On Monitoring The Cardiac Function Based On The Segmentation Of Real-time Echocardiography

At present, real-time echocardiography has become a routine clinical tool. Measuring left ventricular systolic function by using echocardiography is the commonly used clinical noninvasive method in the current examination of left ventricular function. Image segmentation plays an important role in the quantitative and qualitative analysis of the medical ultrasound image and exerts a direct influence on the subsequent analysis and processing. Through image segmentation, the subject, such as the left ventricular, can be split out from a complex background and marked in a specific way, which facilitates doctors'observation and diagnosis.Considering the noisiness, fuzzy boundary and the uneven distribution of gray, based on conventional parallel computing C-V, this paper introduces the information of the local area into the C-V model, utilizes the average edge energy of the zero level set curve to control the speed of its evolution and proposes an improved C-V model segmentation algorithm of supersonic image on the basis of the information of the local area. Compared with conventional C-V model, this algorithm boasts a strong anti-noisiness capability, suitable to process the supersonic images that are unevenly gray and have vague contour goals. This better resolves the problem of contour extraction and segmentation of the weak edge of the ventricular in supersonic echocardiography. The shortages of the algorithm include a large number of iterations and a heavy computing load.CUDA is a parallel programming model and software environment based on CPU+GPU. It boasts an excellent performance and high compatibility. It combines CPU and GPU in calculation and is highly efficient and cost-effective. In order to realize the speedy and real-time segmentation of echocardiography, this paper devises and puts into use the C-V model image segmentation algorithm using CUDA programming model based on CPU and GPU and achieves the segmentation of the two-dimensional echocardiography.Based on the work above, the feasible segmentation of supersonic echocardiograph based on C-V model achieved by CUDA in this paper not only suppresses the problems of noisiness and pseudo-edge and attains accurate segmentation of the ventricular contour information but also improves the solving rate of algorithm. This is of great significance for clinical use. The follow-up work of this paper can study reducing the number of iterations in segmentation algorithm to further improve the solving rate or conduct segmentation studies concerning other medical supersonic images based on CUDA.