Par0001 - elastix

Registration Description

interpatient; affine + B-spline transformation; mutual information

Image data

Screen shot:

alt-text

Application

A set of 50 clinical MR scans were registered interpatiently for the final purpose of atlas-based segmentation of the prostate. In this experiment the effect of the transformation model was studied.

To this end, we used localised MI as a cost function [1], and a four-level Gaussian image pyramid with downsampling. The following transformation models were compared: translation, rigid, affine, and B-spline with different control point spacings: 64, 32, 16, 8, and 4 mm. The result of the registration with translations was only used as an initialisation for all other registrations. For the B-spline registrations, the control point grid was subjected to a multiresolution scheme: registration starts with a coarse control point resolution; with smoother versions of the images, the control point resolution is increased accordingly.

Registration Settings

For parameter files see the Elastix Model Zoo repository on GitHub.

elastix version: 3.9

Command line call:

elastix -f patientA.mhd -m patientB.mhd -p par0001translation.txt -p par0001.txt -out outputdir

with: = one of {rigid, affine, bspline64, bspline32, bspline16, bspline08, bspline04}

Published in

These registration are described in the publication:

S. Klein, M. Staring, K. Murphy, M.A. Viergever, J.P.W. Pluim, "elastix: a toolbox for intensity based medical image registration," IEEE Transactions on Medical Imaging, vol. 29, no. 1, pp. 196-205, 2010.

Other comments

Image data and manual segmentations have been made available via https://bigr-xnat.erasmusmc.nl. See http://xnat.bigr.nl for instructions how to log in.

References

[1] S. Klein, U.A. van der Heide, I.M. Lips, M. van Vulpen, M. Staring and J.P.W. Pluim, "Automatic Segmentation of the Prostate in 3D MR Images by Atlas Matching using Localised Mutual Information," Medical Physics, vol. 35, no. 4, pp. 1407 - 1417, April 2008.

© 2020 Viktor van der Valk