Endocardiac Boundary Extraction Technology And Application In Echocardiogram

Cardiovascular disease is among the three top diseases of highest death ratio for people. Centrifugal ventricle dilation and congestive heart-failure are the basic pathology characteristic of dilated cardiomyopathy(DCM). It is diagnosed by the computation of left ventricle volume and ejection fraction in clinic. The extraction of endocardiac boundary is the crucial basic of quantificational researching 2D echocardiogram. Due to the speckle noise in ultrasound imaging, it is very difficult to detect the endocardiac boundary automatically and accurately in echocardiogram. The study in this thesis has been carried out mainly in three aspects and some innovation methods and preferable results were obtained.Active contour model(ACM) is a preferable semi-automatic boundary detection algorithm applied in complex structure and noisy image. But there is false contour in almost all ACM course of the iterative, especially in the boundary region of great curvature, and result in the failure of the boundary detection. The problem of false contour has been lucubrated in this thesis and the rapid algorithm of removing it was developed. The obtained result has been showed that the contour can approach the actual edge.Traditional endocardiac boundary detection is based on the gray scale information representing anatomic structure. Doppler tissue imaging is a newly analytical technology of ventricular wall movement. It can represent the tissue velocity through the color coding. The multi-characteristic has been introduced to detect the boundary, including the anatomic structure, Doppler tissue velocity and the combine between the both. The exact boundary has been obtained by this method.After the 2D boundary was obtained, the enclosing region area was calculated and the quantitative analysis of cardiac function was carried out. Furthermore, the application of the method has been expanded in 3D domain. The interpolation andcoordinate transform has been developed in rotating scan image. After that, the slice image of the 3D data was performed the forenamed boundary detection. At last the 3D surface of the endocardium has been rendered based on the boundary result in a cardiac cycle. Accordingly, the cardiac function in 3D has been quantificationally analyzed. The Result can represent patient cardiac function and also validate the veracity and feasibility of the method in this thesis.