Parallel imaging technique using localized gradients (PatLoc) reconstruction using orthogonal mode decomposition

Parallel imaging technique using localized gradient (PatLoc) uses the combination of surface gradient coils and an RF receiver array to further improve the efficiency of gradients and to reduce the peripheral nerve stimulation hazard [1]. PatLoc is a generalization of non-Cartesian gradient encoding, such as MR-encephalography [2] or inverse imaging [3]. The imaging and reconstruction algorithm of PatLoc system has been previously reported using two orthogonal sets of linearly combined gradients with circular symmetry and a spin echo imaging sequence [4]. The reconstructed PatLoc image shows reduced sensitivity at the center of FOV as the result of lacking encoding information from both the gradient system and the RF sensitivity. We hypothesize that other linear combinations of gradients can improve the quality of image reconstruction. Specifically, rather than using two sets of polarity reverse gradients [4], we use singular value decomposition (SVD) to reveal two most significant linear combinations of the gradients as the “frequency” and “phase” encoding gradients. We demonstrate the advantage of this mode selection by using 8-channel gradient and RF coil array PatLoc system together with up to 4-fold acceleration using the iterative time-domain reconstruction (iTDR).