Fully Automated Segmentation And Quantification Of Magnetic Resonance Image

A large number of images in magnetic resonance imaging are needed for quantification analysis. These images are traditionally quantified manually, which is time-consuming and subjective. Methods of fully automated segmentation and quantification are studied, including two projects:1.Automated Segmentation of cardiac ventricles in cine Cardiac Magnetic Resonance Image; 2.Fully Automated Measurements of Longitudinal and transverse Relaxation Times of MRI Contrast Agent1. Automated Segmentation of cardiac ventricles in cine Cardiac Magnetic Resonance ImageHeart disease has been the leading cause of morbidity and mortality in the recent years. Accurate detection and evaluation of this disease are of major importance for evaluating disease and determining appropriate courses of treatment. Segmentation of left and right ventricles in cine cardiac magnetic resonance (CMR) image is essential for clinical application, which can be used for the evaluation of functional parameters such as cardiac ventricle volumes, stroke volume, ejection fraction, filling curve.Traditionally, segmentation of cardiac ventricles at end-diastolic and end-systolic phases is manually traced by radiologist using software, which is time-consuming and subjective and ignores a lot of images. Although stroke volume and ejection fraction can be quantified, no dynamic cardiac info can be acquired in a cardiac period. Herein, automated segmentation method is essential for cardiac functional parameters analysis.Various automated and semi-automated algorithms for segmenting the left ventricle have been reported in recent ten years. Automatic segmentation of left ventricle of cardiac cine MR images has been a well-recognized, yet an elusive goal since the early days of cardiac MRI. Traditional segmentation methods including thresholding, regional growth and clustering require user interactions. Compared to the LV, the more irregular geometrical shape of the RV (PTMs) makes it more difficult to segment RV. To date, there is no fully automated image segmentation software of cardiac ventricles available for clinical practice. We have developed a novel approach to segment cardiac cine MR images using the basic knowledge of image signal, without making any contour or geometric based assumptions of the heart. Accordingly, this research consists of the following three specific aims:(1) We propose an LV segmentation method that addresses these challenges in a manner that is fully automated, based on the spatiotemporal continuity (LV-FAST). The LV-FAST algorithm was designed using an iteratively decreasing threshold region growing approach to segment first from the mid-ventricle to the apex, until the LV area and shape discontinued, and then from mid-ventricle to the base, until less than 50% of the myocardium circumference was observable. Region growth was constrained by LV spatiotemporal continuity to improve robustness of apical and basal segmentations.(2) The outflow tracks near the basal slices are not well defined in short-axis view due to the 6-10 mm thick slice used in current clinical protocol. We proposed an LV Automated Segmentation based on Long-axis information and "Neck" Detector in Cine Cardiac Magnetic Resonance Imaging(LV-ASLAN).Long axis images(2- and 4-chamber view images) are combined together through a linear transformation from the original 2D section coordinate to the 3D registration using DICOM header information for apical and basal LV definition. For more accurate segmentation, we propose to develop a neck detection method that can automatically edit these false linkages with bright epicardial fat and other structures. These connecting LV blood to linkages typically are of the form of a thin neck connection the chamber blood and the outside structure.(3) We proposed a Fully Automated Image-Driven Localization of Right Ventricle in Cine CMRI (RV-AID). Due to the low signal noise ratio and contrast, as well as thin and unsmooth myocardium, anisotropic diffusion algorithm is used for the enhancement of image while keep the information of edge. Comprehensive algorithms like OTSU, mathematical morphology were implemented for automated localization of RV based on the prior knowledge of position that right ventricle is at the left of right ventricle2. Fully Automated Measurements of Longitudinal and transverse Relaxation Times of MRI Contrast AgentThis section was inspired by the automated segmentation of cardiac ventricle. Contrast agent for tissue contrast enhancement is widely used in MRI. Measurements of longitudinal and transverse relaxation times (T1 and T2) of solutions of MRI contrast agent (samples) are essential for their characterization and comparison in clinical diagnosis and biomedical applications. Some free software like MRIcro can be used to measure T1 and T2 by manually select ROI, which is time consuming and subjective. Furthermore, other software like Origin need be used for fitting. Herein, a fully automated software-based measurement method was proposed that located tubular samples, draws ROIs and calculates T1 and T2 relaxation times without user intervention.