Comparison of Two Under-sampling Algorithms, Non-uniform Random Sampling and Variable Density Sampling in CS-MRI

This paper is to compare the two undersampling modes NURS and VDRS in Compressed Sensing MRI. In this paper, brain images are reconstructed using Matlab using NURS and VDRS respectively. After analyzing the sampling pattern and probability distribution function of the two undersampling modes and outputting different sampled images to select the appropriate soft threshold, the soft thresholded sampled images are reconstructed by the PCOS algorithm. After evaluating the quality of image reconstruction by utilizing several indexes such as eye observation discrimination, PSNR and SSIM, it is found that VDRS mode is better than NURS mode in this brain image reconstruction application. The experiments show that VDRS mode has less artifacts and higher image quality in brain image reconstruction, while NURS mode has a lot of artifacts, even though most of the noise is removed after noise reduction, but it also reduces the original appearance of the image and the image quality will be degraded as well. This study can provide an idea for future scholars on how to choose the undersampling pattern in CS-MRI applications, and also promote the research on different undersampling algorithms.