Hi Bruce,

A copy of the surfaces would be much appreciated. Just let me know where I can download them from. Ideally if you can tar the whole colin27 directory so I have the aseg and aparc info as well that would be great. I think including this and/or the mni152 and average_305 surfaces would be great, though this might bloat the size of the download.

Not sure if you had a chance to look at my files, but do you think the problem I have is just a bad skullstrip, or is it an intensity issue from the conversion?

thanks, mishkin

On Mon, Nov 30, 2009 at 8:53 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:

Hi Mishkin,

yes, I generates surfaces for the Colin27 dataset. It actually doesn't have great gray/white contrast, probably due to differential B0 distortions in the different sessions. Maybe we should include it as part of our release? In any case, I can send it to you.

cheers Bruce

On Mon, 30 Nov 2009, Mishkin Derakhshan wrote:

...

Hi, I'm trying to get the surfaces for the colin27 brain that is included with the mni packages. Has anyone managed to do this before?

I thought I would have an easy time given that this linear average of 27 brains has very good signal to noise and nice contrasts, and even though recon-all completes without errors, the surfaces are way off. The main problem appears to be the skull stripping.

My input volume is here: http://opus.bic.mni.mcgill.ca/public/colin27_t1_tal_lin.mnc.gz

These are the steps I took and where I think the problem might be:

1. gunzip colin27_t1_tal_lin.mnc.gz
2. recon-all -i colin27_t1_tal_lin.mnc -subjid colin27
3. tkmedit conlin27 orig/001.mgz

At this point the image the brightness and contrast levels of the image look very washed out. I am hoping this is only a display issue and that the underlying intensities of the image are still intact and is not a result of something going on in mri_convert, but maybe this is what is causing the downstream skull strip to fail. Can someone take a look? http://opus.bic.mni.mcgill.ca/public/001.mgz 4. recon-all  -autorecon1 -subjid colin27 5. tkregister2 --mgz --s colin27 --fstal Results are not perfect but reasonable ie. about the same as many other runs I've had that have passed so I don't think it is the registration. 6. tkmedit colin27 brainmask.mgz -aux T1.mgz It is clear this is where the problem is. http://opus.bic.mni.mcgill.ca/public/brainmask.mgz http://opus.bic.mni.mcgill.ca/public/T1.mgz 7. I have subsequently tried: recon-all -skullstrip -wsthresh 10 -clean-bm -no-wsgcaatlas -subjid colin27 (basically extract the entire head) recon-all -skullstrip -wsthresh 5 -clean-bm -no-wsgcaatlas -subjid colin27 (basically extracted the entire head) recon-all -skullstrip -wsthresh 5 -clean-bm -subjid colin27 (extracted the brain, but left large chunks of skull) recon-all -skullstrip -wsthresh 2 -clean-bm -subjid colin27 (extracted the brain, but left quite a few chunks of skull)

I'm concerned that the -wsthresh is so low in order to get a reasonable result, that's why I think the intensities might be off (see step 3).

I'm re-running -autorecon2/3 now to see if the surfaces are better, but if someone could take a look at my input file and the converted 001.mgz to see if it isn't a problem there it would be much appreciated.

thanks, mishkin _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Also the ch2 dataset that comes with MRICro suffers from the same problem. It's the averaging of several acquisition from one brain.

Is it possible to infer that too many acquisitions is worse than one acquisition?

-- PPJ

2009/11/30 Bruce Fischl fischl@nmr.mgh.harvard.edu

Hi Mishkin,

yes, I generates surfaces for the Colin27 dataset. It actually doesn't have great gray/white contrast, probably due to differential B0 distortions in the different sessions. Maybe we should include it as part of our release? In any case, I can send it to you.

cheers Bruce

On Mon, 30 Nov 2009, Mishkin Derakhshan wrote:

...

Hi, I'm trying to get the surfaces for the colin27 brain that is included with the mni packages. Has anyone managed to do this before?

I thought I would have an easy time given that this linear average of 27 brains has very good signal to noise and nice contrasts, and even though recon-all completes without errors, the surfaces are way off. The main problem appears to be the skull stripping.

My input volume is here: http://opus.bic.mni.mcgill.ca/public/colin27_t1_tal_lin.mnc.gz

These are the steps I took and where I think the problem might be:

1. gunzip colin27_t1_tal_lin.mnc.gz
2. recon-all -i colin27_t1_tal_lin.mnc -subjid colin27
3. tkmedit conlin27 orig/001.mgz

At this point the image the brightness and contrast levels of the image look very washed out. I am hoping this is only a display issue and that the underlying intensities of the image are still intact and is not a result of something going on in mri_convert, but maybe this is what is causing the downstream skull strip to fail. Can someone take a look? http://opus.bic.mni.mcgill.ca/public/001.mgz 4. recon-all -autorecon1 -subjid colin27 5. tkregister2 --mgz --s colin27 --fstal Results are not perfect but reasonable ie. about the same as many other runs I've had that have passed so I don't think it is the registration. 6. tkmedit colin27 brainmask.mgz -aux T1.mgz It is clear this is where the problem is. http://opus.bic.mni.mcgill.ca/public/brainmask.mgz http://opus.bic.mni.mcgill.ca/public/T1.mgz 7. I have subsequently tried: recon-all -skullstrip -wsthresh 10 -clean-bm -no-wsgcaatlas -subjid colin27 (basically extract the entire head) recon-all -skullstrip -wsthresh 5 -clean-bm -no-wsgcaatlas -subjid colin27 (basically extracted the entire head) recon-all -skullstrip -wsthresh 5 -clean-bm -subjid colin27 (extracted the brain, but left large chunks of skull) recon-all -skullstrip -wsthresh 2 -clean-bm -subjid colin27 (extracted the brain, but left quite a few chunks of skull)

I'm concerned that the -wsthresh is so low in order to get a reasonable result, that's why I think the intensities might be off (see step 3).

I'm re-running -autorecon2/3 now to see if the surfaces are better, but if someone could take a look at my input file and the converted 001.mgz to see if it isn't a problem there it would be much appreciated.

thanks, mishkin _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer