Compressed sensing MRI using sparsity induced from adjacent slice similarity

We propose a fast magnetic resonance imaging (MRI) technique based on compressed sensing. The main idea is to use a combination of full and compressed sensing. Full sensing is conducted for every several slices (F-slice) while compressed sensing with high compression rate is applied to the rest of slices (C-slice). We can perfectly reconstruct F-slice images, which are used to roughly estimate the C-slices. Since the estimate is already of good quality, its difference from the original image is small and sparse. Therefore, the difference can be reconstructed precisely using the standard compressed sensing technique even with high compression rate. Simulation results show that the proposed method outperforms conventional methods with 3.16dB for arm images, 0.26dB for brain images in average for the C-slices with perfect reconstruction for the F-slices.