Symmetric diffeomorphic modeling of longitudinal structural MRI - PubMed Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2013 Feb 5:6:197.
doi: 10.3389/fnins.2012.00197. eCollection 2012.

Symmetric diffeomorphic modeling of longitudinal structural MRI

John Ashburner  1 , Gerard R Ridgway
Affiliations

Symmetric diffeomorphic modeling of longitudinal structural MRI

John Ashburner et al. Front Neurosci. .

Abstract

This technology report describes the longitudinal registration approach that we intend to incorporate into SPM12. It essentially describes a group-wise intra-subject modeling framework, which combines diffeomorphic and rigid-body registration, incorporating a correction for the intensity inhomogeneity artifact usually seen in MRI data. Emphasis is placed on achieving internal consistency and accounting for many of the mathematical subtleties that most implementations overlook. The implementation was evaluated using examples from the OASIS Longitudinal MRI Data in Non-demented and Demented Older Adults.

Keywords: diffeomorphisms; geodesic shooting; inverse consistency; longitudinal registration; non-linear registration; shape modeling; symmetry; transitivity.

PubMed Disclaimer

Figures

Figure 1
Figure 1
A graphical representation of the full model. Each of the N images (f) is assumed to be a deformed version of the template (μ) scaled by a multiplicative inhomogeneity field [exp(b)] with additive Gaussian noise (precision λ). Each deformation is modeled by the composition of a rigid-body transform (ξ) parameterized by a vector of six parameters (q), and a diffeomorphic deformation (ϕ) parameterized by its initial velocity (v).
Figure 2
Figure 2
A mixture of two Rician distributions fit to an MRI intensity histogram (shown dotted). The fit is shown as a continuous line, whereas dashed lines are used to show the two Rician distributions.
Figure 3
Figure 3
The two simulated images.
Figure 4
Figure 4
Warped simulated images. First column: upper image warped to match lower image (from Figure 3). Second column: deformation fields. Third column: logarithms of Jacobian determinants (color-scales are the same for all examples, and in the range of −3 to 3). First row: results from ω1 = 0.001, ω2 = 0.001, and ω3 = 0.1, with an additional penalty on the square of absolute displacements of 0.0001. Second row: results from ω1 = 0.001, ω2 = 0.001, and ω3 = 2.0. Third row: results from ω1 = 0.05, ω2 = 0.0001, and ω3 = 0.0001. Fourth row: results from ω1 = 0.001, ω2 = 0.5, and ω3 = 0.001.
Figure 5
Figure 5
Warped simulated images. First column: lower image warped to match upper image. Second column: deformation fields. Third column: logarithms of Jacobian determinants (color-scales are the same for all examples, and in the range of −3 to 3). First row: results from ω1 = 0.001, ω2 = 0.001, and ω3 = 0.1, with an additional penalty on the square of absolute displacements of 0.0001. Second row: results from ω1 = 0.001, ω2 = 0.001, and ω3 = 2.0. Third row: results from ω1 = 0.05, ω2 = 0.0001, and ω3 = 0.0001. Fourth row: results from ω1 = 0.001, ω2 = 0.5, and ω3 = 0.001.
Figure 6
Figure 6
Illustration of the results obtained from matching a pair of images of a subject with mild cognitive impairment, which were collected 1895 days apart (OAS2_0002). The three images of the residual difference shown along the bottom are all windowed the same. Black indicates a value of −500 or less, whereas white indicates values of 500 or above.
Figure 7
Figure 7
Detail of the results obtained from matching a pair of images of a subject with mild cognitive impairment, which were collected 1895 days apart (OAS2_0002).
Figure 8
Figure 8
Data from the time-points of a subject with dementia (OAS2_0048). The top row shows the original scans after rigid alignment, whereas the bottom row shows the divergence of the estimated velocity fields.
Figure 9
Figure 9
Plots of main eigenvector from each subject’s divergence maps within the cranium. The columns show plots from those control subjects who were scanned only twice, plots from control subjects who were scanned more than twice, plots from subjects with dementia who were scanned twice, and plots from subjects with dementia who were scanned more than twice. Dotted lines show the best linear fit. Note that the plots are sorted according to their average slope, which was done for easier visualization. Some of the eigenvectors were also rescaled by −1, such that all the gradients are positive.
Figure 10
Figure 10
Mean images. Left: average of all subjects’ warped mean images. Center: average of the warped expansion rate maps of the control subjects. Right: average of the warped expansion rate maps of the subjects with dementia. Mean expansion rates are shown such that values of −0.04 or below are shown as black and values of 0.04 or above are shown as white.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Al-Mohy A., Higham N. (2009). Computing the Fréchet derivative of the matrix exponential, with an application to condition number estimation. SIAM J. Matrix Anal. Appl. 30, 1639–165710.1137/080716426 - DOI
    1. Arsigny V. (2006). Processing Data in Lie Groups: An Algebraic Approach. Application to Non-Linear Registration and Diffusion Tensor MRI. Ph.D. thesis, École Polytechnique, Paris
    1. Ashburner J. (2007). A fast diffeomorphic image registration algorithm. Neuroimage 38, 95–11310.1016/j.neuroimage.2007.07.007 - DOI - PubMed
    1. Ashburner J., Andersson J., Friston K. (1999). High-dimensional image registration using symmetric priors. Neuroimage 9, 619–62810.1006/nimg.1999.0437 - DOI - PubMed
    1. Ashburner J., Friston K. (2005). Unified segmentation. Neuroimage 26, 839–85110.1016/j.neuroimage.2005.02.018 - DOI - PubMed

LinkOut - more resources