Transcript PowerPoint

What 3.2 needs from 1 and 2
I. Needs from genetics (Kennedy and Potkin):
1) Use test file genotypes to begin to
operationalize genetic and imaging data
analyses
– the test file for the Potkin N=28 sample has real BDNF and SNAP25
gene data available.
– We need imaging phenotypes (segmentations) to get started: eg whole
brain grey matter vs BDNF, SNAP25; then brain subregions eg
hippocampus, dorsal parietal, etc)
– From March 05 after ethics approval in Toronto, will have data for
many more genes available eg Dopamine D1, D2, BDNF haplotypes,
myelin system genes…)
What 3.2 needs from 1 and 2
The most important need (requirement!) is the integration of existing
and newly modified tools to be interoperable in a stable,
efficient environment to support the following:
Anatomically accurate and functionally meaningful segmentations
Optimal, dedicated, neuroanatomist-friendly tractography program
Integration of a neuroscientist-friendly tractography with
segmentations, other functional data (SEM, PET, fMRI, fa/ADC, etc)
Further integration of these with genetic data
Interactive, dyNAMIC, Visualization of large, multi-scale,
multi-subject, multi-modal data
What 3.2 needs from 1 and 2
I. Needs from genetics (con’t)
2) Examine the PEDSPLIT software to
understand the first steps to test two genes
in the same sample (problems: the current program
is for parent-to-child transmission of gene variants – we
need to generalize to non-familial singleton cases. Next
important step will be to modify this program or invent
another one that will analyze more than 2 genes
simultaneously)
What 3.2 needs from 1 and 2
II. Anatomical accuracy and integrated data-Jim
Fallon
1- Anatomical accuracy
Tannenbaum – apply Fallon rule–based averaged
functional anatomical unit for DLPFC for
27 of the UCI 28 scz patients
2- Integrating anatomical, functional data
COMT effects
McCarthy
Kilpatrick
Sem corr= -.46
fa/ADC= .72/a
+.26
.67/e
+.42
.75/d
+.26
.45/b
+.72
.35/c
2- Integrating anatomical, functional data
II. Anatomical accuracy, tractography and
integrated data-Jim Fallon
2- Integration of data
Need to integrate the population and individual subject tractography,
with multiple standard segmentations, plus functional anatomical units,
functional imaging data, structural equation modeling, etc to create
more sophisticated circuitry diagrams.
-Problem with existing automatic segmentations is that you can only
segment the cortex in FreeSurfer automatically. Then you can only bring
into Slicer the cortical surface with the colorized labels but have no way
of overlaying the labels on the anatomical slices in Slicer. When you then
add fMRI statistical data in Slicer you can see the “blobs” in the slices views
but without the anatomical labels it isn’t integrated well. That is to say,
the surface labeled object from FreeSurfer doesn’t contain the imported
statistical maps and the T1 anatomicals as the slices views.
-Same as above for subcortical structures which currently aren’t available
via automated segmentations.
-Integrate tractography growing tools (fiber tracts+QBALL gliphs+FA/ADC, etc)
with cortical and subcortical parcellations and statistical activation data, eg. SPMs,
structural equation modeling, fMRI.
II. Anatomical accuracy, integrated data and tractography3- Tractography needs and analyses:
a) We need a seeding/ROI “ABCD filter by March 1
ILF
Occipito-temporal pathways
Inferior longitudinal fasciculus
Occipito-temporal pathways
Unwelcome
extended family
Members: Loop
of Flechsig-Meyer
Occipito-temporal pathways
C
A
Need for ABCD
filters together
with hand-made
tractography
3)Immediate needs in order to start the tractography analyses
b) We need a quick easy way to modify segmented areas, eg, for
us in tractography we need more than segmented cortical areas,
we need the lobular white matter beneath cortex to capture the
tracts. Also need automated subcortical segmentations available
in connecting tracts.
C
A
How to
segment
these
‘grey’
Areas?
Typical cortical
segmentation
Need to fill in the
sublobular white
matter to capture
tract stems for
growing tracts
Immediate needs in order to start the tractography analyses
c) We need to get up close to most tracts to grow them. However,
when you draw in close and start growing fibers, you don’t
know where they end up going, since they leave the field of
view. So you end up growing extraneous fibers, and then when
you pull back to low mag, you realize you have to delete half
of them. So, multiple images are needed to view at low and
high mag simultaneously.
Seeding and admiring the tracts from afar:
The computer scientist’s approach
Up close and
Personal; the
warm and fuzzy
neuroanatomist’s Also need to create “Armor piercing bullets”. Perhaps Q-Ball
can help greatly here, but we’d need the tractography and the
approach
directional glyphs visible in the same window.
II. Anatomical accuracy
b) From Tannenbaum lab – apply Fallon rule–based
averaged functional anatomical units for other brain areas
More segmentations than found in present seg programs
are needed, ie, Functional-Anatomical-Units, eg, eye fields,
parietal attentional, superior temporal gyrus subdivisions
SEF
6
8
6
7
4
40 39
Standard Brodmann segmentation
FEFs
FEFa
FEFi
PEF
Second level needs for tractography, still pressing
a) The features mentioned should be on a tractography
menu with icons/pictures/circuits so that it is visually intuitive
for use by a learning-impaired neuroanatomist.
b) Easy, intuitive import of other tools and databases such as
Freesurfer, Taliarach, Q-Ball, Leventon, LONI, CSAIL, fa/ADC
statistical analyses, literature, etc. info.
c) Need to agree on the meaning of the tracts and structures in
the context of endophenotypic clinical, genetic, cognitive, etc data
d) Need for highly trained students and techs to learn all
the techniques AND the biology. It took me 30 years to begin
to understand the neuroanatomy of the entire brain and I’m really
not sure I can teach another human or a computer do do this in
under 5 years. There are only a handful of “arrow neuroanatomists”
who know the entire brain, and the few of us still alive are either
dying, being politically redistricted, or assasinated.
e) Need a standard RGB color palette that doesn’t need to be
imported each time
f) Deep voxels and visualization
g) Need to integrate all the data into hybrid visualizations
Falko Kuester’s UCI hybrid reality visualization laboratory. The images depict multiple SMART
board displays of 2D orthogonal slices and a synchronized 3D stereoscopic
projection system using Kuester-developed software and middleware. Time
series image sequences (4D) and ability to annotate images are available but not depicted
Also need more shirts like his.
h) The tools are difficult for someone like me to use. If the only thing
I want to do is tractography, I have to press the same
20 buttons or more every time I lay down a tract. Some of the
terminology is confusing and ambiguous and the learning curve is pretty
darn steep. A 2 Gig laptop quickly gets overwhelmed after the first
tract is drawn, so there are numerous crashes, and deleting just one
fiber may take 5 minutes…..
Last, but perhaps most important overall…
THE TRACTOGRAPHERS NEED A SEPARATE FRONT END
ON SLICER DEDICATED JUST TO DTI, USING INTUITIVE
ICONS AND UNAMBIGUOUS TERMINOLOGY, AND
INTEGRATING ALL PRESENTLY AVAILABLE TOOLS FROM
NAMIC COLLABORATORS. THE PROGAM NEEDS TO RUN
MUCH FASTER AND CONSUME LESS ‘EXTRANEOUS’
SUBROUTINES, MEMORY, AND TASKS. REPEAT STEPS
NEED TO BE SCRIPTED INTO ONE-STROKE BUTTON PRESSES.
TAKE ALL THE BEST TOOLS AND INTEGRATE THEM
INTO A SINGLE MACHINE/SOFTWARE WORKING