This program performs multiple comparison correction using a combination of
individual voxel probability thresholding and minimum cluster-size thresholding.
The probability threshold is supplied by the user, the cluster-size threshold
is computed by this program using Monte-Carlo simulations.
Example:
valphasim -in zmap.v -z 2.576 -fwhm 8.0 -report list.txt -out test.v
This command calculates the minimum cluster-size threshold and
corresponding p-values given the original significance
threshold of z=2.576 (uncorrected).
The input image 'zmap.v' serves as a mask for the Monte-Carlo simulations.
It should therefore have the same geometrical properties
as the actual zmap to be analyzed, i.e. the same voxel size, spatial extent, etc.
The Monte-Carlo simulation fills the space
of non-zero voxels in this zmap-mask with randomly generated p-values and counts
the number of false positives.
The parameter '-fwhm' should correspond to the spatial smoothness of the data.
This value depends both on the size of the spatial filter used when preprocessing
the data plus the intrinsic smoothness of the data prior to spatial filtering.
In 'vcolorglm' and 'v2ndlevel', and all second-level programs
this value is automatically estimated, and written
into the header of the zmap. It can be read using the command:
> grep -a smoothness zmap.v
Note, that there no smoothness estimation implemented in 'vwhiteglm'. Please use 'vcolorglm' to perform a smoothness estimation.
Note further, that the smoothness estimation should always be counter-checked
by visual comparison with the output image generated by 'valphasim'.
The output image represents one of the randomly generated test images.
It can be visualized as if it were a zmap using 'vlv' oder 'vlview'.
It can be used to assess whether the 'fwhm'-parameter reflects the spatial smoothness
of the data.
The output of 'valphasim' is a
table of corrected p-values and minimum cluster sizes:
Example output:
fwhm: 5.00mm, seed: 555, numiter: 1000
z: 2.57600, p: 0.00500
image dims: 80 100 80
num voxels: 109429
voxel size: 2.00 x 2.00 x 2.00 = 8.00 mm^3
|
  mm^3   |
  freq   |
maxfreq   |
corrected p |
8.00    | 10163 | 0 | 1.00000 |
16.00    | 6843 | 0 | 1.00000 |
24.00    | 4881 | 0 | 1.00000 |
32.00    | 4152 | 0 | 1.00000 |
40.00    | 3121 | 0 | 1.00000 |
... |
832.00    | 3 | 2 | 0.05500 |
840.00    | 1 | 1 | 0.05300 |
848.00    | 2 | 2 | 0.05200 |
856.00    | 2 | 2 | 0.05000 |
872.00    | 1 | 1 | 0.04800 |
880.00    | 1 | 1 | 0.04700 |
From this table, the minimum cluster size has to be
determined which is associated to the largest p-value < 0.05. In our
example, this is 0.048, and the minimum cluster size is 872 mm^3.
This result can be applied to the data using 'vpretty'
and 'vblobsize' in the following way:
vpretty -in
zmap.v -out correct_zmap.v -pos 2.576 -minsize 872
vblobsize -in zmap.v -out correct_blob.v
-pos 2.576 -minsize 872 -system talairach
The colums 2 and 3 of the above table provide the frequency with which a
cluster size has been detected or was found to be the largest cluster
size within the Monte-Carlo simulations, respectively.
Remark: 'valphasim' is based on the publications cited below.
S.D. Forman, J.D. Cohen, M. Fitzgerald, W.F.Eddy, M.A. Mintun, D.C. Noll.
Improved assessment of significant activation in functional magnetic resonance
imaging (fMRI): use of a cluster-size threshold.
MRM 33:636-647 (1995).
J. Xiong, J.-H. Goa, J.L. Lancaster, P.T. Fox. Clustered Pixels
Analysis for Functional MRI Activation Studies of the Human
Brain. Human Brain Mapping 3:287-301 (1995).