A Study On Rapid Diffusion Tensor Imaging Algorithm Based On Compressed Sensing

Diffusion tensor imaging (DTI) is a kind of magnetic resonance imagingtechnology. DTI is obtained by a series of different dispersion direction ofdiffusion-weighted imaging (DWI), the main advantage is that it can express the finestructure of the organization and the associated. DTI can be measured myocardial andcerebral white matter fiber direction, while this can not be achieved in other MRItechnology.The purpose of my research is reduce the sampling data points and to improvethe quality of image reconstruction. Based on existing algorithms of compressedsensing and further bring compressed sensing theory to diffusion tensor magneticresonance imaging. In this paper, studies how different sampling strategy affectmagnetic resonance imaging based on compressed sensing and diffusion tensor. Andto some extent improve the traditional compressed sensing algorithm.This paper research a method using the joint sparsity constraint of DTI data tofast DTI acquisition which called DCS(distributed compressed sensing).DCS, as an extension of CS by incorporating the joint sparsity prior of multiplesparse signals, has been exploited in numerous studies. It is shown through theoreticalanalysis and numerical study that this new technique can effectively reduce thenumber of measurements to achieve a given reconstruction quality or improve thereconstruction quality for a given number of measurements, compared with the basicdisjoint CS method of reconstructing signals individually without using their jointsparsity propertyDiffusion tensor imaging (DTI) is known to suffer from long acquisition time inthe orders of several minutes or even hours. Therefore, a feasible way to accelerateDTI data acquisition is highly desirable. In this paper, the feasibility and efficacy ofdistributed compressed sensing(DCS) to fast DTI is investigated by exploiting thejoint sparsity prior in diffusion weighted images(DWIs).Fully-sampled DTI datasets were obtained from both phantom and experimentalheart sample, with diffusion gradient applied in six directions. The k-space data wereunder-samples retrospectively with acceleration factors from2to6. DWIs werereconstructed by solving an l2-l1norm minimization problem. Reconstructionperformance with varied signal-to-noise ratio(SNR) and acceleration factors wereevaluated by root-mean-square error and maps of reconstructed DTI indices.Superiority of DCS over basic CS was confirmed with simulation, and the reconstruction accuracy was influenced by SNR and acceleration factors.Experimental results demonstrate that DTI indices including fractional anisotropy,mean diffusivities, and orientation of primary eigenvector can be obtained with highaccuracy at acceleration factors up to4.Sometimes we use the finite difference (Finite-Differences) to be the sparsityfunction, which is called TV (total variation), TV is the sum of the absolute differencein the whole image. When use the other sparsity function as the objective function ofthe magnetic resonance image, we usually use the TV as a penalty function, the paperalso studied the impact of the wavelet sparsity parameters and TV sparsity parameterswhen reconstruct the images.Due to the characteristics of DTI data, b0image and other diffusion direction ofthe image value are quite different, this case may lead great impact to the accuracywhen reconstruct the DTI using DCS. So b0image sampling strategy is also studiedto research its influence on DTI reconstruction quality by using three samplingstrategy and comparing the results. The results show that different sampling strategieshave a great influence on the reconstruction quality.DCS is shown to be able to accelerate DTI and may be used to reduce DTIacquisition time practically.