Print

Print
We reanlayzed data with SPM2 that had originally been analyzed with SPM99.
The task was a block design with a simple fixation baseline.
The images from 30 subjects were normalized and smoothed (FWHM = 6 mm).
The fixed-effect model was based on the canonical HRF.  The random effects
analysis showed robust activation for TASK>REST with both SPM99 and SPM2. 
The networks were essentially identical with the two versions of software,
with SPM2 more sensitive as indicated by larger max T, larger cluster sizes,
and larger number of clusters.
 
For the REST>TASK contrast, the results were quite different between
the two software versions.  With SPM99, the typical "default mode" network
was observed, including large clusters in the orbital frontal cortex,
posterior cingulate,and angular gyrus.  With SPM2, only two small clusters
(orbital frontal and posterior cingulate) were observed.  If the a less conservative
threshold is chosen in SPM2, then the network looks very similar to the one
observed with SPM99.  The peak tables for the two largest clusters for each
contrast and software version are shown below.
 
It seems that the TASK>REST contrast is enhanced and the REST>TASK contrast is
diminished in the SPM2 analysis as compared to the SPM99 analysis.  There is no
obvious difference in the respective ResMS images for the two versions of
software.  Does this make sense?
 
Thanks for your help
 
Rob
 
Robert J. Ogg, Ph.D.
Division of Translational Imaging Research
Department of Radiological Sciences
St. Jude Children's Research Hospital
tel: 901.495.2502 fax: 901.495.4398
email: [log in to unmask]
 

SPM2:
 
cluster   voxel    
p(cor) k   p(unc)  p(FEW) p(FDR) T     equiv Z p(unc)  x,y,z {mm}
 
[TASK > REST]
0        546 0           0      0          14.18  7.69    0         -4    6 52   
0      1136 0           0      0          10.61  6.73    0       -44 -74   -2
 
[REST > TASK]
0      26  0            0.006  0.001    7.15   5.39    0       -4   48 -10 
0      10  0.001      0.012  0.001    6.9     5.27    0       0 -50   36
   
Height threshold: T = 6.37, p = 0.000 (0.050)
Extent threshold: k = 5 voxels, p = 0.010 (0.001)
Expected voxels per cluster, <k> = 0.665
Expected number of clusters, <c> = 0.00
Expected false discovery rate, <= 0.00
Degrees of freedom = [1.0, 29.0]
Smoothness FWHM = 6.7 6.8 8.6 {mm}  = 3.4 3.4 4.3 {voxels}
Search vol: 1389168 cmm; 173646 voxels; 3276.6 resels
Voxel size: [2.0, 2.0, 2.0] mm  (1 resel = 48.97 voxels)
 
SPM99:
 
cluster   voxel    
p(cor) k   p(unc)  p(cor) T     equiv Z p(unc)  x,y,z {mm}
 
[TASK > REST]
0      333 0            0      11.22 6.92    0         -4    6 52     
0      479 0            0      10.92 6.83    0       -44 -74   -2
 
[REST > TASK]
0      671 0            0      9.64  6.41    0         2 -64  46   
0      119 0            0      9.39  6.31    0        -6  48 -14
 
Height threshold: T = 6.40, p = 0.000 (0.050)
Extent threshold: k = 5 voxels, p = 0.014 (0.001)
Expected voxels per cluster, <k> = 0.762
Expected number of clusters, <c> = 0.00
Degrees of freedom = [1.0, 29.0]
Smoothness FWHM = 7.4 7.4 9.2 {mm}  = 3.7 3.7 4.6 {voxels}
Search vol: 1287712 cmm; 160964 voxels; 2363.2 resels
Voxel size: [2.0, 2.0, 2.0] mm  (1 resel = 62.22 voxels)