Coil Correction

Before After

Before (left) and after (right) coil correction
Overview
We will demonstrate a novel, fully automated, method of image intensity correction that does not rely on tissue identification or RF coil modeling. Methods dependent on  true tissue intensity levels are subject to error with pathological data whilst intensity correction algorithms based on phantom RF coil models ignore the effects of patient  anatomy. The new algorithm uses two orthogonal, scaled, error weighted, contiguous pixel gradient images which are smeared using a linear sequential filter.  The differential maps are re-integrated in two paths constrained by a chi squared fit into a gain correction mask for intra-slice normalisation. Interslice intensity variation of 30% in phantoms can be adjusted to within the statistical limitations of noise. Two iterations of the fit bring post contrast T1  weighted images to within 1% of a stable correction map. White matter intensity consistency is improved by 50% from an image with an original gain drift of 30%.  For inter-slice normalisation a stable correction to within 0.1% can be achieved in both clinical brain images and phantoms, with intensity non-uniformities of up to 40%. This algorithm offers a simple machine independent technique for image normalisation prior to analysis.

Downloads
source
Library toolkit
binary 1
Sun Sparc - Solaris 2
binary 2
Intel x86 - Linux 2
Publications
[1] E.Vokurka, N.Thacker, A.Jackson, `A Fast Model Independant Method for Automatic Correction of Correction of Intensity Non-Uniformity in MRI Data' JMRI, 10, 4, 550-562, 1999.
[2] N.A.Watson, Y.Watson, E.Vokurka, A.Jackson, N.A.Thacker., `High Resolution MR imaging og the Orbit Using Intensity non-Uniformity Correction. proc. ECR, p 197, SS908, 658, Vienna, March, 2000
the end
[Back to software]