High-speed compressed sensing reconstruction in parallel MRI using modified grappa.

Project report by an external student from Sree Chitra Thirunal College Of Engineering,Thiruvananthapuram

Master of Technology in Electronics & Communication Engineering (Signal Processing) 2014-2015 INT Joseph Suresh Paul

High image quality with short acquisition time is desirable for any fast MR imaging application. Parallel MRI is an important technology to achieve these goals. By de-ploying multiple receiver coils, parallel MR imaging can achieve several-fold reductions in scan-time. To reconstruct high quality full-FOV images from under-sampled data, a good reconstruction algorithm is needed.The GeneRalized Autocalibrated Partially Parallel Acquisition (GRAPPA) is a significant Parallel MR Imaging reconstruction technique in frequency domain for subsampled multiple-coil MR data. The additional k-space lines in low frequency region, named as auto calibration signal (ACS) lines, are used to find the relationship between sampled and nonsampled k-space data. Finally, the image is recovered by using sum of squares to combine all coil images. GRAPPA adopts a single finite-size kernel to approximate the true relationship between sampled and non-sampled signals. However, the estimation of this kernel, based on k-space signals is imperfect. Better reconstruction can be achieved by considering low frequency and high frequency regions separately. Here a method of reconstruction dealing with this frequency variation in k-space is implemented.