High Pass SENSE

Introduction Artificially increasing the sparsity of an image before reconstruction has been used to improve the performance of partially parallel imaging techniques [1, 2]. Similar to the requirement of explicit sensitivity maps, an explicit image support (non-zero regions) definition is also required to take fully advantage of image sparsity in conventional SENSE [3]. However, it is not trivial to provide an accurate image support definition. To overcome this challenge, this work uses regularized SENSE as proposed in Ref [4], which adopts a regularization image. A sparse image, which was generated by high pass filtering and g-factor map suppression, is used as the regularization image. The explicit definition of image support is avoided. Experiments show that the proposed method can reconstruct images with noise levels significantly lower than conventional SENSE. Excellent image quality has been achieved using an 8-channel head coil and a 1D net acceleration factor of R = 4.