functional mask
32 views
Skip to first unread message
Martine van Schouwenburg
May 27, 2008, 7:19:17 AM5/27/08
to Distance toolbox
I’m experimenting a bit with the distance toolbox. However I don’t
really get the difference between the global mask and the functional
mask. I applied the global mask, the functional mask with an
uncorrected p-value of 0.01 and the functional mask with an
uncorrected p-value of 0.0001, and all three give me the same results.
Can anyone tell me some more about the use of the functional mask?
Cheers,
Martine van Schouwenburg
really get the difference between the global mask and the functional
mask. I applied the global mask, the functional mask with an
uncorrected p-value of 0.01 and the functional mask with an
uncorrected p-value of 0.0001, and all three give me the same results.
Can anyone tell me some more about the use of the functional mask?
Cheers,
Martine van Schouwenburg
Sébastien Mériaux
May 30, 2008, 7:50:30 AM5/30/08
to Distance toolbox
Dear Martine,
When we started to work on inter-subjects distances, we observed that
the distance between two subjects can be highly influenced by which
voxels are used for the computation (it seems that it is not the case
for your dataset). Therefore we proposed to use a "functional mask
image" to restrict the distances computation on all activated area
across subjects for the specified contrast of interest.
Our "functional mask" is computed by thresholding the SPM statistical
images ('spm_T*') of all subjects for the specified contrast of
interest and by pooling all the suprathreshold voxels. One limitation
of using this kind of "functional mask" is that the SPM statistical
images are obtained from the corresponding SPM contrast images
('con_*'), which are also used to compute the inter-subjects
distances. We rather recommand to construct your own functional mask
(for example by drawing on anatomical images some regions of interest
for your contrast) regardless of the SPM contrast or statistical
images and then use it as "user mask" during distances computation.
Regarding the fact that you obtained the same results with a
functional mask computed with an uncorrected p-value of 0.01 and with
an uncorrected p-value of 0.0001, it seems that the inter-subjects
variability (observed in the suprathreshold regions for your contrast
of interest) is not affected when restricting the number of voxels
used for distances computation.
Don't hesitate to ask if you need more information.
Best regards,
Sébastien Mériaux.
On 27 mai, 13:19, Martine van Schouwenburg
When we started to work on inter-subjects distances, we observed that
the distance between two subjects can be highly influenced by which
voxels are used for the computation (it seems that it is not the case
for your dataset). Therefore we proposed to use a "functional mask
image" to restrict the distances computation on all activated area
across subjects for the specified contrast of interest.
Our "functional mask" is computed by thresholding the SPM statistical
images ('spm_T*') of all subjects for the specified contrast of
interest and by pooling all the suprathreshold voxels. One limitation
of using this kind of "functional mask" is that the SPM statistical
images are obtained from the corresponding SPM contrast images
('con_*'), which are also used to compute the inter-subjects
distances. We rather recommand to construct your own functional mask
(for example by drawing on anatomical images some regions of interest
for your contrast) regardless of the SPM contrast or statistical
images and then use it as "user mask" during distances computation.
Regarding the fact that you obtained the same results with a
functional mask computed with an uncorrected p-value of 0.01 and with
an uncorrected p-value of 0.0001, it seems that the inter-subjects
variability (observed in the suprathreshold regions for your contrast
of interest) is not affected when restricting the number of voxels
used for distances computation.
Don't hesitate to ask if you need more information.
Best regards,
Sébastien Mériaux.
On 27 mai, 13:19, Martine van Schouwenburg
Reply all
Reply to author
Forward
0 new messages