Software Overview

1
Software Overview

• 1) Installation
• 2) Scanning Procedure
• 3) File organization
• 4) Segmentation
• 5) Unfolding
• 6) Importing data into HFM toolbox
• 7) Demarcation
• 8) Analysis in Flat Space

2
Before you begin ...

• Read all of the following
• Zeineh et. al. NeuroImage 11(6)668-83, 2000
• Zeineh et. al. The Anatom. Record New Anatomist
  265111-120, 2001
• Zeineh et. al. Science Jan 24299(5606) 577-80,
  2003
• Amaral Insausti 1990. Hippocampal formation. In
  The Human Nervous System (G. Praxinos, Ed.), pp.
  711755. Academic Press, San Diego
• Duvernoy, H. M. 1998. The Human Hippocampus
  Springer, Berlin.
• Insausti et. al. Am. J. Neuroradiol. 19
  659671.

3
Instructions

• http//airto.loni.ucla.edu/BMCweb/HowTo/Hippocampu
  s/

4
Requirements

• PC segmentation
• UNIX MATLAB
• patience, forttude, computer skills

5
I. Installation

• PC download mrGray - segmentation program.
• http//white.stanford.edu/brian/mri/segmentUnfold
  .htm

6
I. Installation

• UNIX download, gunzip, and untar 5 files
• http//sourceforge.net/projects/mtl-unfolding/
• HFM Main Module
• Stanford Unfolding Code
• (Brian Wandell)
• Stanford mrLoadRet
• Visualization Utility
• Image utilities
• Example data

7
I. Installation

• UNIX download updated image utilities
• Automated Image Registration (AIR)
• http//bishopw.loni.ucla.edu/
• imconvert (UCLA image conversion utility)
• http//airto.loni.ucla.edu/BMCweb/SharedCode/ImgLi
  b/imconvert.c.html
• http//airto.loni.ucla.edu/BMCweb/SharedCode/Share
  dSoftware.htmlAnchor-UCLA-35829
• Compile everything
• Make data directories and links
• Test with the sample data

8
I. Installation

• Sign up for the mailing list!
• http//lists.sourceforge.net/mailman/listinfo/mtl-
  unfolding-help

9
II. Scanning - Structural

• Need high in-plane resolution (512 x 512) for
  less than 0.5 mm in plane pixel size
• Go perpendicular to hippocampal long axis
• Get as many structural slices as possible
• An adjunct 3-D volume can help discern sulci

10
II. Scanning - Functional

• Also need high in-plane resolution (128 x 128
  minimum) for less than 1.5 mm pixel size
• Alignment with structural slices is most
  important - get them coplanar to make it easy
• If acquired in different planes, will need
  accurate 3D alignment via AIR / SPM / other tools.

11
III. File Organization

• For each subject, 4 main directories
• raw - all fxnal aw data goes in here
• segment - all structural data for segmentation
• air - motion correction for fxnal data
• loadret - it all gets synthesized here

12
IV. Segmentation
13
IV. Segmentation
White

• 1) Segment white matter

CSF
Gray
Anterior
1) PHG 2) FG 3) lat. vent. 4) fornix
Posterior
14
IV. Segmentation
White

• 2) Segment CSF

CSF
Gray
Anterior
5) hippocampal and collateral sulci 6) adjacent
vessels/cisterns 7) boundaries - encapsulate
segmentation in 3D
Posterior
15
IV. Segmentation

• 3) Interpolate segmentation

Especially for white matter 1) smooth
transitions 2) thicken boundaries (all gt size 1
fxnal voxel) 3) eliminate topological errors
16
IV. Segmentation

• 4) Grow out layers of gray matter

17
V. Unfolding

• Use the matlab unfolding routine to
  computationally stretch gray matter mesh so it is
  uniplanar, then compress all planes or layers
  onto one.
• Select a seed pixel in the subiculum, enter in
  data parameters
• The algorithm does the rest!

18
VI. Importing data into HFM

• Imports
• Raw Structural Data
• Unfolded MTLs
• Motion Corrected Fxnal Dat

19
VI. Importing data into HFM

• Enter parameters
• slices, of fxnal runs
• image sizes

20
VI. Importing data into HFM

• Roughly align structurals w/ functionals

21
VI. Importing data into HFM

• Import all fMRI time series

22
VI. Importing data into HFM

• Finely align fMRI with structurals

23
VII. Demarcating the MTL

• Boundaries can be created using ROI generation
  tools and stored in one of 12 slots for each
  hemisphere.

24
VII. Demarcating the MTL
1
3
2

• 1. Select the boundary, 2. Draw, 3. Store

25
VII. Demarcating the MTL

• You can project the boundary to flat space
  (smaller projection is better).

26
VII. Demarcating the MTL

• 1) Locate the 1st slice where HC head starts -
  this divides anterior and posterior HC
• This corresponds to the
• boundary between ERC/PRC
• and PHG
• Mark this boundary from
• ERC vertex to FG (12)

27
VII. Demarcating the MTL

• 2) Locate the slice with the hippocampal feet in
  it (pez), likely next slice anterior

• Anterior to this boundary, CA 1, 2, 3, and DG
  are all present and indistinguishable

• Mark this boundary from the superior tip of CA 1
  all the way medially (10)

28
VII. Demarcating the MTL

• 3) Mark these posterior boundaries
• Medial Fusiform Vertex - lateral startpoint of
  CoS (1)
• CoS - depth of collateral sulcus (2)
• Sub PHG - most medial point of PHG (4)
• CA 1 Sub - medial termination of HC body (7)
• CA23DG CA 1 - 45º off HC body/fissure (8)

8
7
4
8
2
7
1
1
4
2
29
VII. Demarcating the MTL

• 4) Mark these anterior boundaries
• PRC FG - lateral startpoint of CoS (1)
• CoS - depth of collateral sulcus (3)
• ERC PRC - medial startpoint of CoS (5)
• Sub ERC - medial/superior vertex of PHG (6)
• CA 1 Sub - inferior to middle of HC head (7)
• CA23DG CA 1 - 45º off HC body/fissure (8)

Shift with CoS depth
1
3
8
5
6
7
8
7
3
6
5
1
30
VII. Demarcating the MTL

• 5) Demarcations can all be projected to flat
  space

• 6) Smooth boundaries can be automatically fit!

31
VIII. Analysis - Images

• Paradigm Correlation Analysis
• Need paradigm file
• Generate from your behavioral paradigm, smooth by
  HRF

32
1. Create correlation map ...
3.Adjust threshold
2. View the phase map
increase
decrease
33
VIII. Analysis - Images
Color enhance activations

• Make it snazzy!

Export to photoshop!
Superimpose boundaries
34
VIII. Analysis - ROI

• ROI analysis
• create ROIs

• edit as necessary

35
VIII. Analysis

• plot ROI timeseries

36
VIII. Analysis

• Export all timeseries to spreadsheet

37
Time Series for Subregions
Sustained
Late Activation
No Activation
Parahippocampal
CA 2, 3, DG
CA 1
Fusiform
Subiculum
Entorhinal
38
Future Releases

• Hippocampal Flat Template
• Warping
• Automated Segmentation

39
Acknowledgements
UCLA Medical Scientist Training Program NIH
National Research Service Award Ahmanson
Foundation Pierson-Lovelace Foundation Brain
Mapping Medical Research Organization Tamkin
Foundation Alma and Nick Robson Norma and Lyn
Lear Jennifer Jones-Simon UCLA School of
Medicine Neuropsychiatric Institute Department of
Neurology
Susan Bookheimer Stephen Engel John
Mazziotta Barbara Knowlton Joaquin Fuster Itzhak
Fried Charles Wilson Mark Cohen Paul
Thompson Bernice Wenzel Gary Small Roger
Woods Arthur Toga Russ Poldrack Paul Rodriguez