Results of automatic image registration are dependent on initial manual registration

Abstract

Measurement of static alignment of articulating joints is of clinical benefit and can be determined using image-based registration. We propose a method that could potentially improve the outcome of image-based registration by using initial manual registration. Magnetic resonance images of two wrist specimens were acquired in the relaxed position and during simulated grasp. Transformations were determined from voxel-based image registration between the two volumes. The volumes were manually aligned to match as closely as possible before auto-registration, from which standard transformations were obtained. Then, translation/rotation perturbations were applied to the manual registration to obtain altered initial positions, from which altered auto-registration transformations were obtained. Models of the radiolunate joint were also constructed from the images to simulate joint contact mechanics. We compared the sensitivity of transformations (translations and rotations) and contact mechanics to altering the initial registration condition from the defined standard. We observed that with increasing perturbation, transformation errors appeared to increase and values for contact force and contact area appeared to decrease. Based on these preliminary findings, it appears that the final registration outcome is sensitive to the initial registration.

Keywords::

• kinematics
• image registration accuracy
• voxel-based image registration
• manual image registration
• magnetic resonance imaging

Acknowledgements

We would like to acknowledge Columbia University for use of Joint_Model software, and Allan Schmitt for technical assistance with MR imaging.

Conflict of interest disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

^1. Email: [email protected]

Additional information

Funding

This research was supported by the National Institute of Biomedical Imaging and Bioengineering of the National Institutes of Health [grant number R01EB008709].