided into some regular geometrical shapes, and calculate size of chambers by summing up all the geometrical shapes. Such is an easy way but just effective to healthy hearts.The aim of this paper is to provide a reliable and fully automatic endocardium segmentation method. Firstly, our method enhances the ultrasound images according to its intrinsic feature, including denoising and blurring, and finds the gray level ranges for chambers and myocardium using gray histogram. Secondly, while segmenting endocardium, our method uses the chamber area found in the previous step as the initial area for fast marching method, and segment endocardium conforming to several conditions and restrictions. Finally, the segmentation result of central slice is applied to other slices to avoid fault segmentation. Our method totally depends on features of 3D ultrasound images which assure that our result is able to provide credible clinical suggestion.The experimental result shows that our method achieves better result than traditional geometrical methods. Our accuracy rate is near 93% and has promise for clinical purpose.