MUKHERJEE Pratik Diffusion MRI Connectome UCSF Mar 23 2012
<p class="MsoNormal" >MUKHERJEE Pratik Diffusion MRI Connectome UCSF Mar 23 2012</p>
<p>
intro by Bob Dougherty (head of NCI, Stfd)</p>
<p>
he is a neuroradiologist at UCSF...
he did training at Wash Univ St Louis.. spent time in Marcus Raichles lab... doing tractography..</p>
<p>
he is in Dept of Radio (and does clin med a couple of days per week)
he also at Center for Molec and Func Ingaging at China Basin / UCSF...</p>
<p>
slide: DTI... he describes.. get 3 eigenvalues...</p>
<p>
can get the primary eigenvector (corresponds to prime eigen value) shows direction of fiber...</p>
<p>
fractional anisotropy.. correl with behavior...
many studies...</p>
<p>
he has done chronic concussion patients...</p>
<p>
Niogi, Mukh. in AJNR 2008</p>
<p>
spatial extent of white matter injury on DTI...</p>
<p>
large # of tracts with low FA have bad function...</p>
<p>
slide: struc dissociation of attentional control and memory in adults with and without mild TBI...</p>
<p>
LDFR long delay free memory score...</p>
<p>
people with high FAs do better than one with low FAs...</p>
<p>
left hemisphere ACR correlates with conflict in normal adults..
ACR = anterior corona radiata...</p>
<p>
also looked at concussion patients...</p>
<p>
same with bilater uncinate fasciculus...</p>
<p>
*</p>
<p>
are DTI parameters of difft white matter tracts independent?? are they?</p>
<p>
mean diffusity = md, ad =avg diffus. and rd = radial diffusivity...</p>
<p>
hypthesis some tracts are more closely related than others by DTI micro params...</p>
<p>
looked at 6 difft tracts..
IFO = infer fronto-occip fascic
ILF = infer longitudinal fasc... (missed the rest)</p>
<p>
slide: compare R and L tracts..eg IFO frac aniso in R vs L v high corr r = 0.88 (spearman)
but depends on tracts...</p>
<p>
UF = uncinate fascic...</p>
<p>
CST corticospinal tract</p>
<p>
looked at 12 tracts across 2 hemispheres.</p>
<p>
look at corr matrix 12 by 12...</p>
<p>
look at Wahl, Li, Ng et al Neuroimage 2010 in 51:page 521...</p>
<p>
look at dendrogram... based on fA values</p>
<p>
look at microstruc correl of white matter tracts...</p>
<p>
R and L IFO v high corr.</p>
<p>
note L and R AF = arcuate fascic are NOT symmetric (language)</p>
<p>
CST R and L correl with each other but NOT with any other tracts...</p>
<p>
so, summary: wide variation in FA correl strengths among homol tracts</p>
<p>
big Q: are these micros struc corr. present at birth or driven by usage-dependent palsticity...</p>
<p>
do acallosal indivs have reduced microstruc corr btwn homol tracts
due to reduced interhemisph functional connectivity...?</p>
<p>
(other great Q's I missed)</p>
<p>
TBSS = tract-based spatial statistics...</p>
<p>
TBSS use FA skeletonization... look at centers of tracts to get best data...</p>
<p>
fist do PCA/IA to decompose resting stae fMRI data on single subject...</p>
<p>
did 25 indep componets...</p>
<p>
completely data driven analy... of WM = white matter FAs...</p>
<p>
decomposes CC = corpus callos.. into 3 difft areas (data driven)... difft funcitons corres to difft FAs</p>
<p>
SLF = super. longit. fascic... </p>
<p>
ref for this is Li, Yang, Nguyen Hum brain Mapp 2011 ...</p>
<p>
got 21 components (tracts)...</p>
<p>
anter commis of CC etc.</p>
<p>
how might this be used practically...</p>
<p>
improve voxel based analy.... studying 29 matched TBI patients
at 1 month post injury vs 16 controls</p>
<p>
see reduced microstruc integrity... doing SBA = source based analyssi of group DTI data..</p>
<p>
prefrontal tracts are most commony injured... prefrontal areas...</p>
<p>
*</p>
<p>
now, looking at Connectome...</p>
<p>
<b>Networks of the Brain </b>book by <b>Olaf Sporns </b>(regional connectome)</p>
<p>
vs Connecotme (neuronal connectome...microscopic... synaptic level)
see Connectome new book by Sebastian Seung</p>
<p>
how to do it..</p>
<p>
do standard parcellation using Freesurfer...</p>
<p>
use HARDI probabilistic tractography... by Julia Owen in his lab..</p>
<p>
400 by 400 matrix who is connected to who... color shows connection..</p>
<p>
see obvious diff btwn agenesis people and virtual agenesis people...
huge differ.</p>
<p>
network metrics (from olaf sporns... Neuroimage 52)</p>
<p>
compare metric btwn virtual split brain controls and complete agenesis pts</p>
<p>
he has a great post doc Etay Ziv md, phd in his lab..</p>
<p>
info theoretic modularity... (he did phd in graph theory)</p>
<p>
cingulum is highly abnl in the agenesis pts...</p>
<p>
look at RA index, modularity...</p>
<p>
</p>
<p>
</p>