Re: [HCP-Users] Distorted MNI space result (T?w_restore) -- maybe due to "bad" initial brain mask?
Note that the scenes we’ve assembled for “StructuralQC” include a check of the FNIRT registration. You can find the scripts necessary to generate those scenes here: https://github.com/Washington-University/StructuralQC cheers, -MH -- Michael Harms, Ph.D. --- Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave.Tel: 314-747-6173 St. Louis, MO 63110Email: mha...@wustl.edu On 12/1/16, 1:04 PM, "hcp-users-boun...@humanconnectome.org on behalf of Gaurav Patel" wrote: We've been noticing similar masking issues with our non-HCP data acquired on a GE MR750, and have found that fat sat in our sequence was turned off, so we will attempt to do the same and report whether that changes things. __ gaurav patel gauravpa...@gmail.com www.neurofreak.net On Dec 1, 2016, at 2:01 PM, Julien Dubois wrote: > It looks like the Freesurfer pipeline does not make use of the MNI space >registration from the last Prefreesurfer step. > Hence, to correct for the suboptimal MNI registration due to the fat sat >discrepancy, my plan of action is to: > 1) generate a "no fat sat" T1w template > 2) re-run AtlasRegistrationToMNI152_FLIRTandFNIRT.sh (last step of the >PreFreesurfer pipeline) > 3) re-run the PostFreesurfer pipeline > Does that sound reasonable? > > If 2) doesn't work well, I'll look into using the T2w image for MNI >space registration instead. > > - Julien > > > Julien Dubois, PhD > Postdoctoral Scientist > Department of Neurosurgery, Cedars-Sinai Medical Center, Los Angeles > Department of the Humanities and Social Sciences, California Institute >of Technology, Pasadena > Tel: +1 (310)423-8377 > Web: http://nigiri.caltech.edu/~jdubois > > > On Thu, Dec 1, 2016 at 10:24 AM, Glasser, Matthew >wrote: > If you have a decent number of subjects that go through the pipelines >okay, you can just generate another average template after MNI >registration. Else, there isn’t any particular reason why a T1w-based >MNI registration is better than a T2w-based registration. We probably >should make a flag that automatically allows you to base the brain mask >and atlas registration on the T2w image. > > Peace, > > Matt. > > From: Julien Dubois > Date: Thursday, December 1, 2016 at 11:49 AM > > To: Matt Glasser > Cc: "hcp-users@humanconnectome.org" > Subject: Re: [HCP-Users] Distorted MNI space result (T?w_restore) -- >maybe due to "bad" initial brain mask? > > Thank you Matt. > > I actually don't think the brain extraction is the culprit, after >spending more time with the intermediate outputs. Indeed, >T1w/T1w_acpc_dc_restore_brain.nii.gz actually looks quite good. > > The issue mostly arises at the fnirt step in >AtlasRegistrationToMNI152_FLIRTandFNIRT.sh. And indeed, it is clear that >the template from the HCP pipelines (MNI152_T1_2mm.nii.gz) is very dark >between the brain and the skull, while my input image >(T1w/T1w_acpc_dc_restore.nii.gz) is not. The best solution will thus >consist in using a template with fat sat off. > > Any idea where I might find such a template? I'll look around, but if >you have suggestions, let me know. > > - Julien > > Julien Dubois, PhD > Postdoctoral Scientist > Department of Neurosurgery, Cedars-Sinai Medical Center, Los Angeles > Department of the Humanities and Social Sciences, California Institute >of Technology, Pasadena > Tel: +1 (310)423-8377 > Web: http://nigiri.caltech.edu/~jdubois > > > On Wed, Nov 30, 2016 at 5:41 PM, Glasser, Matthew >wrote: > If you had a volume template in MNI space that had fat sat off, it might >register better. Alternatively you might modify the pipelines to use the >brain mask from the T2w scan and to do the atlas registration from the >T2w scan. > > Peace, > > Matt. > > From: Julien Dubois > Date: Wednesday, November 30, 2016 at 5:51 PM > To: Matt Glasser > Cc: "hcp-users@humanconnectome.org" > Subject: Re: [HCP-Users] Distorted MNI space result (T?w_restore) -- >maybe due to "bad" initial brain mask? > > I am not sure -- I am checking with the people who acquired that data. > I have had similar issues before with in-house data which does not have >fat sat turned on. > Do you recommend a tweak to the PreFreeSurfer pipeline if fat sat is not >turned on? > > I now see that you recommmend fat sat in the HCP T1 protocol [“Fat >suppr. = Water excit. Fast”] to reduce signal from bone marrow and scalp >fat > > - Julien > > Julien Dubois, PhD > Postdoctoral Scientist > Depar
Re: [HCP-Users] Distorted MNI space result (T?w_restore) -- maybe due to "bad" initial brain mask?
We've been noticing similar masking issues with our non-HCP data acquired on a GE MR750, and have found that fat sat in our sequence was turned off, so we will attempt to do the same and report whether that changes things. __ gaurav patel gauravpa...@gmail.com www.neurofreak.net On Dec 1, 2016, at 2:01 PM, Julien Dubois wrote: > It looks like the Freesurfer pipeline does not make use of the MNI space > registration from the last Prefreesurfer step. > Hence, to correct for the suboptimal MNI registration due to the fat sat > discrepancy, my plan of action is to: > 1) generate a "no fat sat" T1w template > 2) re-run AtlasRegistrationToMNI152_FLIRTandFNIRT.sh (last step of the > PreFreesurfer pipeline) > 3) re-run the PostFreesurfer pipeline > Does that sound reasonable? > > If 2) doesn't work well, I'll look into using the T2w image for MNI space > registration instead. > > - Julien > > > Julien Dubois, PhD > Postdoctoral Scientist > Department of Neurosurgery, Cedars-Sinai Medical Center, Los Angeles > Department of the Humanities and Social Sciences, California Institute of > Technology, Pasadena > Tel: +1 (310)423-8377 > Web: http://nigiri.caltech.edu/~jdubois > > > On Thu, Dec 1, 2016 at 10:24 AM, Glasser, Matthew wrote: > If you have a decent number of subjects that go through the pipelines okay, > you can just generate another average template after MNI registration. Else, > there isn’t any particular reason why a T1w-based MNI registration is better > than a T2w-based registration. We probably should make a flag that > automatically allows you to base the brain mask and atlas registration on the > T2w image. > > Peace, > > Matt. > > From: Julien Dubois > Date: Thursday, December 1, 2016 at 11:49 AM > > To: Matt Glasser > Cc: "hcp-users@humanconnectome.org" > Subject: Re: [HCP-Users] Distorted MNI space result (T?w_restore) -- maybe > due to "bad" initial brain mask? > > Thank you Matt. > > I actually don't think the brain extraction is the culprit, after spending > more time with the intermediate outputs. Indeed, > T1w/T1w_acpc_dc_restore_brain.nii.gz actually looks quite good. > > The issue mostly arises at the fnirt step in > AtlasRegistrationToMNI152_FLIRTandFNIRT.sh. And indeed, it is clear that the > template from the HCP pipelines (MNI152_T1_2mm.nii.gz) is very dark between > the brain and the skull, while my input image > (T1w/T1w_acpc_dc_restore.nii.gz) is not. The best solution will thus consist > in using a template with fat sat off. > > Any idea where I might find such a template? I'll look around, but if you > have suggestions, let me know. > > - Julien > > Julien Dubois, PhD > Postdoctoral Scientist > Department of Neurosurgery, Cedars-Sinai Medical Center, Los Angeles > Department of the Humanities and Social Sciences, California Institute of > Technology, Pasadena > Tel: +1 (310)423-8377 > Web: http://nigiri.caltech.edu/~jdubois > > > On Wed, Nov 30, 2016 at 5:41 PM, Glasser, Matthew wrote: > If you had a volume template in MNI space that had fat sat off, it might > register better. Alternatively you might modify the pipelines to use the > brain mask from the T2w scan and to do the atlas registration from the T2w > scan. > > Peace, > > Matt. > > From: Julien Dubois > Date: Wednesday, November 30, 2016 at 5:51 PM > To: Matt Glasser > Cc: "hcp-users@humanconnectome.org" > Subject: Re: [HCP-Users] Distorted MNI space result (T?w_restore) -- maybe > due to "bad" initial brain mask? > > I am not sure -- I am checking with the people who acquired that data. > I have had similar issues before with in-house data which does not have fat > sat turned on. > Do you recommend a tweak to the PreFreeSurfer pipeline if fat sat is not > turned on? > > I now see that you recommmend fat sat in the HCP T1 protocol [“Fat suppr. = > Water excit. Fast”] to reduce signal from bone marrow and scalp fat > > - Julien > > Julien Dubois, PhD > Postdoctoral Scientist > Department of Neurosurgery, Cedars-Sinai Medical Center, Los Angeles > Department of the Humanities and Social Sciences, California Institute of > Technology, Pasadena > Tel: +1 (310)423-8377 > Web: http://nigiri.caltech.edu/~jdubois > > > On Wed, Nov 30, 2016 at 2:18 PM, Glasser, Matthew wrote: > Out of curiosity, does this data not have fat sat on the T1w scan? > > Peace, > > Matt. > > From: on behalf of Julien Dubois > > Date: Wednesday, November 30, 2016 at 2:51 PM > To: "hcp-users@human
