Ventricle Segmentation And Quantitative Analysis Of Cardiac Function Based On Anatomical Features Of The Heart

With the continuous development of medical technology,the death rate of human epidemics has gradually declined,and chronic diseases have gradually become the main cause of death for modern people.As a common chronic disease,early quantitative diagnosis and risk assessment of cardiovascular and heart disease play an important role in prolonging life expectancy and improving the life quality of patients.As a non-invasive means of observation,medical imaging can obtain the internal anatomical structure of the patient's heart,which provides the possibility for quantitative assessment of cardiac function.Cardiac medical image segmentation method is the basis for quantitative assessment of cardiac function.The purpose of segmentation is to obtain accurate internal structural boundaries of the heart.The gray level inhomogeneity in the medical image of the heart and the overlapping of surrounding tissue are the hardness of medical image segmentation.So the performance of the algorithm affects the segmentation results of cardiac medical images.In this thesis,we focus on the needs of the assessment of cardiac function in the clinical study of two-dimensional and three-dimensional data segmentation algorithm for cardiac medical images.Based on the segmentation,we use static and dynamic cardiac function parameters to analyze cardiac function quantitatively.And we verified the algorithm performance in the open data set.The main research and innovation of this paper are as follows:The imaging principle and preprocessing method of CT and MR images are studied.Hough transform is used to detect position of heart in the short-axis images of cardiac MR images.The image is automatically segmented based on the detected center coordinates using the region grow algorithm.And the obtained rough segmentation results were used as the initialization of level set.The anatomical geometry of the heart is studied,and the anatomical features are refined and added to the level set segmentation algorithm.For the anatomical characteristics of uniform myocardial thickness changes,the bi-level model of the level set algorithm is used to segment the endocardium and epicardium at the same time,constraining the distance between the internal and external borders.For the anatomical characteristics of convexity of the left ventricle at the short-axis images,a periodic convexity-preserving mechanism is added during the level set segmentation to ensure that the segmentation is convex.Finally,a geometrical decomposition method of the ventricle is proposed to decompose the left and right ventricles into two convex structures,and the segmented results of the entire ventricular region are obtained by combining the two convex structures.The phenomenon of slice misalignment in cardiac magnetic resonance data is analyzed.To solve the slice misalignment,a least-squares method is used to fit the center of gravity of the cardiac region to a straight line.Finally,on the basis of accurate segmentation of the endocardium and epicarduim,the cardiac function is quantified using static cardiac function parameters and dynamic cardiac function parameters,and compared with the cardiac function parameters calculated by the expert's manual segmentation results.