Back to Search Start Over

Comparison of two landmark based image registration methods for use with a body atlas

Authors :
Crafoord, Joakim
Mahmoud, Faaiza
Kramer, Elissa L.
Maguire Jr., Gerald Q.
Noz, Marilyn E.
Zeleznik, Michael P.
Crafoord, Joakim
Mahmoud, Faaiza
Kramer, Elissa L.
Maguire Jr., Gerald Q.
Noz, Marilyn E.
Zeleznik, Michael P.
Publication Year :
2000

Abstract

We describe preliminary work registering abdominal MRI images from three healthy male volunteers. Anatomically selected 3D homologous point pairs (landmarks), from which eigenvalues were generated to form the basis for a 3D non-affine polynomial transformation, were placed on axial slices alone and on axial, coronal and sagittal slices. Registration accuracy was judged visually by comparing superimposed 3D isosurfaces from the reference, untransformed, and transformed volume data and by comparing merged 2D slices projected fi om the transformed and reference volume data superimposed with 2D isolines. The squared sum of intensity differences between the transformed/untransformed and the reference volume was significant at the 0.05 (p >0.05) confidence level. The correlation coefficient improved by an average of 38% and the cross correlation between pixel values improved by an average of 22%. In each trial, the standard deviation of the landmarks after transformation was within one voxel and the standard error of the mean was not significantly different from zero at the 0.05 confidence level. Abdominal isosurface volume differences (between individuals) changed from an average of 14.5% before registration to 2.9% after registration. This experiment shows that it is possible to choose landmarks such that abdominal data from different subject volumes can be mapped to a common reference, and thus that it is possible to use this combined volume both to form an atlas and to warp abdominal data from an atlas to a patient volume.<br />QC 20100525 NR 20140804

Details

Database :
OAIster
Notes :
English
Publication Type :
Electronic Resource
Accession number :
edsoai.on1234874215
Document Type :
Electronic Resource