A new framework for real-time MR imaging by using time and gradient sparsities
Xiaoying Cai,Feiyu Chen,Enhao Gong,Shi Wang,Kui Ying
2012-01-01
Abstract:l A new framework for real-time MR imaging by using time and gradient sparsities Xiaoying Cai, Feiyu Chen, Enhao Gong, Shi Wang, and Kui Ying Biomedical Engineering, Center for Biomedical Imaging Research, Tsinghua University, Beijing, Beijing, China, Department of Biomedical Engineering, Tsinghua University, Beijing, Beijing, China, Electrical Engineering, Stanford University, Stanford, CA, United States, Department of Engineering Physics, Tsinghua University, Beijing, Beijing, China Target audience Researchers on real time MR images with high spatial/ temporal resolution. Purpose Real-time MR imaging has drawn great attention in clinical practice including medical diagnosis and therapy treatment monitoring. But there is always a tradeoff between temporal and spatial resolution: a highly under-sampled k-space data set is good for improving temporal resolution but at the cost of reducing spatial resolution and thus introducing artifacts which can be challenging for reconstruction. Since medical dynamic or 3D images are usually sparse in z-direction or tdirection after certain transformation, several methods have been proposed to reconstruct MR image using this nature of sparsity, like k-t FOCUSS[1] or 3D CS[2] However, k-t FOCUSS uses only temporal sparsity of images and is specially effective for dynamic process with periodic motion, while 3D CS utilizes the sparsity of the gradient of images. In this study, we propose a framework for high spatial/ temporal dynamic imaging reconstruction combining k-t FOCUSS and 3D CS to improve quality of reconstructed images from highly under-sampled k-space by exploring sparsity of images in k-t domain and gradient in image domain. The results demonstrated that our method performs better than the existing methods in eliminating artifacts and keeping structure details. Theory (1)K-t FOCUSS[1] explores sparsity of k-t domain by solving: