Jan Petr, Jan Kybic, Sven Müller: Fast Parallel MRI Reconstruction Using B-spline Approximation

Parallel MRI (pMRI) is a way to use data obtained simultaneously from several receiver coils to increase the speed of the MRI acquisition. The receiver coils have distinct spatial sensitivities over the imaged area. Therefore, the measured data contains more information about the position of the signal than data obtained by a single coil. The idea is to speed up the acquisition by sampling the imaged object more sparsely and compensate the data loss with the use of the additional information obtained from more receiver coils. Methods that work in image domain estimate the reconstruction transformation independently in each point. This makes the reconstruction time to increase significantly with high image size. We propose an algorithm that uses B-spline functions to approximate the reconstruction coefficients. This reduces the number of parameters to estimate and thus makes the reconstruction faster and more robust to noise. We have tested our method on both phantom and in vivo data. We compared our method with Siemens implementation of GRAPPA and mSENSE in terms of time complexity and quality of the reconstructed images.