Accelerated 2D Laplace NMR of porous media with compressed sensing at low SNR

Abstract

Abstract Multidimensional Laplace NMR methods can provide insights into pore structures and confined fluid information. However, these methods are time-consuming and could be accelerated. In this research, we shorten the measurement time for T2–T2, D–T2, and D–T2 MRI experiments with compressed sensing (CS) for low signal-to-noise ratio (SNR) NMR data. Pseudo-random sampling is employed as a subsampling scheme. The spectra are reconstructed by l1 minimization with wavelet transform regularization. The influence of the reconstruction is then examined (SNR: 10–60). Also, spectra from fully sampled and subsampled data were compared to explore the accuracy of the CS reconstructions. Both simulations and experiments were performed to verify this method. Sandstone samples were measured, in one case saturated with water, and the other with water and oil. The experiments were performed on a 2 MHz rock analyzer (Oxford Instruments, UK). The results show that the accuracy of reconstruction increases with the SNR, and an accelerated ratio (AR) of 5 can be achieved for SNR 20 with a normalized mean square error of around 6%. The oil and water can also be clearly distinguished from the subsampled D–T2 and D–T2 MRI correlation maps. This method is expected to be useful in low-field (such as NMR well logging) and spatially resolved NMR, where the SNR is low.