|Title||Quasi-Monte Carlo estimation approach for denoising MRI data based on regional statistics|
|Publication Type||Journal Article|
|Year of Publication||2011|
|Authors||Wong, A., and A. Mishra|
|Journal||IEEE Transactions on Biomedical Engineering|
|Date Published||1076 - 1083|
|Keywords||Bayes methods, biomedical MRI, contrast measures, data adaptive method, general Bayesian least-squares estimation, Image Denoising, mean structural similarity, medical image processing, Monte Carlo methods, MR data post processing step, noise-free signal, nonlocal means MR noise reduction, QMCE, quasi Monte Carlo estimation approach, regional statistics, stochastic approach, stochastic processes, tissue structural boundaries, WAVE, wavelet transforms, wavelet-based MR noise reduction|
An important postprocessing step for MR data is noise reduction. Noise in MR data is difficult to suppress due to its signal-dependence. To address this issue, a novel stochastic approach to noise reduction for MR data is presented. The estimation of the noise-free signal is formulated as a general Bayesian least-squares estimation problem and solved using a quasi-Monte Carlo method that takes into account the statistical characteristics of the underlying noise and the regional statistics of the observed signal in a data-adaptive manner. A set of experiments were per formed to compare the proposed quasi-Monte Carlo estimation (QMCE) method to state-of-the-art wavelet-based MR noise reduction (WAVE) and nonlocal means MR noise reduction (NLM) methods using MR data volumes with synthetic noise, as well as real noise-contaminated MR data. Experimental results show that QMCE is capable of achieving state-of-the-art performance when compared to WAVE and NLM methods quantitatively in SNR, mean structural similarity (MSSIM), and contrast measures. Visual comparisons show that QMCE provides effective noise suppression, while better preserving tissue structural boundaries and restoring contrast.