Hi everyone,
I ran a VBM analysis between unpaired groups and got 4 significant clusters.
Cluster –i tfce_corrp_tstat1 –t 0.95 -o cluster_size osize=cluster_size > results.txt
fslmaths –dt int cluster_size -thr -uthr -bin
Where –thr and -uthr is the minimum-maximum (threshold) value from the index_cluster column at the results.txt where every significant cluster is displayed (1-4). In my case the -thr and -uthr values were the same (1), because only cluster 1 was of my interest.
I checked the mask and everything went ok at this point (1 cluster = 1 mask).
fslmeants –i –m > results_cluster.txt
I would like to calculate the total volume of that particular cluster (voxels, mm3) for every subject (n=56) to do correlations/regressions with other behavioral/cognitive variables in SPSS. I think my problem is in Step 3, do someone know a better way to calculate volumes on a particular cluster?
I also performed a manual extraction for the data using kernel spheres but I’m not happy with the results/method because even if I merge several spheres to create a mask to cover all the cluster I still got some voxels out of it (and viceversa).
Thanks!
If I understand right, you want the GM volume in that cluster for each person.
How about binarising the GM segmentation
fslmaths GM -thr 0.5 -bin GM_binary
, and then running the fslmeants on that?
So that would give you the average of the GM in the cluster, and mutliplying by the cluster volume should give GM volume in mm.
I think I found the solution:
fslstats -K –V -M | awk {‘print $1 * $3’}
fslstats calculates the number of non-zero voxels (-V) of that cluster and mean intensity (-M) within the cluster (-K)
This command will print in terminal the volume of each subject for that cluster.
Hi Jonatan,
Thank you for posting the solution you used. It has helped me arrive at what I think may be a solution for a similar problem. I'd like to know the volume of GM in a certain mask that DIDN'T come up significant. I basically want to say 'Group X had a volume of Y in the Z region.'
I wonder if this is an OK approach. Please let me know if you have any thoughts:
1) split up GM_mod_merg_s2 into individual volumes using fslsplits
2) Then use fslstats with the -V flag to calculate the voxels and volume in each region specified by the mask (with the -k flag). If I do this I believe I still need to convert the volume to mm3. I'm not quite sure how to do this.
Does anybody have any thought on this approach or other suggestions? Thanks very much!
Best
Sheena
Hi Sheena,
I'm glad that this solution worked for you. My thoughts:
(1) For what I know the "GM_mod_merg_s2" contains the smoothed GM mask for every subject as volumes (volume 1 = subject 1). In step (3) of the original message I used fslmeants to get the intensity value for each subject in that particular cluster, so I'm afraid I don't understand well why you're running fslplits here.
If you know the order of your subjects (p.e. template_list) you should easily know which value belongs to each when you get the fslmeants output (results_cluster.txt). Anyway, don't take my words for granted because I'm just starting with FSLl!!
Besides statistical significant clusters I also tested some ROI's as you did in other areas and I followed the same steps. I just skipped steps 1 and 2, because I didn't have to extract any cluster for any statistical map. You have to be aware of the fact that usually people calculates the total brain volume for each subject as well in order to controll this variable (to make sure that particular subject didn't have a lower volume in that ROI because they have a lower total brain volume).
(2) I'm still not sure of what kind of measure is exactly the volume extracted. I believe the regular size of a voxel is actually in mm3 (2x2x2mm).
Hi Jonatan,
I see what you mean now and I am just starting out with FSL too. I think you're right that the volume is reported in mm3 by fslstats.
The reason I thought of fslsplit is so that I can get an accurate volume for each person. It seems to me that in your approach you're taking the total volume of all subjects (fslstats on the GM_mod_merg image) and dividing it by n. Therefore, aren't you using an average volume per subject rather than each subject's specific volume? Please correct me if I'm wrong! I thought by doing fslsplit that I would be able to get each unique volume, but perhaps I'm adding an unnecessary step!
Best,
Sheena
Hi Sheena,
yes, I'm dividing the total volume of that cluster by n, but then I'm multiplying that result by the voxels intensity value for each individual in a particular cluster (total volume / n * intensity value -> this last one came out from the results_cluster.txt).
Maybe your approach is valid as well! I'm just following some protocols we have in our research group.
Thanks Jonatan - Your approach sounds good - just wanted to make sure I understood it. I will try both and see how they compare :)
Best
Sheena
Hi Jonatan and Sheena,
I'm curious to know why you're interested in GM volume? The intensity of the voxel (i.e. grey scale) I think is a representation of the grey matter content. That is, both volume and density. So, if you're looking for a measure of grey matter content, say the amount of grey matter, the fslmeants command should give you just that. Please see the attached references. Have a look at the attached paper for a discussion about density vs volume and how after modulation the grey scale represents both.
Of course, you could have a specific hypothesis about GM volume.
http://www.sciencedirect.com/science/article/pii/S1053811901907864
Hi Peter,
Thanks for that suggestion! I did end up using the fslmeants command as described in this forum: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;3c23a239.1011
Cheers
Sheena
Hi Guys
I wanted to calculate the longitudinal brain volume. I am using FSL Sienax. baseline scans were acquired on one Scanner same strength. The scanner was upgraded and their followup at 18 months and 3 years were scanned on the upgraded scanner same strength. Do you guys suggest me if I can copere these data or not. Also Should I use the same parmeters for Sienax for calculating Brain volume for all the three time points I mean pre and post upgraded scans. (However my results doesn’t show decreasing trend). Could anyone suggest regarding the same.
- Badges
- Science topic
- Similar topics
- Correction