Research On Modeling And Simulation Of Diffusion Magnetic Resonance Imaging Of Human Cardiac Fibers

Diffusion magnetic resonance imaging (dMRI) is able to measure indirectly thetissue structures by detecting the diffusion of water molecules therein. It appearscurrently as the unique imaging technique to investigate noninvasively both ex vivoand in vivo three-dimensional fiber architectures of the human heart. However, dueto the specific structural and functional properties of the cardiac muscle, there arestill several main problems in the research of cardiac dMRI. Firstly, it is difficult toexplain the reason for the diffusion anisotropy of molecules in the cardiac fiber, toknow how well the diffusion properties calculated from diffusion images reflect themicrostructure of the myocardium and to analyze the relationship between the dMRImeasures and the microstructures of cardiac fiber, since there is no ground-truthinformation available and add to that the influence of various factors such as spatialresolution, noise and artifacts, etc. Secondly, optimizing the dMRI scanningparameters is very significant for obtaining the high-quality diffusion weightedimages and analyzing accurately the complex cardiac fiber structure. However, up tonow, the optimization of the imaging parameters can only be achieved by using therepeated scans, which is lack of the theoretical basis, such experiment process iscomplicated and expensive. Finally, since dMRI is very sensitive to the motion,during the in vivo dMRI scanning, the motion caused by the heart beating, patientmoving and arrhythmias will generate the artifacts and introduce additional signalattenuation, it is therefore difficult to describe quantitatively the properties ofdiffusion images of in vivo cardiac fibers.For solving the above problems, based on the nature of dMRI theory and takinginto account the cardiac fiber structure properties, this thesis develops a realisticmodel-based dMRI simulator to simulate diffusion-weighted images for both ex vivoand dynamic cardiac fibers by integrating different imaging modalities. Thissimulator provides an effective mean for evaluating the measurement accuracy ofdMRI for cardiac fiber structure, analyzing the relationship between dMRI measuresand the microstructure of cardiac fiber, optimizing dMRI scanning parameters anddescribing quantitatively dynamic cardiac fiber structure. The following are themain issues addressed in this thesis.The first part concerns the issue that the clinical dMRI is unable to explain thecauses of the diffusion anisotropy of water molecules in the cardiac fiber structureand to evaluate the accuracy of dMRI detection for cardiac fiber orientations due tothe limit of the spatial resolution. For solving this problem, two local cardiac fiberstructure models are firstly constructed, the corresponding diffusion weighted and tensor images with different spatial resolutions are then simulated using aMonte-Carlo method, the main cause for diffusion anisotropy is finally given byanalyzing the influence of cardiac fiber structural and physical characteristics on thediffusion anisotropy. Meanwhile, the dMRI detection accuracy at different scales isanalyzed.The second part addresses the problem of analyzing the relationship betweendMRI measures and the microstructures of the cardiac fiber. The3D cardiac fiberstructure model of an entire heart is firstly constructed using the high-resolutionpolarized light imaging data, and then the corresponding diffusion weighted andtensor images properties at multi-scales are described through the simulation, finally,by controlling the cardiac fiber modeling parameters, the influence of the cardiacmyocyte structure variation on the diffusion image properties is investigated, andthe results show that the proposed dMRI simulator can provide an auxiliary meanfor exploring the relationship between dMRI properties and cardiac fibermicrostructures, and also for cardiac disease analysis and diagnosis.The third part deals with the issue of optimizing the dMRI scanning parametersfor cardiac fibers. To realize the optimization, a novel improved dMRI signalanalysis and simulation method is proposed, which takes into account the weightsall the gradients used in the imaging sequence on the diffusion of water moleculesfor simulating the process of dMRI more realistically. By investigating the influenceof each parameter on the diffusion image properties, the optimization principle ofsuch parameter is finally given.The last part puts the emphasis on the modeling of dynamic cardiac fiberstructures and the simulation of the corresponding diffusion images. It also analyzesthe variation of the diffusion image properties of cardiac fibers with the heartmotion. This work provides a basis for the in vivo myocardial fiber image analysis.