Meeting Banner
Abstract #2573

Comparing Anisotropic Diffusion Filters: Enhancement of 23Na MRI of Human Kidney

Frank G. Zoellner1, Christoph M. Decker1, Simon Konstandin1, Lothar R. Schad1

1Computer Assisted Clinical Medicine, Heidelberg University, Mannheim, Germany

In-vivo MR imaging of sodium is limited due to its electro-physiological characteristics. A logical step forward is to enhance image intensity by applying post-processing filters while Gibbs ringing could also be suppressed. The increase in signal-to-noise ratio (SNR) by linear filters as previously proposed is paid by a loss in resolution (blurring). Filters that preserve contours while removing noise are anisotropic diffusion filters (AND). The aim of this study was to compare different anisotropic filters for enhancing sodium MR images of the kidney.

Keywords

acquisition adapted addition amidst analyses angular anisotropic applied assisted automatically available background behind best bisector blurring body boundaries brain canny characteristic choice chosen clinical close coefficient coil combination commercially comprising computed computer computing coordinates correlation decker decrease density depends depicted determined diffusion edges electro elements enhance enhanced enhancing envision estimator examine fails filter filtered filtering filters forward frank frequencies function gradient gradients histogram human idea ideally identical in vivo initial intensities intensity isotropic iterations iterative kidney length linear logical losing loss many measured middle model much noise nominal observe omitted optimal original patterns physiological plotted post preserve preserving previously process processing projections properties proposed quantile quantitative radial readout recently reception reduced removing reported residual resolution revealed ringing sampling satisfactory scanner selected sensitivity sets several simplest slices slightly smoothed smoothing sodium space spatial spoke squares step strong structures successfully sums suppressed tensor threshold trajectories trans tuned typical unfiltered upper variants versions volunteer whole worse