MR Imaging Based On Iterative Shrinking Threshold Algorithm And Adaptive Sampling

Magnetic resonance imaging(MRI),as an important detection method in clinical medicine and medical research,is widely used at present.However,MRI has the problems of slow imaging speed and long scanning time,which limits the further promotion of MRI.The Compressed Sensing(CS)theory provides a theoretical basis for fast magnetic resonance imaging(CS-MRI).CS-MRI can be divided into three parts: sparse representation,compressed sampling and reconstruction.At present,most CSMRI methods adopt fixed sparse representation and sampling methods,and different sparse conversion and sampling methods are selected according to different types of images.This phenomenon increases the difficulty of compressed sensing MRI and may increase the unnecessary loss of reconstruction accuracy.Based on the above issues,the research content of this article can be divided into the following two aspects:First,this paper proposes a CS-MRI network framework based on iterative contraction threshold algorithm and adaptive sampling,which combines the advantages of compressed sensing algorithm and deep learning network and implements sparse representation and sampling mode adaptation.Second,regarding the CSMRI reconstruction network based on iterative shrinkage threshold algorithm and adaptive sampling,this paper proposes LS-ISTA-Net,{0,1}-ISTA-Net,{-1,+1}-ISTA-Net,DSLS-ISTA-Net,these four reconstruction methods based on different sampling sub-networks,while achieving adaptive sampling,reduce the computational complexity of image sampling and the storage requirements for the sampling matrix and accurately and quickly reweight magnetic resonance images Structure.