Dear all
I am a young researcher just starting on fMRI/ freesurfer analysis in Cape Town. I got a good start after a Course the MGH guys gave us a few months back. Ran 53 subjects, T1 MPRAGE 3T Allegra Siemens on our local cluster (8 core nehalem X 300) (www.chpc.ac.za) : at one subject per node, took 12 -13 hrs on them, all recon-ed without a hitch as far as I can tell with tkmedit within 2 weeks. I have a few problems though in my initial "get feet warm" analysis:
First off I just wanted to run by you guys my stats analysis: I wanted to start with a very basic ROI using four different predefined regions: (Right and left) Medial Orbital Frontal Cortex, Lateral Frontal Cortex, Caudate and Globus Pallidus. I took the mean cortical thickness of one group and did a unpaired t-test two-tailed t-test with the other group in R (Which I have also started to use BTW).(13 in OCD group; 15 in control group) I also did a non-parametric Wilcoxon test in the same way as I inspected the data on box-charts in R and they looked a little suspicious to me. The t-tests were all insignificant. The Wilcoxon turned up as 0.054 on the Right MOFC which is a bit dicey but nevertheless right on the money as a colleague working independently on the DTI found some good results in the Right Uncinate. The recruiting is still ongoing though and the numbers will increase, so maybe that will sway the P values one way or the other.
Next I want to run a regression analysis in qdec using their ybocs scales(OCD severity): cortical thickness on the one arm and the score on the other.
Am I doing this part right? Is there a simpler or more effective way? Comments or concerns...
Also:
I found that the parcelation for the MOFC and LOFC are spot on as far as I can tell from inspecting the scans in tkmedit. The segmentation of the Caudate looks okay, but the Globus Pallidus is a bit off. Sometimes it includes bits of the (?)anterior commisure and surrounding white matter (?CS tracts). Is it a bit pale for this type of analysis? (Also, I might be asking a bit much, but has anyone tried and measure the pars interna and externa of the Globus Pallidus?)
Any suggestions?
Thanks again everyone for a fantastic course and an even more brilliant software package.
regards
Dr Stefan du Plessis CUBIC South Africa
Hi Stefan,
that sounds pretty reasonable. Pallidum is hard. We do have some newer tools that do better on the pallidum, but they are not ready for distribution yet. If you upload one dataset where the pallidum segmentation is significantly off, we'll take a look.
And thanks for the positive feedback :)
Bruce
On Thu, 22 Apr 2010, Stefan Du Plessis wrote:
Dear all
I am a young researcher just starting on fMRI/ freesurfer analysis in Cape Town. I got a good start after a Course the MGH guys gave us a few months back. Ran 53 subjects, T1 MPRAGE 3T Allegra Siemens on our local cluster (8 core nehalem X 300) (www.chpc.ac.za) : at one subject per node, took 12 -13 hrs on them, all recon-ed without a hitch as far as I can tell with tkmedit within 2 weeks. I have a few problems though in my initial "get feet warm" analysis:
First off I just wanted to run by you guys my stats analysis: I wanted to start with a very basic ROI using four different predefined regions: (Right and left) Medial Orbital Frontal Cortex, Lateral Frontal Cortex, Caudate and Globus Pallidus. I took the mean cortical thickness of one group and did a unpaired t-test two-tailed t-test with the other group in R (Which I have also started to use BTW).(13 in OCD group; 15 in control group) I also did a non-parametric Wilcoxon test in the same way as I inspected the data on box-charts in R and they looked a little suspicious to me. The t-tests were all insignificant. The Wilcoxon turned up as 0.054 on the Right MOFC which is a bit dicey but nevertheless right on the money as a colleague working independently on the DTI found some good results in the Right Uncinate. The recruiting is still ongoing though and the numbers will increase, so maybe that will sway the P values one way or the other.
Next I want to run a regression analysis in qdec using their ybocs scales(OCD severity): cortical thickness on the one arm and the score on the other.
Am I doing this part right? Is there a simpler or more effective way? Comments or concerns...
Also:
I found that the parcelation for the MOFC and LOFC are spot on as far as I can tell from inspecting the scans in tkmedit. The segmentation of the Caudate looks okay, but the Globus Pallidus is a bit off. Sometimes it includes bits of the (?)anterior commisure and surrounding white matter (?CS tracts). Is it a bit pale for this type of analysis? (Also, I might be asking a bit much, but has anyone tried and measure the pars interna and externa of the Globus Pallidus?)
Any suggestions?
Thanks again everyone for a fantastic course and an even more brilliant software package.
regards
Dr Stefan du Plessis CUBIC South Africa _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu