Fast Magnetic Resonance Diffusion Tensor Imaging Study Based On Compressed Sensing

As a special magnetic resonance imaging method, diffusion tensor imaging (DTI)not only can evaluate the completeness of a tissue structure in microscopic field, butalso is the only method that can show the trend of nerve fiber bundles. However, arelatively long acquisition time is needed in diffusion tensor imaging, so it is easilysuffered from the motion artifact, and maybe it would exceed patients’ endurance.Therefore, some extent its clinical application will be limited. As a result, how toimprove the imaging speed has been a hot topic in this research field. In recent yearscompressed sensing (CS) as an accelerated imaging method growing up, has beenapplied to magnetic resonance DTI successfully. This thesis has explored andresearched how to speed up diffusion tensor imaging preliminarily. The paper’s mainlywork can be shown as follows:The undersampling form has been optimized and measurement values have beencompensation by rotation interpolation, making full use of relevant information betweenimages. First of all, radial undersampling masks with angle rotation have beenconstructed based on the structure characteristiCS of k-space of diffusion tensor imagesto. So that the form of undersampling and measurements can be got. Second accordingto relevant information of different diffusion tensor images, to interpolate compensationthe sampling values of images by rotating undersampling masks. Finally using thereconstruction method of compressed sensing to rebuild diffusion tensor images.Results shown that rotating interpolation undersampling method reduces the size ofraw data, also it can preserve image edge information and image quality better.Fast diffusion tensor imaging is realized based on distributed compressed sensing(DCS) method. Because diffusion tensor images have high correlation coefficients andmeet with joint sparse model, we can use DCS theory to jointly reconstruct DTI Images.In this study, DCS has been used to obtain common components and innovationcomponents of DTI images, by adding them together we can get final DTI images.Results showed that the reconstruction error of DCS method is smaller than traditionalCS method and the calculation of diffusion coefficients are more accurate. In addition,results of DCS have better edge information compared with rotation interpolation CS,reconstruction results of DCS and rotation interpolation CS have equal contributionboth in improving the structural similarity and image peak signal noise rate.