The Implementation Of Image Reconstruction Algorithm On CS—MRI Based On VC

Magnetic resonance imaging (MRI) has many advantages, such as no invasion,no ionizing radiation, high resolution ratio, and high flexibility. It can free to choosethe location of the body through multi-parameters. But it hinder) the clinicalapplications because it takes a long imaging time. Then people continue to explore avariety of fast MRI imaging technologies, such as spiral acquisition, radial acquisition,parallel acquisition, etc. Although these collection methods greatly accelerate the datasampling speed, the spiral acquisition and radial acquisition have high requirementson magnetic resonance equipment. A little equipment meets the hardwarerequirements. When the parallel imaging acceleration factor is too large, it will lead todecrease the SNR and quality of the reconstructed image. Moreover, when the datacollection meets the Nyquist sampling theorem, the reconstruction process alsorequires a lot of storage space.Compressed sensing (CS) theory surpasses the limitation of traditional Nyquistsampling theorem. It can only collect few samples to reconstruct the original imagewith high quality, and greatly improve the imaging speed. The techniques based oncompressed sensing magnetic resonance imaging(CS_MRI) has the followingadvantages: First, It puts the signal acquisition and compression together, so that itcan greatly reduce the amount of sampling data and save storage space; Second, theuse of appropriate sparse image reconstruction algorithm can recover enoughinformation of the reconstructed image with high quality. The technology includestwo methods: One is the design of MRI pulse sequence with under-sampling, theother is the research of the reconstruction algorithm with under-sampled data. Wemainly research the CS-MRI reconstruction algorithm, using VC software to achieveour idea and lay the foundation of the technology in the MR scanner.We first introduce the principle of MRI，the sparse transform, and the dataacquisition based on CS-MRI. Most MR images in a transform domain can be sparserepresented. For example, the human vascular images are sparse, so we just use asimple differential transform that change them into sparse; And for some of the othercomplex medical images, such as the brain images, we need to adopt a method whichcombined wavelet transform and the finite difference to change them into sparse. Andin this way, those medical images can be sufficient sparse and then get a betterreconstructed image with the specific reconstruction algorithm.Then we research one of the compressed sensing reconstruction algorithmsapplied in MRI. And we realize the nonlinear conjugate gradient descent algorithm to get the reconstructed image. The reconstructed image is compared with the originalimage with the error analyze. The result shows that, when we apply the compressedsensing algorithm to MRI reconstruction, we can successfully and quickly reconstructthe image with high quality.Finally, we implement the CS algorithm in MRI using C++, which is proposed forengineering application. The software system is developed with VC++6.0based onMFC dialog. The software includes three parts: the phantoms import system, the dataacquisition system, and the image reconstruction system. The phantoms importsystem imports the original images collected by MR scanner. First, it can provide theoriginal image data, and second, it is convenient to compare with the reconstructedimage. The data acquisition system uses the full acquisition technic and theunder-sampled acquisition technic. We research those two technics, and know thatwhich technics can obtain the better reconstruction with CS-MRI. The imagereconstruction system mainly implements the nonlinear conjugate gradient descentalgorithm to reconstruct the image with the data.The platform provides a tool to realize image reconstruction with CS_MRItechnology. It can further enrich the subsequent reconstruction algorithm and improvethe speed of the reconstruction, also can build the foundation of the technology onMR scanner.