Premium
SU‐GG‐I‐138: Registration and Fusion of MRI and Interventional X‐Ray Images Based On External Markers in a XMR Suite
Author(s) -
Simard D,
Beaudoin G,
Deschenes S
Publication year - 2008
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.2961536
Subject(s) - imaging phantom , orientation (vector space) , artificial intelligence , computer vision , image registration , magnetic resonance imaging , computer science , nuclear medicine , medical imaging , medicine , radiology , mathematics , image (mathematics) , geometry
Purpose: To report a method for registration and fusion of 3D magnetic resonance (MR) volume and 2D X‐ray images obtained from a C‐arm mounted flat panel detector. This XMR (X‐ray/MR) technique has been designed as a support for X‐ray guided embolization of venous malformations using pre‐interventional magnetic resonance imaging (MRI). Method and Materials: Registration and fusion are performed through a custom‐made graphical user interface and are based on multimodality external markers. A high resolution MR sequence has been optimized to obtain a precise pre‐interventional localization of the markers. After the localization, a rigid body registration method is implemented, based on one or two X‐ray acquisitions, which can be apply to subsequent images with different C‐arm angulations. The precision of the registration, with respect to the number of markers used, has been evaluated on a phantom and tested off‐line with patients. Results: The computation of registration parameters from one MR volume, two X‐ray images and four external markers on a phantom, is a good example of realistic clinical conditions. For this computation, an excellent registration accuracy is obtained for any subsequent image taken at different C‐arm orientations (0.9±0.6mm). Furthermore, we find that only one X‐ray view is sufficient to reach equivalent accuracy when the registration is performed on images taken at similar C‐arm orientation (0.8±0.6mm). Conclusion: As far as we know, this level of accuracy has never been reported in the literature using a similar registration method. Starting with only two X‐ray acquisitions, this work shows that excellent registration accuracy with an MRI acquisition is possible for any subsequent image taken at different C‐arm angulations. Moreover, we show that when the registration is performed on images taken at about the same C‐arm orientation, one view is definitely enough for an accuracy of approximately 1 mm.