Segmented Region Based Reconstruction of Magnetic Resonance Image

Compressed Sensing theory promises to reconstruct the magnetic resonance images from partially sampled k-space data. Through this Compressed sensing - magnetic resonance imaging CS-MRI technique, we accelerate the reconstruction process but at the cost of high artifacts especially with the increase of high reduction factor and high reconstruction time. To minimize these artifacts, we proposed a segmented region based reconstruction technique to enhance the quality image without affecting much more the reconstruction time. In this algorithm, the partial k-space data segmented into two parts according to their frequencies. At central part which has lower frequency components selected and predicted by nuclear norm minimization. After that the part is fused with peripheral part of the k-space components and apply this recovery technique another time to reconstruct more accurate images in terms of conventional techniques. To analyze the performance of proposed algorithm, we compare the results for different data sets of brain with CS techniques. Better results in term of NMSE and time shows the effectiveness of proposed method with high reduction factor of data.