Noise-adaptive anisotropic diffusion filtering of MRI images reconstructed by SENSE (sensitivity encoding) method

SENSE (SENSitivity Encoding) imaging provides significant acquisition speedups in MRI. The main drawback of the method is that it generates images that have increased and spatially nonuniform noise levels and, hence, will often require retrospective filtering. In this paper, we show that standard anisotropic diffusion filtering, while being an effective technique for edge-preserving denoising of images with uniform noise levels, is often non-optimal for SENSE-reconstructed data. We have developed a modification of this filter for SENSE images using a robust statistical analysis of the anisotropic diffusion process. The new method utilizes the image noise matrix that is available from the SENSE reconstruction to automatically adjust filtering parameters with local noise levels. The effectiveness of the method and its advantage over standard anisotropic diffusion filtering for SENSE images were demonstrated with phantom and patient MRI data.