Elsevier

NeuroImage

Volume 41, Issue 2, June 2008, Pages 243-251
NeuroImage

Cortical thickness measured from MRI in the YAC128 mouse model of Huntington's disease

https://doi.org/10.1016/j.neuroimage.2008.02.019Get rights and content

Abstract

A recent study found differences in localised regions of the cortex between the YAC128 mouse model of Huntington's Disease (HD) and wild-type mice. There are, however, few tools to automatically examine shape differences in the cortices of mice. This paper describes an algorithm for automatically measuring cortical thickness across the entire cortex from MRI of fixed mouse brain specimens. An analysis of the variance of the method showed that, on average, a 50 μm (0.05 mm) localised difference in cortical thickness can be measured using MR scans. Applying these methods to 8-month-old YAC128 mouse model mice representing an early stage of HD, we found an increase in cortical thickness in the sensorimotor cortex, and also revealed regions wherein decreasing striatal volume correlated with increasing cortical thickness, indicating a potential compensatory response.

Introduction

A recent study by our group examined differences between the YAC128 mouse model of Huntington's Disease (HD) and wild-type mice using deformation-based analyses of high-resolution post-mortem MR scans (Lerch et al., 2008). Along with the shrinkage of the striatum (Vonsattel et al., 1985), we found a volume increase in the sensorimotor cortex, a finding mirrored by a study showing that prodromal human HD also features increased cortical thickness (Paulsen et al., 2006) and exhibits compensatory functional responses in the thalamocortical circuit (Feigin et al., 2006).

The tools to investigate shape changes in the mouse cortex are, however, still limited. Histology and stereology-based techniques, while extremely powerful and capable of providing cellular level detail, are labour intensive and thus do not easily produce whole cortex coverage. They are also subject to tissue shrinkage and distortion from sectioning.

MR-based assessment of murine brain shape has recently gained increasing traction (Badea et al., 2007, Pitiot et al., 2007, Ma et al., 2005). The dominant methodology involves automated non-linear alignment of all examined mice towards a common space and is followed by either statistical analyses of the deformation fields or mapping an anatomical atlas back to each animal to measure structure volumes (Ma et al., 2005, Kovacevic et al., 2005, Chen et al., 2006). Among anatomical measurements, the thickness of the cortex in particular is endowed with biological significance, given that this measure can approximate the path taken by the cortical column, the prime computational unit of the cerebral cortex.

The last decade has produced a field of research based on accurately extracting shape models of the human cortex from MR images. As these techniques have become more sophisticated, intriguing new insights into the anatomy of the cortex have been gained, including mapping the pattern of thickening and thinning during development (Sowell et al., 2004) and ageing (Salat et al., 2004), in degenerative (Lerch et al., 2005) and psychiatric disorders (Shaw et al., 2006b), and in relationships between anatomy and behaviour (Shaw et al., 2006a, Walhovd et al., 2006).

Algorithms to measure cortical thickness from MRI follow one of four prototypes. The simplest, and rarest due to the labour intensive nature of the task, are manual measures using the digital equivalent of callipers (Ross et al., 2001, Meyer et al., 1996). A second set of methods which has gained increasing prominence involves the extraction of the inner and outer cortical surfaces using deformable models (MacDonald et al., 2000, Fischl and Dale, 2000, Kim et al., 2005). A third set of methods works without explicit polyhedral models, instead computing image boundaries to find the cortex then measuring the thickness of the cortex at the voxel level (Jones et al., 2000, Preul et al., 2005). The last type involves a mix of explicit surface parameterisation and voxel-based cortical thickness methods (Thompson et al., 2005, Miller et al., 2000).

The human cortex is convoluted with the folding pattern differing considerably between individuals. The mouse cortex, on the other hand, is lissencephalic and, given the similarities within inbred strains, almost identical among individuals. This key difference between mice and humans simplifies the creation of a cortical thickness algorithm considerably:

  • 1. The complex deformable model process employed to map the convoluted human cortex is not necessary for the mouse.

  • 2. The similarity across animals allows for near perfect image registration, a task still impossible within human populations.

  • 3. The ability to compute accurate registrations implies that the definition of the cortex and its boundaries can be created by mapping a pre-existing atlas to each animal.

These points suggest that a volume-based cortical thickness algorithm with boundary definitions gained from registration towards an anatomical atlas will be the easiest solution. Explicit surface parameterisation does, however, provide one key advantage: the creation of a surface coordinate system with a controllable number of nodes. Rather than comparing cortical thickness at every voxel, thickness statistics can instead be computed on a much smaller set of vertices over the cortical surface, leading to obvious advantages in controlling for multiple comparisons. Moreover, it allows the data to be smoothed along the manifold, an anatomically more sensible operation than using volumetric Gaussian kernels. A combined method, using voxel-based cortical thickness measures together with a surface coordinate system, suggests itself as an ideal choice of algorithm. It has the further advantage that every voxel in the cortex is part of a path between the inner and outer cortical surface.

The goal of this study, motivated by the intriguing finding of shape differences in the YAC128 Huntington Disease model mice as well as the lack of available tools to investigate the mouse cortex, is to design a framework for automatically measuring cortical thickness from mouse MRI and to use these new tools to further investigate the YAC128 mouse model. These new methods encompass a combination of mouse registration algorithms with adaptations of tools designed to measure cortical thickness from human MRI scans.

Section snippets

Methods

The proposed method estimates the thickness of the entire cortical manifold. The basic procedure used follows: (1) MRIs of all mice in the study are acquired. (2) A model independent consensus representation of all mice in the study is created using linear and non-linear registration techniques. (3) A previously existing atlas with an associated segmentation of the cortex is then mapped to the result of step 2. (4) The remapped cortical boundaries are then in turn mapped to each of the

Variance and power analysis

In order to understand the cortical analysis system described herein, a group of 20 male C57BL/6 mice, first described in Spring et al. (2007), were processed. The goal was to gain an understanding of the variance inherent in the thickness and surface area measures and thus know their statistical power.

The results are shown in Fig. 3. The mean cortical thickness in these mice was 0.89 mm ± 0.016 mm. Given two groups of 10 mice, one can expect to recover a 0.05 mm difference in cortical thickness

YAC128 mouse model

HD is a neurodegenerative disorder characterised by motor dysfunction, psychiatric disturbances and cognitive impairment that is caused by a CAG trinucleotide expansion in the HD gene on chromosome 4 (Huntington's Collaborative Research Group, 1993). HD is classically associated with initial reduction in volume and neuronal loss primarily localised to the striatum (Vonsattel et al., 1985), but widespread brain atrophy has been described in early to mid-stage HD in humans using magnetic

Conclusions

Here we have presented an automated algorithm for measuring cortical thickness at every point of the cortex from mouse MRI. This technique can greatly compliment the existing set of tools for analysing differences in image deformations by providing a more biologically relevant index of cortical change. The application to the YAC128 mouse model of Huntington's Disease showed an increase in cortical thickness in response to striatal degeneration, a potential compensatory response also seen in

Acknowledgments

The authors would like to thank Dr. Claude Lepage of the Montreal Neurological Institute for valuable fixes to the code. This work was supported by grants from the Michael Smith Foundation for Health Research, the Canadian Institutes of Health Research, the Huntington's Disease Society of America and the High Q Foundation. J.P.L is supported by the Canadian institutes for Health Research. J.B.C is supported by the Huntington Society of Canada. M.R.H. is supported by the Canadian Institutes of

References (46)

  • MillerM.I. et al.

    Bayesian construction of geometrically based cortical thickness metrics

    NeuroImage

    (2000)
  • PaulsenJ.S. et al.

    Brain structure in preclinical Huntington's disease

    Biol. Psychiatry

    (2006)
  • SpringS. et al.

    Sexual dimorphism revealed in the structure of the mouse brain using three-dimensional magnetic resonance imaging

    NeuroImage

    (2007)
  • TyszkaJ.M. et al.

    Statistical diffusion tensor histology reveals regional dysmyelination effects in the shiverer mouse mutant

    NeuroImage

    (2006)
  • WalhovdK.B. et al.

    Regional cortical thickness matters in recall after months more than minutes

    NeuroImage

    (2006)
  • BartensteinP. et al.

    Central motor processing in Huntington's disease. A PET study

    Brain

    (1997)
  • CollinsD.L. et al.

    Automatic 3D intersubject registration of MR volumetric data in standardized Talairach space

    J. Comput. Assist. Tomogr.

    (1994)
  • CollinsD.L. et al.

    Automatic 3D model-based neuroanatomical segmentation

    Hum. Brain Mapp.

    (1995)
  • FeiginA. et al.

    Preclinical Huntington's disease: compensatory brain responses during learning

    Ann. Neurol.

    (2006)
  • FischlB. et al.

    Measuring the thickness of the human cerebral cortex from magnetic resonance images

    Proc. Natl. Acad. Sci. U. S. A.

    (2000)
  • HenkelmanR.M. et al.

    High throughput microimaging of the mouse brain

  • HofP.R. et al.

    Comparative Cytoarchitectonic Atlas of the C57BL/6 and 129/Sv Mouse Brains

    (2000)
  • A novel gene containing a trinucleotide repeat that is expanded and unstable on Huntington's disease chromosomes. The Huntington's Disease Collaborative Research Group

    Cell

    (1993)
  • Cited by (105)

    View all citing articles on Scopus
    View full text