Research On Reconstruction Algorithms For Quantitative Imaging Of Diffusion And Magnetic Susceptibility

Quantitative magnetic resonance imaging provides biomarkers for studying human central nervous system.Quantitative diffusion weighted imaging(qDWI)can measure the changes in the diffusivity of water molecules in tissue.The quality of-the reconstructed diffusion weighted image directly determines the accuracy of diffusion parameter map.Quantitative susceptibility mapping(QSM)can measure the concentration of magnetic sensitive material such as iron in tissue.The solution of QSM field-source inverse problem directly determines the accuracy of the magnetic susceptibility map.In this thesis,the following three high-quality reconstruction algorithms are developed,respectively for diffusion weighted image reconstruction and magnetic susceptibility map reconstruction:(1)An eigen-analysis based algorithm for reconstructing multi-shot diffusion weighted image,eIRIS for short.Multi-shot diffusion imaging technique suffers from the phase inconsistency between multiple shots.The typical IRIS algorithm uses navigator-echo to correct the phase inconsistency between multiple shots.However,IRSI suffers from geometry mismatch between navigator-echo and image-echo,especially in the case of spinal cord diffusion imaging which requires large shot numbers.In order to redcue the sensitivity to the mismatch between navigator-echo and image-echo for IRIS,in this thesis,eIRIS is developed.Firstly,eIRIS proposed to treat the phase inconsistency between multiple shots as the phase modulation of the coil sensitivity maps(CSM).Thus,shot dimension can be treated as virtual coil dimension.The CSM of virtual coil is called phase modulated CSM.Secondly,eIRIS proposed to use an eigen-analysis scheme to extract the phase modulated CSM from navigator-echo.Since the eigen-analysis scheme is insensitive to data errors such as image distortion.eIRIS can reduce the sensitivity to mismatch between navigator-echo and image-echo.Experimental results showed that eIRIS reduced artifacts caused by mismatch between navigator-echo and image-echo in the reconstructed diffusion weighted image.(2)A joint reconstruction algorithm based on diffusion weighed images from different directions,joint reconstruction for short.The scan time for qDWI is long,which increases the possibility of motion during qDWI acquisition and thus reduces the image quality.To accelerate qDWI acquisition,inthis thesis,a joint reconstruction algorithm is developed.The joint reconstruction algorithm proposed to use the correlation between diffusion weighted images from different directions and use the Tikhonov regularization term to constrain the reconstruction of anisotropic image.Experimental results showed that the propsoed joint reconstruction algorithm improved the reconstructed image quality at high acceleration factor and less signal averages.(3)A morphology-adaptive total variation for the reconstruction of quantitative susceptibility map,MATV for short.The field-source inverse problem in QSM is ill-condit:ioned.The typical MEDI algorithm improves the quality of the reconstructed magnetic susceptibility map by imposing total variation(TV)constraint to the voxels inside smooth regions.However,MEDI does not impose any constraints in regions near edges and thus results in inaccurate reconstruction near edges.In order to improve the image quality near boundaries,in this thesis,MATV is developed.Firstly,MATV proposed to use a global TV constraint,that is,MATV imposed TV constraint to both voxels inside smooth regions and voxels near edges.Secondly,MATV proposed to use a morphology-adaptive TV constraint to avoid over-smoothing of the tissue edges.The TV weight in MATV is a monotonically decreasing function of magnitude gradients.Thus,voxels inside smooth regions are assigned with larger TV weights than those near edges.Experimental results showed that MATV can reduce artifacts near boundaries.