Diffusion applications

Romain Valabregue CENIR ICM heavily inspired from https://fsl.fmrib.ox.ac.uk/fslcourse/lectures/fdt1.pdf and fdt2.pdf

Diffusion applications I Tractography Pitfalls Dissection of major bundles : Tract specific Region of Interest Subregion classification given the connectivity profiles : Cortical or sub-cortical Region subdivision

Quantitative measure

II

Tensor metrics : FA MD L1 Lr Advanced model : Noddi Quantitative measure of connectivity : tract density

Connectomme : Group comparison VBM / TBSS / Fixel based

A connection + a stregth

Tractography

Tractography ●

●

connection probability : “connectivity index” or “ connectivity strength” BUT it does not quantify how strong a connection is ! Rather : how robust it is against noies/uncertainty : The dominant path through the diffusion field...

Confounding factor : • • •

Connection length (longer are less probable) Geometric complexity Resolution of the spatial grid

Tractography ●

●

Density of tractograms : a quantitative measure ? Sift : Spherical deconvolution Informed filtering of tractograms

(Smith neuroimage 2013)

Tractography

Tract densité estimated from : b fod c streamline d streamline after sift

Tractography

Tractography

Tractography

Tractography

Tractography : disection To find specific track : Add anatomical prior Seed ROI : where the tract start (5000 fiber per voxel) Waipoint ROI : keep only tract that go through Exclusion ROI: Remove tract that goes through Termination ROI : stop the tract when it goad through

Alternative : Seed from the all white matter or gray-white interface → filter tract

Tractography : disection Examle : the Cortico-Spinal Tract

Tractography : disection Examle : the Cortico-Spinal Tract

Tractography : disection Examle : the Cortico-Spinal Tract

Tractography : disection

Tractography : classification ● ● ●

Define different target Seed to target tract Find the biggest

Tractography : classification

Tractography : classification Classification based on connectivity profie (no hypothesis on target) Example with area 6 : separate SMA and pre SMA

Tractography : classification

Diffusion applications I Tractography Pitfalls Dissection of major bundles : Tract specific Region of Interest Subregion classification given the connectivity profiles : Cortical or sub-cortical Region subdivision

Quantitative measure

II

Tensor metrics : FA MD L1 Lr Advanced model : Noddi Quantitative measure of connectivity : tract density

Connectomme : Group comparison VBM / TBSS / Fixel based

A connection + a stregth

Quantitative metrics ●

Tensor metrics

FA ~ eigenvalues Variance MD ~ Eigenvalues mean Longitudinal AD : λ1 Tranverse ADC : (λ2+λ3)/

FA decrease / MD increase → loss of structure → Decrease of longitudinal ADC ~ Axonal breakdow ? → Decrease of transverse ADC ~ Myelin breakdow ?

Quantitative metrics ●

Tensor metrics : BUT

Sensitive but not specific Crossing fiber

Quantitative metrics ●

Tensor metrics :

Sensitive but not specific : FA correlate do axon density

Quantitative metrics ●

Advances model ●

Axon caliber / charmed model –

●

Noddi : neurite orientation dispersion and density imaging – – –

●

Very long acquisition

Zhang NeuroImage 2012 2 or 3 shell. (64 dir B2000; 32 dir B700; 8 dir b 300) Simplify model : ● 3 compartiment intra-cellular, extra-cellular, CSF ● → Viso Vic + OD (orientation dispersion)

MAP : Mean apparent Propagator – – – –

Ozarslan neuroimage 2013 Extension of the gaussian diffusion process (tensor metric) to characterized the true average propagator ( modecular displacements in “r-space” ) RTOP / RTPP / RTAP Implementation in DIPY.

Quantitative metrics ●

Noddi example

Factors contributing to FA changes, can be disentangled with NODDI OD and Vic measuer

Quantitative metrics ●

Advances model ●

AFD apparent Fiber density –

●

Raffelt, neuroimage 2012

Fiber density

Group comparison ●

Normalisation to standard space

●

VBM on FA

●

Roi based

●

Tract-based

●

TBSS

●

Fixel based

Group comparison : normalisation ●

Normalise Anat scan corregister diffusion acquisition and apply normalise fsl / SPM / freesurfer

●

Normalise Diffusion B0 volumes or FA fsl / SPM

●

Normalise the tensor dtitk / ants

●

Normalise the ODF mrtrix / ants

Group comparison : VBM

Group comparison : ROI

Group comparison : ROI ●

Automatic ROI creation

●

Track based . –

●

●

To have a subject specific roi definition

Atlas based : –

Define ROI in MNI space (atlases)

–

Use the inverse transform to de-normalised ROI

–

Automatique Sub-cortical segmentation (freesurfer or first from fsl)

Compare mean FA in ROI across subjects

–

Group comparison : along tract

Group comparison : TBSS

Group comparison : TBSS

Group comparison : TBSS

Group comparison : TBSS

Group comparison : TBSS

Group comparison

Group comparison : fixel based

AFD Apperent Fiber Density: (Raffelt neuroimage 2012) Tract specific comparison of the FOD amplitude lob. ODF based normalisation Quantitative measure : AFD

Group comparison : fixel based

Differences in a single fibre bundle, in a region containing multiple fibre population