Thank you Paul,
Kindly one last question.
If I use the command  "fslmaths <SUV_in_MNI_T1_2mm.nii> -inm 1 SUVR.nii -odt float "

This command output the image "SUVR.nii" which is intensity normalized. If I open this image in "fslview" and on top of this image I open the input image  "SUV_in_MNI_T1_2mm.nii" I can see in the intensity field in "fslview" how the formula is applied
(intensity within each voxel = intensity within the voxel / mean image intensity). I understood this part and it is clear to me.

But when I add mask to the previous command
fslmaths <SUV_in_MNI_T1_2mm.nii> -mas <MNI152_T1_2mm_brain_mask.nii.gz> -inm 1 SUVR.nii -odt float

The calculation is not really (intensity within each voxel = intensity within voxel / mean image intensity). What the mask do for this formula?

Thank you for any clarification!

Jon

Hi Jon,

I suspect that you would still need to include the covariate. Even if normalising the images would account for different doses, including the covariate anyway will do no harm.

But I don't know much about PET imaging, so hopefully somebody else will be able to give you advice on this.

Cheers,

Paul

On 24 October 2016 at 21:55, John anderson <[log in to unmask]> wrote:
Hi paul,
Thank you very much for your response. Kindly I have one more question, and I highly appreciate your response or any of FSL experts.

I have PET images for 20 subjects. The dose of the radio tracer used to collect these PET images is not the same between the subjects and as a results some images were acquired with higher doses than the other images. I need to include all these images (after processing and setting it in MNI) in a group analysis (FSL/randomise) and I think that I need to use dose as a covariate in GLM.

My question is: when I uses the command bellow and normalize all the images using the flag "inm 1" does this result with an images free of the effect of dose and I remove the effect of the covariate. In other words after normalizing the images. Do I need to use the covariate?

fslmaths <SUV_in_MNI_T1_2mm.nii> -mas <MNI152_T1_2mm_brain_mask.nii.gz> -inm 1 SUVR.nii -odt float


Thank you for any comment

Jon



Hi Jon,

Sorry - no - this has nothing to do with voxel volume. I was using the word 'volume' to refer to the entire 3D image. The "-inm 1" option to fslmaths simply results in the image divided by its mean.

Cheers,

Paul

On 22 October 2016 at 14:57, John anderson <[log in to unmask]> wrote:
Hi Paul,
Your response is highly appreciated. Thank you very much!

Kindly I would like to inquire about the volume? Do you mean we divide the volume for every voxel by the mean signal over all the voxels?

In my case I am trying to normalize PET images using "fslmaths" is it correct :

inm = (Volume in a specific voxel)/mean PET signal over all the voxels

Thanks again
Jon


Hi Jon,

With the '-inm <value>' flag, every voxel in the input volume is multiplied by <value> / M, where M is the mean across all voxels.

As you have used '-inm 1', this has the effect of simply dividing the volume by its mean.

Cheers,

Paul

On 22 October 2016 at 01:25, John anderson <[log in to unmask]> wrote:
Dear FSL experts,
I am working on a PET analysis and I created the standardized uptake value (SUV) images. From SUV images I want to use "fslmaths" to generate the standardized uptake value ratio maps (SUVR)

I used the command :
fslmaths <SUV_in_MNI_T1_2mm.nii> -mas <MNI152_T1_2mm_brain_mask.nii.gz> -inm 1 SUVR.nii -odt float

I highly appreciate if any of FSL experts explain to me how the flag "inm" in the command "fslmaths" works to do intensity normalization?

Depending on my literature review (which is not enough to fully understand the facts behind this). I found that in intensity normalization, the SUV values (or any other values in the voxels like FA, MD ,...) in every voxel in the brain are divided by the average of intensity over all the voxels in the brain. Is this correct?

Thank you for any input

Jon