USING GLASS PATTERNS AND fMRI TO IDENTIFY AREAS THAT PROCESS GLOBAL FORM IN MACAQUE VISUAL CORTEX.

1
USING GLASS PATTERNS AND fMRI TO IDENTIFY AREAS
THAT PROCESS GLOBAL FORM IN MACAQUE VISUAL CORTEX.





1. Max-Planck-Institute for Biological
Cybernetics, Spemannstrasse 38, D-72076 Tübingen,
Germany 2. HHMI and Center for Neural Science,
New York University, N.Y., N.Y. 10003. 3.
Present address Dept. of Psychological and Brain
Sciences, Dartmouth College, Hanover N.H. 03755
Peter.Tse_at_dartmouth.edu
P.U. Tse1,3, M.A. Smith2, M. Augath1, T.
Trinath1, N.K. Logothetis1, and J.A. Movshon2.
286.6
We set out to locate areas of the macaque
brain involved in processing form using fMRI.
Natural scenes or objects are not satisfactory
stimuli for this purpose because their
characteristics are difficult to parameterize and
control. Glass patterns, created by pairing
each dot in a random texture with another at a
specified spatial offset (Glass 1969), are useful
because form is defined by the global statistics
of dot pairs whose spatial correlation is purely
local. Glass patterns are well-controlled for
fMRI because the number of dot pairs and the
spatial power spectra are the same for all
stimuli. Wilson et al (1997) showed that
concentric Glass patterns are processed more
efficiently than other patterns with the same
local but differing global statistics, suggesting
that there may exist higher-order 'grouping'
filters tuned to particular patterns of
activation among local filters. We used the
patterns below 1. 9-pane concentric, 2. 9-pane
radial, 3. randomly oriented dot pairs, 4. 9-pane
translational, 5. translational. The patterns
subtended 20x20 deg, dot spacing was 0.2 deg, and
density was 100 dots/deg2. Each pattern was
replaced every 0.5 sec by a new pattern with
identical statistics.
Logic Differences in the BOLD signal between
Glass patterns will arise only in brain areas
that process non-local relationships among dot
pairs. Method We used the Glass patterns shown
below to activate visual cortex in 3 anesthetized
macaque monkeys. Eight-segment T2 weighted EPI
fMRI images (13 slices, FOV12.5, matrix128x128)
were collected on a 4.7T/40cm Biospec vertical
scanner with 50mT/m gradients, using quadrature
transmit/receive RF coils and gradient-recalled
EPI fMRI sequences. Voxel volume was 0.5x0.5x2mm,
TE40ms, TR750, and FA20 to 25 degrees. Data
are represented by significance maps calculated
using a general linear model (GLM) to test the
effects of different patterns and pattern
contrasts on the BOLD activation.
Results 2 Differential V1 activation from
different Glass patterns
translational x 9
translational x 9
random
random
concentric x 9
translational
concentric x 9
translational
radial x 9
radial x 9
of brain volumes (16 minutes)
1 concentric -random
-
Results 3 One monkey of three showed
differential activation in extrastriate areas
beyond V2, probably including V4.
icon1 icon2
1 concentric -translational
3 concentric -translational
-
Results 1 Activation of cortex revealed by
comparing all Glass patterns with blank
-

2 concentric -random
Monkey 1
Approximate area locations in sample horizontal
section
-
1.
2.
3.
Conclusion The BOLD signal in V1/V2 (and perhaps
V4) is greater for certain Glass Patterns than
others in the anaesthetized macaque. This may
reveal processing of large-scale spatial
correlations by neural networks in these areas.
2 9frame transl.-translational
4.
5.
-
The five stimuli
translational x 9
translational x 9
translational
concentric x 9
concentric x 9
translational
radial x 9
random
radial x 9
random
of brain volumes (16 minutes)
The lack of activation anterior to the lunate in
these scans may be an effect of anesthesia.