Flow Field Information Extraction And Analysis Based On Contrast Echocardiographic Images

Cardiovascular disease is the major threat to human health. It is also one of thoseleading causes of the death in developing countries. Currently, with the help of medicalcontrast ultrasound imaging technology, the quantitative analysis of the heart flow fieldhas been the central issue of Cardiovascular research. Among them, left ventricular isthe most important chamber and undertakes the major function of blood transfer. Theproperties of left ventricular blood flow field represent the efficiency of heart function.So it is very important to precisely measure and depict the left ventricular blood flowfield. The left ventricular abnormal in shape and motion could be regard as essentialevidence for clinical diagnosis. It could help diagnosis the existence and location ofheart disease. Hence, it is significant to research on algorithm that extract leftventricular blood field based on contrast ultrasound image sequence and the quantitativeindexes to evaluate the heart function.Taking the practical application of the medical diagnosis into consideration, thisthesis introduces the background and objective of the research. Analyze the currentresearch situation of particle image velocimetry, vector field modification andevaluation indexes. The thesis mainly focuses on the research of contrastechocardiographic particle image velocimetry technology, flow vector fieldmodification technology and left ventricular evaluation indexes extraction. The researchincludes the following four parts:First, research the single frame vector filed extraction algorithm. After the ROIarea segmentation pre-processing of the images based on level set algorithm, analyzethe particle image velocimetry based on grey level cross correlation and update thepatch matching similarity measurement, which is based on the noise model ofultrasound image and Bayesian estimation theory. The experiment results prove that thenovel matching algorithm is more suitable to the ultrasound images compared with thetraditional matching algorithm.Second, the existence of the invalid vectors in the flow field will deteriorate theaccuracy of the extraction of the evaluating indexes in the future, so a flow fieldmodifying algorithm is needed. Then this thesis researches on the spatial-temporalmodification of the vector field. First is the single modification algorithm based onHessian kernel, then the spatial-temporal modifying algorithm based on threedimensional anisotropic diffusion, which utilizes the redundant information contained inthe temporal axis. The experiments of simulated images, simulated vector field and vivoimages indicate the efficiency of the modifying algorithm.Third, to help doctors treat the patients, several quantitatively indexes are needed. Then the thesis researches on the left ventricular function quantitative evaluation. Theindexes were based on the evaluation of Left ventricular cavity blood flow and Leftventricular myocardium motion field. In details, the indexes include vector fields seriesfrequency energy distribution, flow field evaluation based on differential operator andmyocardium motion evaluation. Then analyze the correlation between the quantitativeindexes and the classification based on doctors’ experience, the results show that thisresearch will do favor in assisting the diagnosis in the future.