中文摘要：

针对并行磁共振在欠采样率较高情况下重建图像存在的混迭伪影和噪声问题,提出一种非二次正则化的保边性图像重建算法.基于SENSE技术,该算法以保边平滑性的非二次凸函数为正则化项,构建一个非二次代价函数,并运用非线性共轭梯度算法求解该最小化问题,实现并行磁共振图像的保边性重建.为了评价算法的有效性和鲁棒性,以归一化均方误差作为评价准则,分析并行磁共振欠采样率最大时真实数据和仿真数据的图像重建.结果表明,该算法显著减少欠采样率较高时并行磁共振图像的混迭伪影,并能够有效抑制噪声和保留边缘信息.相比于其他图像重建算法,该算法能够快速收敛.

英文摘要：

Aiming at the images of poor quality resulted from the aliasing artifacts and noise in parallel magnetic resonance imaging,which was reconstructed from high reduction undersampling sensitivity-encoding data,a non-quadratic regularized edge-preserving reconstruction algorithm was proposed.Based on Sensitivity Encoding technique,the algorithm used an edge-preserving non-quadratic convex function as the regularization term,and then a non-quadratic cost function was constructed.Using nonlinear conjugate gradient method,reconstruction image was obtained by minimizing the objective function.In order to evaluate the robust and validity of the proposed algorithm,analysis on severe undersampling data was presented and discussed.Based on the analysis indicator known as normalized mean squared error,the results show that for high acceleration factors,the proposed algorithm evidently reduces the aliasing artifacts in the reconstruction images,and noise is effectively restrained as well as edge information is preserved.Furthermore,the proposed algorithm can be quick convergence.