Deformable registration of DT-MRI data based on transformation invariant tensor characteristics


A. Guimond, C. R. G. Guttmann, S. K. Warfield, C.-F. Westin
ISBI, Washington (DC), USA
Proceedings of the IEEE International Symposium on Biomedical Imaging (ISBI'02)
Pages 1-4
July 7-10, 2002

Download full paper

Abstract

In this paper, we present an automated method for colon registration. The method uses dynamic programming to align data defined on colon center-line paths, as extracted from the prone and supine scans. This data may include information such as path length and curvature as well as descriptors of the shape and size of the colon near the path. We show how our colon registration technique can be used to produce synchronized fly-through or slice views.

Several regions have been highlighted in Fig. 2 in which important differences are observed between the various registration results presented. One can clearly see the advantage of using deformable registration to match tensors by comparing regions A and B (Fig. 2-a and -b respectively). Including the eigenvalues in the registration process further improves the registration result (Compare regions C and D in Fig. 2-b and -c). In particular, note that the tensors define much smoother tracts, as should be the case in the corpus callosum, and that the general structure of the source tensors (yellow) is improved as compared with the target tensors (green). A less clear-cut improvement is present when comparing the result of using the full tensor to perform registration as opposed to the eigenvalues. Region F (Fig. 2-d) seems to be an improvement between region E (Fig. 2-c), but tensors in region H seem to be less well aligned than in region G. This last difference might be due to a better match in other orientations, but this is only a presumption. 3D visualization of the tensors maps from both subjects might help better explain these differences and work is currently underway to develop such a visualization tool.


Reference

Guimond A, Guttmann CRG, Warfield SK, Westin CF. Deformable registration of DT-MRI data based on transformation invariant tensor characteristics. In Proceedings of the IEEE International Symposium on Biomedical Imaging (ISBI'02). Washington (DC), USA, 2002;1-4.

Bibtex entry

@InProceedings{guimondISBI02,
  author         = {A. Guimond and C. R. G. Guttmann and S. K. Warfield and    
                   C.-F. Westin},                                              
  title          = {Deformable registration of {DT-MRI} data based on          
                   transformation  invariant tensor characteristics},          
  booktitle      = {Proceedings of the IEEE International Symposium on         
                   Biomedical  Imaging  (ISBI'02)},                            
  year           = 2002,                                                       
  pages          = {1-4},                                                      
  address        = {Washington (DC), USA},                                     
  month          = {July 7--10}
}                                               

Grants

NIH P41-RR13218 (NAC)

Research areas

DTMRI, Tensor