Functional Neuroimaging of Perceptual Decision Making

1
Functional Neuroimaging of Perceptual Decision
Making

• Group E
• Elia Abi-Jaoude, Seung Hee Won,
• Sukru Demiral, Angelique Blackburn
• Faculty Mark Wheeler
• TA Elisabeth Ploran

2
Background
http//whyfiles.org/209autism/images/slide3.gif
Philiastides and Sajda, 2007
3

• Objective
• Does perceptibility (visibility) affect decision
  making?
• Does activity in the FFA predict decision making
  activity?
• Hypothesis
• Relative activity in areas identified in facial
  processing will vary proportionately with
  visibility of face images likewise with object
  activity in those areas identified in object
  perception.
• As difficulty increases, activity in the ACC, AI,
  and DLPFC will increase. This will vary
  inversely with perceptual activity.

4
PART IBLOCK DESIGN
To identify areas of perceptual activity of faces
and objects
5
Perception Task
To identify areas of perceptual activity of faces
and objects
every 2s For 30s
30s
30s
every 2s For 30s
30s
Scan Parameters
2 runs each with 4 blocks Run 1
Face/Object/Face/Object Run 2 Object/Face/Object/
Face Run order counterbalanced across
participants 15 images per block, random
presentation order

• 3T Siemens scanner
• TR 2s
• TE 40ms
• Voxel Size
• 3.2 x 3.2 x 3.2mm
• Flip angle 70 degrees
• Slices 38
• Structural MP-RAGE

6
Data Processing

• Structural/Functional Alignment
• All functional scans were aligned to the MP-RAGE
  structural scan
• Talairach Transformation
• Reconstructed images were transformed into
  Talairach space
• Smoothing
• Smoothed to 6.4 x 6.4 x 6.4mm (2 voxels)

• Slice Time Correction
• To compensate for slices taken over 2s interval,
  used sinc function to time correct all slices to
  first slice
• Motion Correction
• In 6 directions x, y, z rotational and
  translational
• Intensity Normalisation
• Set most frequent intensity in each subject to
  1000 to normalise intensities across participants

RW Cox. AFNI Software for analysis and
visualization of functional magnetic resonance
neuroimages. Computers and Biomedical Research,
29162-173, 1996.
Avi Preprocessing Script http//nrg.wikispaces.co
m/page/code/4dfptools
7
Block Design Individual Analysis
FacegtObject
R
L
ObjectgtFace
Consistant with previous findings e.g. Scherf,
S. et al. 2007. Developmental Science,
10(4)F15-F30.
Plt0.01
8
Block Design Group Analysis
As FFA is highly variable across individuals, we
were unable to localize the FFA in the group
analysis. This is a common problem with small
sample sizes and could be ameliorated with a
larger sample size.
All Images at Talairach Coordinates X49.0
mm Y55.0 mm Z-14.0 mm
Plt0.01
S4
S3
S6
S2
9
Variable FFA Location Across Participants
S4 X-1mm Y38mm Z4mm
S6 X49mm Y55mm Z-14mm
S3 X41mm Y37mm Z-29mm
10
Block Design Summary

• We were able to localize face and object areas in
  the individual analysis which conformed to
  previous findings
• Our group analysis did not have enough power to
  identify the FFA

11
PART IIEVENT RELATED DESIGN
Determine how decision making varies with
perceptual difficulty. Determine face and object
differences as a result of perceptibility using
ROIs defined in the Block Design and comparing to
ACC differences due to difficulty.
12
Discrimination Task Face vs. Object
To determine how decision making varies with
perceptual difficulty
Randomized Jitter 0,2,4,6s
100ms
200ms
1600ms
75ms
320 Trials in 2 ER runs, same scanning
parameters as BLOCK
5 Visibility
40 Visibility
13
Optimization of Task
Pilot Data Accuracy as a function of Mask Levels
at 100ms Stimulus
Percent Accuracy
5 10 20 25 30 35 40 50
Percent Visibility
14
ResultsBehavioural Data



Visibility Level
15
ER Individual Analysis

• Markers for each stimulus type
• 3 visibility levels (Low, Med, High)
• 2 stimulus types (Face and Object)
• 2 Accuracy (Correct and Incorrect)
• Due to time constraints we were unable to adjust
  our analysis to fix the Signal to Noise.

16
Future Expectations ROI analysis of ER
Object Presentation 5 low predicted
activity 40 high predicted activity
For Face Presentation 5 low predicted
activity 40 high predicted activity
ACC 40 low predicted activity 5 high predicted
activity
17
Summary

• Using a block design, we were able to identify
  face and object areas in our population.
• We would like to use these regions to identify
  relative changes in these areas and the ACC,
  DLPFC, and AI at an individual level during our
  event related design.

18
We have learned

• How to design an fMRI experiment
• About the steps in data preprocessing
• How to do individual subject analysis using the
  GLM
• Reasonable data at an individual level becomes
  less reasonable once averaging starts, need a
  larger sample size.
• Ideas about how to incorporate fMRI into research
  using our current modalities (EEG, NIRS) when we
  return home.

19
Acknowledgments

• The MNTP Program
• Seong-Gi Kim
• Bill Eddy
• Mark Wheeler
• Elisabeth Ploran and Jeff Phillips
• Tomika Cohen and Bec Clark
• NIH