Hi SPM users,
I am new to SPM and I wanted to make sure I get a good handle on the process to see if I am correct for VBM analysis. I have 20 control subjects and 20 patients with no repeated scans. I am using a GE Signa Hdx 3.0 Tesla with SPGR acquisition of T1 weighted images
File Conversion
1) Convert DICOM files to a compressed NIFTI file per subject (i.e. 120 Dicom images compressed into 1 NIFTI file per subject) Repeat procedure for all 40 individual patients. All 40 should NOT be merged into a single NIFTI file.
Voxel size of T1 weighted images is .982x.982x1.5mm^3
2)Realignment Procedure
Would I only use estimate and not the estimate/reslice option due to the fact that reslicing(from my understanding)is only used for time series analysis(fmri) when you are just looking for changes in the same slice over a period of time? If my assumption is incorrect, does reslicing the image create isotropic voxels (if so what size) and would Images 2..n be recommended for VBM?
Also I see options for Realign, Realign and Unwarp and Coreg. Which of these would be recommended for VBM?
For this step(whichever option is chosen) do i do each NIFTI file (1 per patient) one at a time? If I enter all 40 Nifti scans of all patients at the same time does the software recognize this and does the realignment per individual patient or does it assume all 40 NIFI scans belongs to one patient and tries to realign all scans together?
3)New Segment or segment for VBM since it says in the manual New Segment has newer features but not fully integrated with SPM8?
-What BIAS FWHM is recommended for vbm analysis if there is not too much noise?
4)Run Dartel (Create Template)
Voxel size of T1 weighted images is .982x.982x1.5mm^3
The procedue states that voxels should be isotropic(stated Ch. 26 pg 185 of manual but not stated Ch. 39). Is there any way to make them isotropic in SPM in order to use this toolbox? If not is there another toolbox recommended to use in SPM for VBM analysis. I have also read that there is a tool from the Mayo Clinic which can be used to create isotropic voxels but I would rather use SPM tools if there are any built in tools, if this step is indeed necessary.
5)Normalize to MNIspace
--for this step do i normalize the grey matter first and then rerun this step for white matter?
--If not, Do I enter all 40 flow fields of grey matter (20 control then 20 patients) for grey matter and then 40 flow fields of white matter (20 control and 20 patients) for a total of 80 flow fields at once and make sure images are in same order?
Voxel Size - should i normalize to 1.5x1.5x1.5 to make it isotropic (originally .982x.982x1.5mm^3) or leave this option alone
Boundary box - How do i determine this?
6) Factorial design specification
2 sample t test
--Group 1 all the grey matter scans of 20 control
--Group 2 all the grey matter scans of 20 patients
--Grand mean scaling? It says only used for PET data, but does that also mean include for VBM analysis since they are on the same tab?
--ANCOVA - same as above...its says only PET but do they mean PET/VBM?
--Covariates - would any calculated paramters such as deformations etc be used as a covariate?
--Masking
---Threshold none
---Implicit - yes
---Explicit - none selected
Global Calculation - Would i choose Mean for VBM analysis or leave as omit (specified as PET only but not sure if VBM using T1 images is ok)
Global Normalization - Again it says PET only but not sure if they really mean PET/VBM since its the same initial option for starting the program
--Overall Grand Mean scaling - YES - 50ml/dl/min Grand mean scaled value
--Normalization - Proportional or ANCOVA
I am testing to look for GM and WM changes between patient and control subjects. I am not sure if ANCOVA by subject (disease vs control) or proportional would be the better option in this case (or none of the above).
Thank you very much in advance for the help.
Sincerely,
Ajay Kurani
University of Illinois at Chicago
[log in to unmask]
|