EE599 Magnetic Resonance Imaging and Reconstruction Final Project Report Jungwoo Lee (610-21-4670) Reference : An Improved Gridding Method for Spiral MRI Using Nonuniform Fast Fourier Transform

Final Project Report Jungwoo Lee (610-21-4670) Reference : An Improved Gridding Method for Spiral MRI Using Nonuniform Fast Fourier Transform Authors: Liewei Sha, Hua Guo, and Allen W.Song Introduction The objective of this project is to develop improved reconstruction algorithm using nonuniform fast Fourier transform(NUFFT) for spiral MRI. In conventional gridding methods, while explicit forms of kernel functions were used, in this paper optimized kernels are suggested minimizing reconstruction error in the least square sense between real and reconstructed images. To accomplish its goal well-known techniques such as Kaiser-Bessel gridding(KB)[3] and a generalized fast Fourier transform(GFFT)[4] are being used to put an emphasis on the new NUFFT method called LS_NUFFT[1,2] compared to conventional ones. With phantom image(128x128) provided in the homework set, the concept of improved NUFFT method and the corresponding results are presented and the difference between LS_NUFFT and previous algorithms including KB and GFFT is discussed. In order to be able to succeed in getting good reconstruction, results with different values of parameters are shown for justification. Finally, some possible modification of LS_NUFFT method is addressed in final remarks.