Abstract
Compressed sensing is a technique that acquires less MRI signals than conventional reconstruction, and thus reduces scan times. Quality of MR image in compressed sensing depends strongly on the method used for the undersampling of phase-encoding. However, optimization of the image quality using an actual machine involves time and effort. The purpose of this work is to establish a relationship between the method of undersampling and the image quality in the Cartesian setup using a MRI simulator. This simulator enabled effective investigation on influence of three factors on the quality of image: ratio of the constantly selected central portion of phase-encoding to full phase-encoding in k-space, shape of a probability density function for randomly selecting phase-encoding, and randomness of the phase-encoding. The quality was estimated using RMSE, inclination, SSIM, and SNR. The results show that the image quality is decided by both the ratio of the central portion and the full width at half maximum (FWHM) of normal distribution for randomly selected phase-encodings and that randomness of phase-encoding exhibits less influence on the image quality and has a small positive correlation with the image quality.
