Quantitative Analysis Of Left Ventricular Function Based On Two Phases MSCT Data

Cardiovascular diseases (CVDs) continue to be the number one cause of death globally. Functions evaluation of left ventricle (LV), especially ejection fraction (EF) and mass, is the significant predictor of CVDs, and can provide accurate information of clinical cardiovascular disease treatment and prognosis.So far, the research of clinical application of Multi-Slice CT mainly concentrates on coronary artery disease diagnosis, less refers to cardiac function analysis and pathological diagnosis. Because of the accurate positioning, the capacity of MSCT in cardiac disease diagnosis is being accepted by clinicians. Taking the place of extremely time-consuming manual segmentation, accurate extraction of the cavity and myocardium of LV is the key step for analyzing heart functions quantitatively. And the image segmentation of cardiac and other soft tissue has been one of the diffcult problems unresolved in image processing field.In this thesis, several typical denoising algorithm first studied according to the characteristic of dataset, and experiments and contrast analysis are realized, especially, the wavelet denoising algorithm is emphasessed. The preprocessed images are obtained which are denoised cardiac CT dataset.Aiming at the difficulty of cardiac MSCT segmentation, an improved robust semi-automated approach is presented for segmentation of cavity and myocardium from 3D cardiac multi-slice CT (MSCT) dataset. Based on random walks, a novel seeds selection method composed of region growing technique and morphological operation is introduced to locate and identify the cavity and myocardium of LV. The seeds selection method can obtain accurate labels in 3D cardiac MSCT dataset, and deliver the seeds to the next label precisely.At the same time, 6-connected lattice topology has been applied in the random walker algorithm to make sure the consecutive segmentation result among slices in 3D dataset. And Conjugate Gradient method has also been applied in the random walker algorithm to solve the large sparse linear equation system, in order to promote the segmentation performance of 3D dataset. The consecutive result of 3D reconstruction and similarity comparation of results between the manual segmentation and the proposed segmentation, show the efficacy and advantage of our method for the segmentation of LV in MSCT images. At last, the obtained segmentation results are applied in the computation and estimation of various cardiac indexes of left ventricle. The key characteristic parameters are extracted effectively, and the good evaluation results are obtained. The cardiac indexes of left ventricle can be used to prevention and detection of coronary heart diseases and other cardiovascular diseases. The research work has obvious clinical application value.This research was sponsored by Nature and Science Foundation of China(Project No.60771007).