A post-processing method for multiexponential spin–spin relaxation analysis of MRI signals

Abstract

Quantitative MR imaging is a potential tool for tissue characterization; in particular, proton density and proton relaxation times can be derived from MR signal analysis. However, MR image noise affects the accuracy of measurements and the number of tissue parameters that can be reliably estimated. Filtering can be used to limit image noise; however this reduces spatial resolution. In this work we studied, using both simulations and experiments, a filter called a 'selective blurring filter'. Compared to other classical filters, this filter achieves the best compromise between spatial resolution and noise reduction. The filter was specifically used to reliably determine the bi-component transverse relaxation of protons in adipose tissue. Long and short relaxation times and the relative proton fraction of each component were obtained with a degree of uncertainty of less than 10% and an accuracy of 95%.