Meeting Banner
Abstract #4651

Noise Bias Correction for Signal Averaged Images

Elena Olariu1, Arturo Cardenas-Blanco2, Ian Cameron1,3

1Physics, Carleton University, Ottawa, Ontario, Canada; 2Ottawa Health Research Institute; 3Diagnostic Imaging, Ottawa General Hospital

Clinical MR images are corrupted by noise which may reduce the reliability of quantitative analyses. The extraction of the true MR signal intensity from noisy MR magnitude images is confounded by a bias, which will be referred to here as Rician Bias (RB), caused by noise rectification in the magnitude calculation for low intensity pixels. Averaging in the image domain reduces the effective noise but not the noise bias. For low SNR a post-processing scheme to correct the noise bias combined with a limited amount of signal averaging is preferable. The RB correction method discussed here, which is an implementation of the theory developed by Koay and Basser, has been previously described2. The results are extended here to consider the effect of signal averaging and inaccuracies in the value of g used.