Print

Print
Thanks a lot, Steve!

Sorry to bother you again for a couple of questions regarding to Gaussian random field theory correction. :)

1. If we want to correct our results in a two-tailed manner (e.g., see ADHD > Controls and ADHD < Controls significant results simultaneously), we need to correct the positive values (ADHD > Controls) at a Cluster P < 0.025, then correct the negative values (ADHD < Controls) at a Cluster P < 0.025, finally merge the negative part and the positive part and will get a final P < 0.05, is this correct?

2. For Z Statistical image, from the documents (Flitney, D.E., & Jenkinson, M. 2000. Cluster Analysis Revisited. Tech. rept. Oxford Centre for Functional Magnetic Resonance Imaging of the Brain, Department of Clinical Neurology, Oxford University, Oxford, UK. TR00DF1.) and the source code (smoothest.cc and infer.cc in FSL 4. 1.9), it seems the degree of freedom will not be used (set to default value 100). The only place I can find degree of freedom, v, used is transforming the estimated covariance matrix (for smoothness) from the residual field to Z0 for calculating the probabilities of clusters found in the Z-statistic image (Section 2.3 in Flitney & Jenkinson, 2000).
Screen shot 2012-01-26 at 2.30.22 PM.png
For Z Statistical image inputted into easythresh, if the smoothness is estimated from Z-statistical image directly, then the transformation from residual field to Z0 is no longer need i.e., degree of freedom will not be used. Is this understanding correct?

For the transformation, Interpolate in smoothest.cc (Ln 132):
retval = (j->second - i->second)/(j->first - i->first)*(v - i->first) + j->second;

Could this be:
retval = (j->second - i->second)/(j->first - i->first)*(v - j->first) + j->second;

Screen shot 2012-01-26 at 4.44.54 PM.png

3. For smoothness estimation,  the individual voxel variance is estimated by (Section 2.2.2 in Flitney & Jenkinson, 2000)
Screen shot 2012-01-26 at 2.09.14 PM.png
But in the code, it's calculated by (smoothest.cc Ln 324):
S2[X] += 0.5 * (Sqr(R(x, y, z, t)) + Sqr(R(x-1, y, z, t)));

Is using the average of the individual voxel variance and its neighbor voxel variance better for estimation?

Thank you very much for your patience and time!

Best,

Chao-Gan


On Fri, Jan 20, 2012 at 2:02 AM, Stephen Smith <[log in to unmask]> wrote:
Right - yes, for (e.g.) the Z case it's set to 100 as anything larger than that gives virtually the same result.
Cheers.



On 19 Jan 2012, at 20:56, YAN Chao-Gan wrote:

Thanks, Steve!

Sorry for my unclear statement.

To my understanding, easythresh will call smoothest without passing the DOF parameter. However, in smoothest, DOF will default to 100 if nothing specified, is that right? What DOF should we use if we perform easythresh on a Z statistical image?

Best,

Chao-Gan

On Thu, Jan 19, 2012 at 12:48 PM, Stephen Smith <[log in to unmask]> wrote:
I don't think it does - easythresh (at least my version) doesn't pass in a DoF.
Cheers.


On 18 Jan 2012, at 22:14, YAN Chao-Gan wrote:

Dear FSL experts,

Recently I used easythresh on a Z stats map for multiple comparison correction.

I found easythresh will call smoothest with the degree of freedom set at the default value of 100. The DOF of statistical Z is not 100, right? Is there any reason to use 100 as DOF for Z stats map?

Thanks!

Chao-Gan


---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------






---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------