Meeting Banner
Abstract #3026

Confounds in Image Registration

Sune Darkner1, Matthew George Liptrot1, Jon Sporring1

1Department of Computer Science, University of Copenhagen, Copenhagen, Denmark

Many of the analysis methods applied to neuroimaging are dependent upon correct co-registration of datasets. The direction in which the registration is applied (image A registered to image B, or vice-versa) may vary depending upon the topic of interest without consideration of possible confounds this may involve. Here we show how even affine registration is subject to inverse inconsistency, where the co-registration result is different depending upon the direction chosen. A difference of 0.5 voxels is demonstrated for a pure translational, affine registration. Care must therefore be taken when performing co-registration in order to minimize subsequent directionality-based confounds.

Abstract PDF Presentation Video

Keywords

accepted addition affine analytical applications applied approaches approximate assessed asymmetry atlas brain care cause choice chosen clear computer confounds constrained correct curvature datasets density dependent depending determined differed discrete distributions either employ employed equivalent error especially estimation even examinations example exhibit exhibited extraction fast features framework general generalized gradient histogram histograms impact inconsistency inconsistent increasing intelligence interpolation inverse inversely joint locally location machine many maximization medical minimize modal modalities much must mutual nearest neighbor normalized novel often optimal optimum orderless otherwise partial pattern popular population probability produce prohibit pronounced pure purely recent referral registered registration registrations remove rigid risk scheme schemes science simulated structural subject subjects subsequent surface suspected symmetric symmetrical taken termed theoretical theory topic trans transactions translation turn upon utilizing vary versa vice viola volume whilst window