Dear FSL email list,
I have some questions regarding the interpretation of Oxford_asl output. I am using the latest version of Oxford_asl (version 3.9.13 for CentOS 7) downloaded from GitHub.
1) I understand from previous posts that in the folder containing partial volume corrected results (/.../native_space/pvcorr), the perfusion_calib_masked.nii.gz image is CBF for GM only, and perfusion_wm_calib_masked.nii.gz is CBF for WM. I tried to use "fslstats
perfusion_calib_masked -M" and "fslstats perfusion_wm_calib_masked -M" to get the mean CBF for GM and WM, respectively. However, the output is different from the numbers in perfusion_calib_gm_mean.txt and perfusion_wm_calib_wm_mean.txt. For example:
[jiyang@dhcp pvcorr]$ fslstats perfusion_calib_masked.nii.gz -M
60.859487
[jiyang@dhcp pvcorr]$ fslstats perfusion_wm_calib_masked.nii.gz -M
4.719992
[jiyang@dhcp pvcorr]$ cat perfusion_calib_gm_mean.txt
30.913142
[jiyang@dhcp pvcorr]$ cat perfusion_wm_calib_wm_mean.txt
4.747511
I was wondering why this is so, and which one is the current GM/WM CBF?
The values calculated within oxford_asl and reported in the text files will be different due to the different ROI used. For oxford_asl reported values we consistently use a GM (or WM) ROI mask defined by a high fraction of GM (or WM) - namely 70% (or 90%).
The masked perfusion image in the pvcorr subdirectory has been masked at 10% GM, i.e. it only shows voxels in which it might be reasonable to attempt to interpret the GM perfusion values. This means that when you do fslstats you are including a lot more voxels,
in practice those voxels with a low fraction of GM will still be somewhat less certain. Thus, I would still tend to calculate mean GM from only those voxels with a higher proportion, i.e. the 70& we use in oxford_asl.
2) Is it correct to get the global CBF from the perfusion_calib.nii.gz in "/.../native_space" folder (i.e. non-partial-volume-corrected), by using
fslstats /.../native_space/perfusion_calib -M?
This would be reasonable, but bear in mind that the non-zero voxels will potentially include the CSF - although the signal the signal there will be small, it might still count in the average. That said, when people report whole brian CBF it is likely they
just use a simple whole brain mask that includes CSF in the ventricles etc.
3) In order to calculate lobar CBF and CBF in subcortical structures, is it correct to warp MNI Structural Atlas to /.../native_space/pvcorr/perfusion_calib_masked.nii.gz, and use fslstats -M to get the mean CBF in each region?
Yes, warp the structure ROIs into the native space (threshold) then use with fslstats to calculate mean CBF. Like above, you might want to be stricter about including only voxels with a high proportion of grey matter - we provide a mask for that, it is
called gm_roi.nii.gz in the native_space directory.
4) I got the following error/warning messages when running Oxford_asl. Are they concerns or something I can ignore?
Cannot open volume /tmp/fsl_Ry3Ru2_ox_asl/init/basil/step2/mean_fwm for reading!
Cannot open volume /tmp/fsl_Ry3Ru2_ox_asl/init/basil/step2/mean_deltwm for reading!
Cannot open volume /tmp/fsl_Ry3Ru2_ox_asl/basil/step2/mean_fwm for reading!
Cannot open volume /tmp/fsl_Ry3Ru2_ox_asl/basil/step2/mean_deltwm for reading!
/bin/cp: '/tmp/fsl_Ry3Ru2_ox_asl/mask.nii.gz' and '/tmp/fsl_Ry3Ru2_ox_asl/mask.nii.gz' are the same file
/bin/cp: '/tmp/fsl_Ry3Ru2_ox_asl/tissseg.nii.gz' and '/tmp/fsl_Ry3Ru2_ox_asl/tissseg.nii.gz' are the same file
Cannot open volume /data_pub/jiyang/ASLprocessing/reconvertedRawData_reprocessing/pCASL_FSL/test11004/subjects/11004/11004_invKineticMdl_wT1/native_space/pvcorr/deltwm for reading!
Since you are getting pvcorr results, I am not worried by these errors particularly.
The command I ran to processing was:
${newOxfordAsl_folder}/oxford_asl \
-i
${indFolder}/${ID}_pCASL \
--iaf
${acq_order} \
--tis
${TI} \
--casl
\
--bolus
${bolus} \
--bat
${BAT} \
--slicedt=${slicedt}
\
--fslanat=${indFolder}/${ID}_T1.anat
\
--sbrain
${indFolder}/${ID}_T1_brain \
-c
${indFolder}/${ID}_M0 \
--tr
${TR} \
--cmethod
single \
--tissref
csf \
--csf
${indFolder}/${ID}_M0_ventricle \
--te
${TE} \
--fixbolus
\
--pvcorr
\
--spatial
\
--mc
\
-o
${indFolder}/${ID}_invKineticMdl_wT1 \
>>
${indFolder}/${ID}_logfile.txt
Any help is much appreciated!
Best regards,
Jiyang