Using diffusion MRI for neurosciences
1) Voxel Based Morphometry for diffusion data
4) Connectivity-based parcellation
2) Bundle-based morphometry
5) Connectome analysis
3) Combining diffusion and functional stuff
7) Fiber bundles 6) Diffusion and and cortical folds spatial normalisation in development J.-F. Mangin
Architectural features that could be seen in vivo before diffusion: Macroscopic/global scale (10-2m) Brain regions
Morphology Mesoscale (10-3/-4m) Cortical columns Regional assemblies
Function
Architectony, connectivity
Visual areas?
Microscopic scale Local networks (1014 synapses) Neurons (10-5m), synapses (10-6m) J.-F. Mangin
Diffusion imaging is a new hope for mapping architecture
Cat Let us map brain architecture in vivo!
Behrens et al.
El Kouby et al.
Macaque J.-F. Mangin
What is beyond reach for diffusion MRI!
Santiago Ramon y Cajal Brainbow-coloured nerve cells, tracer-based tracing J.-F. Mangin
Adapting Voxel-Based Morphometry to diffusion data
You just have to change input
Mass univariate statistics (correction for multiple tests) Ashburner et al., SPM
J.-F. Mangin
What input for voxel or ROI-based analysis of diffusion data?
A lot of possibilities, from scalar features to fiber ODF
Anisotropy map Use input or input related data for normalization ! J.-F. Mangin
Fractional anistropy (FA) for VBM or ROI-based analysis Hundred of such studies, straightforward extention of VBM
Douaud et al, Huntington Disease You will find differences, you will have to interpret them… You do not really know where are the bundles J.-F. Mangin
Key issues when performing VBM
?
How much warping? Which template? How much smoothing?
Mass univariate statistics (correction for multiple tests) Ashburner et al., SPM
J.-F. Mangin
Overcoming some weaknesses of VBM: TBSS (FSL, Oxford)
Identify the most typical subject as the target for nonlinear registration of FA maps
1) Compute averaged map, 2) Skeletonization, 3) Alignment with MNI152
S. Smith, et al., et T. Behrens, Neuroimage, 2006 J.-F. Mangin
TBSS : individual FA maps versus mean FA skeleton
J.-F. Mangin
TBSS : Projecting individual subjects’ FA onto the skeleton
J.-F. Mangin
TBSS, an example: Controls > schizophrenia group comparison
Red means significant
TBSS
VBM
mean FA
controls
schizophrenics
J.-F. Mangin
Let us define bundles before thinking to morphometry
Millions of diffusion-based tracts
J.-F. Mangin
Sorting the tracts from various VOIs
Diffusion Tracking of Commisural Fibers
Diffusion Tracking of Parietal Association Fibers
Conturo T.E. et al., PNAS: 96;10422-27, 1999 J.-F. Mangin
Virtual dissection to get bundles
Catani et al., Ann Neurol, 2005 J.-F. Mangin
Splitting the bundles for morphometry: brainVISA toolbox
http://brainvisa.info
J.-F. Mangin
Qualitative analysis of the variability of bundle morphology
Catani et al., Ann Neurol, 2005 J.-F. Mangin
Atlas of white matter by DTI
From S. Mori
J.-F. Mangin
Statistical Anatomical Parametric Maps Averaging the bundles in Talairach space (you need to define the bundle of interest) Averaged left and right Broca and Wernicke
Averaged left and right Arcuate bundle
Left hemisphere
Right hemisphere
Parker et al., Neuroimage, 24, 2005 A lot of variations around that idea can be imagined J.-F. Mangin
Spatial normalization and coordinate systems: the 1D case !
Tubular objects global measurements: length, average section area
Travelling along the bundles: BrainVISA toolbox (http://brainvisa.info)
Dubois et al., 2006 J.-F. Mangin
Can we define bundles automatically ?
1.
Use Roi-based atlas normalized to the subject
3.
Use clusters of fMRI contrast (group-based or individual)
5.
Use balls centered around coordinates in Talairach space
(more adapted to connectome study, see further)
But do we have a list of the main bundles? a very preliminary one…
J.-F. Mangin
Can we infer a list of large fiber bundles ? We see actual folds, do we see actual bundles?
Cortical folding patterns
Fiber bundles
?
S. Mori
J.-F. Mangin
Inference of a structural model of bundles (DTI)
I Tracts
II Tract clusters
III Model of bundles El Kouby et al., MICCAI 2005
J.-F. Mangin
Mapping U-fiber bundles with MR diffusion imaging (HARDI data)
One subject
Twelve subjects Guevara et al., Neurospin
J.-F. Mangin
FA versus reaction times!
Tuch et al., PNAS, 102(34), 2005
Visual self-paced choice
(Add Dubois paper, J Neurosci.) J.-F. Mangin
The finest parcellations stem from connectivity
Architecture is the referential
Connectivity is the basis of architecture
J.-F. Mangin
Structural, architectural, structuralism?
We decompose everything… Is it always meaningful?
J.-F. Mangin
Projecting a parcellation elsewhere
Does not require perfect mapping of anatomical connectivity T.E.J. Behrens, H. Johansen-Berg et al., Nature Neuroscience, 6(7):750, 2003
Thalamus segmentation From cortex parcellation J.-F. Mangin
Diffusion-based versus fMRI-based (SMA/preSMA) Reordered connectivity crosscorrelation matrix identifies a change in connectivity profile
fMRI clusters Clusters mapped back onto brain
Johansen-Berg, Behrens et al, PNAS 101(36), 2004
Connectivity-defined clusters J.-F. Mangin
Collapsing connectivity profiles
Cachia et al., MedIA, 2003 Perrin et al., IJBI, 2008
Connectivity profiles of nodes of the cortical surface mesh Guevarra et al., ISBI, 2008
J.-F. Mangin
Parcellation of postcentral gyrus for 3 subjects
Guevarra et al., ISBI 2008
J.-F. Mangin
The main bundles creating the orange class subject 1
subject 3
subject 2
Guevarra et al., ISBI 2008
J.-F. Mangin
Whole cortical surface parcellation (connectome) Most of diffusion-based tracts go to the top of the gyri
Dealing with the curse of dimensionality Roca et al., submitted
J.-F. Mangin
Connectivity of clusters of activation? Surprisingly few results… Geniculo-Calcarine Tracts
LGN
Conturo T.E. et al., PNAS: 96;10422-27, 1999
J.-F. Mangin
Mapping the « connectome » of the human brain
Hagmann et al., PLOS biology, 2008
J.-F. Mangin
Modeling very large networks One world wide web / 6G human brains
J.-F. Mangin
Graph theoretical analysis
Bullmore et Sporns, Nature Review Neuroscience, 2009
J.-F. Mangin
Simple graph analysis (5 subjects)
Top 20% strength Hagmann et al., PLOS biology, 2008
J.-F. Mangin
Connectivity Backbone (998 ROIs, 1 subject)
Hagmann et al., PLOS biology, 2008
J.-F. Mangin
Connectivity Backbone (66 cortical regions, 5 subjects)
Hagmann et al., PLOS biology, 2008
J.-F. Mangin
Modularity and hub classification
Hagmann et al., PLOS biology, 2008
J.-F. Mangin
Comparison of Structural and Functional Connectivity
Hagmann et al., PLOS biology, 2008
J.-F. Mangin
Structural and Resting State Functional Connectivity Same data as in Hagmann paper
SC = diffusion-based connection
One example of connectivity profile Do you believe the in this data? Honey et al., PNAS, 2009
J.-F. Mangin
Is there a consistency with long distance connections?
Honey et al., PNAS, 2009
J.-F. Mangin
Diffusion-data could change spatial normalization Atlas warping approaches, coordinate systems MNI template
Fischl et al., MGH
Fischl et al. Lyttelton et al., MNI, 2007
Morphology-based approaches are blind to architecture J.-F. Mangin
Main bundles should be aligned !
LONI: Gyral matching to improve spatial normalization
J.-F. Mangin
Registration of cortical connectivity matrices
one male one female
Cathier et al., MMBIA’06, Newyork J.-F. Mangin
Diffusion data to understand the variability of the folding patterns
Variability
Connectivity
Development JF Mangin, Neurospin, CEA, France
J.-F. Mangin
Variability of the cortical folding patterns Is there some meaning behind?
Brain development architecture
tensions
Van Essen, Nature, 1997 J.-F. Mangin