1. Govinda R. Poudel, PhD
Senior Research Fellow
Mary Mackillop Institute of Health Research
Australian Catholic University
Melbourne, Victoria
Email: Govinda.poudel@acu.edu.au
Modelling network spread of degeneration and
disconnection in Huntington’s disease
2. Outline
Network degeneration in HD
Cortico-striatal network in HD
Atrophy and disconnection
Evidence from neuroimaging studies
Network diffusion model
Background
Implementation
Network diffusion as a predictive model in HD
Predicting volume loss in HD
Predicting disconnection in HD
3. Context: HD impacts the human brain at multiple scales
Cellular
Ensembles
Regions
Topological
Spatial
Temporal
Microscale
Mesoscale
Macroscale
Instantaneous
Changes with disease progression
Evolutionary
changes
mHTT
cell
damage
Spiny
neurons
Cortico-striatal
4. HD
Post-mortem studies provided the first evidences of network-level changes in
HD brain
• Caudate/putamen atrophy correlated with the
severity of cortical atrophy.
• Loss of frontal white matter correlate with both
cortical and striatal atrophy.
• Loss of white matter around the basal ganglia
correlated with atrophy of the cortex and putamen
but not with the caudate nucleus.
Halliday et. al., 1998
5. Atrophy in the striatum covaries with cortical atrophy
0
0.50
Caudate atrophy covariance
N=70 HD (pre and manifest) compared to controls (N=36)
Z-scores= HD volume – mean(control volume)
standard deviation (control volume)
Putamen atrophy vs middle frontal cortex atroph
R2=0.25, p<0.001
Putamen atrophy covariance
6. HD
Covariance network show abnormal connectivity in HD
Track-HD study, Minkova et. al., HBM., 2015
“Fronto-striatal circuits are
among the earliest and
most consistently affected
in the prodrome”
Predict-HD study, Ciarochi et. al., 2016
control > low > medium > high
7. Mapping white matter networks in HD
DWI based tractography
Fronto-parietal regions connected to caudate
9. Poudel et. al., 2014, Neurobiology of Disease
White matter networks dysfunction in HD
10. HD
Why is there a stereotypical pattern of degeneration in HD?
Emergent property of the brain damage at multiple scales
Abnormal aggregation and clearance of polyQ-htt protein.
Transcriptional dysregulation
Disruptions of axonal transport and excitotoxicity.
Network spread
Topological vulnerabilities
Trans-connectome spread of pathology
Active transport
12. X2= Disease factor in R2
X1= Disease factor in R1
β=Diffusivity constant
C1,2=Connection strength
C1,2
X2 X1
On the whole brain, this can be solved as:
Where H is Laplacian matrix
X0 is initial condition
Network diffusion model
Network diffusion model is based on passive
diffusion of disease factors from high density
regions to low density regions
13. • Data from the IMAGE-HD study
• 26 Manifest Huntington’s Disease
• Age-matched Healthy Controls (N=26)
• Neuroimaging data
• T1-weighted (3D-MPRAGE)
• DWI data (64 directions, b=1200)
Key questions
Can a model of diffusion explain volume loss in HD?
Are the brain pathways most vulnerable to diffusion also most susceptible?
Network diffusion to identify vulnerable brain networks
18. Are there specific epicenters of disease in HD?
P<0.001
(A)
(A)
Seeding the diffusion from Accumbens area was associated with most
stable pattern
Diffusion seeding from all brain regions
Step 1: Network diffusion is run repeatedly
through out the brain using brain regions as seeds
Step 2: Predicted degeneration is correlated with
measured degeneration to generate correlation versus
time curves.
19. Diffusion from the striatum best predict atrophy
Poudel et. al., under review
20. Can network diffusion also identify susceptible connections?
Identify the edges most vulnerable to
disconnection by taking disease spread
process into consideration.
For an edge ei,j, it’s susceptibility to
disconnection is determined by diffusion
vulnerability of the corresponding nodes (vi
and vj) .
Edge vulnerability = Total diffusion in the
nodes connected by the edge.
X2= Disease factor in R2
X1= Disease factor in R1
β=Diffusivity constant
C1,2=Connection strength
C1,2
X2 X1
EV = X1+ X2
21. Measuring disconnection in HD using DWI
Generation of connectome maps in HD using HARDI data
HD versus Controls using Network based statistics
22. P<0.05, FDR corrected using NBS
Disconnection in HD using network based statistics
Cortico-striatal tracts
Fronto-parietal tracts
Bilateral tracts
23. Remaining ROIs
Can network diffusion also identify susceptible connections?
Seeding the diffusion process from the Putamen, Pallidum, makes the connections most susceptible
24. White matter connections linking most vulnerable nodes are
most susceptible to disconnection.
Poudel et. al., under review
26. • HD is the disease of brain networks. It is associated with a spatial pattern
which covaries over time and with the progression of degeneration.
• White matter changes shows disconnection of stereotypical cortico-striatal
pathways.
• Network models allow us to better explain the behaviour of brain networks in
HD, which has been characterized by the large scale imaging studies in HD.
• Simple models of diffusion on network may help explain/predict the pattern of
grey matter degeneration and white matter disconnection that are hallmarks
of Huntington’s disease.
Conclusions
27. Acknowledgment
IMAGE-HD Study
Lead PI: Prof. Nellie Georgiou-Karistianis
(Monash University)
Prof. Gary Egan (Monash University)
Prof. Julie Stout (Monash University)
Assoc. Prof. Phyllis Chuah
Other RAs, Post-docs, and Imaging Technicians associated
with the project
Participants
Funding bodies