Re: [HCP-Users] cifti-correlation-gradient settings and surface-based math?
Hi Alex, For group data I usually use a gradient pre-smoothing of 1mm sigma, though for some modalities like myelin maps, even this reduces some detail (you might try both ways…). For seed exclusion I use 2mm (the acquisition resolution), basically the goal is to exclude the central spike in correlation. Recently we’ve found that the fisher z transform distorts gradients some (relative to other modalities) and global signal regression/MGTR distorts them more (basically moving gradients towards cortical areas that are more correlated with the global sensory systems signal). This is one of the reasons we’ve stopped doing global signal regression/MGTR, though there remains debate on this issue in some quarters of the HCP. To avoid using things like global signal regression, use of a denoising approach like ICAFIX is absolutely critical (and is important anyway, as past approaches to denoising did not address spatially specific temporal artifacts in a data driven way, possibly leading to things like distance dependent artifacts that are possibly then exacerbated by using global signal regression). There is also a difference of opinion on edge detection and we don’t use it (preferring to stay closer to the original data and avoid thresholding), though others do. We have not implemented edge detection in Connectome Workbench, but I believe someone has code that uses wb_command outputs and computes this in matlab. This is how you visualize CIFTI dense connectomes .dconn.nii, and so this is a core feature of Connectome Workbench. Note that one uses -cifti-correlation to compute the dense connectome and then -cifti-correlation-gradient to compute the gradients. Peace, Matt. From: Alex Cohen alexco...@gmail.com Reply-To: alexco...@gmail.com alexco...@gmail.com Date: Wednesday, April 22, 2015 at 2:19 PM To: hcp-users@humanconnectome.org hcp-users@humanconnectome.org Subject: [HCP-Users] cifti-correlation-gradient settings and surface-based math? Hi all, Two quick (ok three) fcMRI questions: - What settings have you all been using for calculating surface-based gradients with resting BOLD data for surface presmoothing, seed exclusion and the like? - Is there a a list of methods using wb_command for doing surface-based image processing beyond simple gradient (e.g., like edge detection, etc...) I know there is extrema detection, but just wondering what else is available... and/or how to implement things like this... - As a side-note, back in the Caret days there was a function where you could click on the surface and the associated correlation map would appear on the surface (assuming it was already made)... Did this make it into wb_view? and/or was adapted for CIFTI files containing correlation maps? -Alex Alexander Li Cohen, M.D., Ph.D. Resident Physician Child and Adolescent Neurology E-mail: cohen.alexan...@mayo.edu (Medical/Science Email) E-mail:alexco...@gmail.com (Lifetime Email) Mayo Clinic 200 First Street SW Rochester, MN 55905 mayoclinic.org ___ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail. ___HCP-Users mailing listHCP-Users@humanconnectome.orghttp://lists.humanconnectome.org/mailman/listinfo/hcp-users
Re: [HCP-Users] cifti-correlation-gradient settings and surface-based math?
Ah, I didn't realize that you were referring to that in the 3rd paragraph. thanks. -alex *Alexander Li Cohen, M.D., Ph.D.* Resident Physician Child and Adolescent Neurology E-mail: cohen.alexan...@mayo.edu (Medical/Science Email) E-mail: alexco...@gmail.com (Lifetime Email) *Mayo Clinic* 200 First Street SW Rochester, MN 55905 mayoclinic.org On Wed, Apr 22, 2015 at 3:23 PM, Glasser, Matthew glass...@wusm.wustl.edu wrote: -cifti-correlation-gradient does a second order correlation (where you had used eta^2), taking a .dconn.nii as its input. The first order correlation is done with -cifti-correlation from a .dtseries.nii to a .dconn.nii. I already explained about click view maps for .dconn.nii. ;) Peace, Matt. From: Alex Cohen alexco...@gmail.com Reply-To: alexco...@gmail.com alexco...@gmail.com Date: Wednesday, April 22, 2015 at 3:19 PM To: Matt Glasser glass...@wusm.wustl.edu Cc: hcp-users@humanconnectome.org hcp-users@humanconnectome.org Subject: Re: [HCP-Users] cifti-correlation-gradient settings and surface-based math? Interesting, thanks for the numbers. I am using FEAT and ICA+FIX as my preprocessing as well, so that should be less of an issue. Of note, my impression from the wb_command help was that correlation-gradient was doing BOTH correlation and gradient-average... wrapped up into one tool. is this not correct and one needs to run cifti-correlation prior??? (also, any idea regarding the click-view of correlation maps? Donna?) -Alex *Alexander Li Cohen, M.D., Ph.D.* Resident Physician Child and Adolescent Neurology E-mail: cohen.alexan...@mayo.edu (Medical/Science Email) E-mail: alexco...@gmail.com (Lifetime Email) *Mayo Clinic* 200 First Street SW Rochester, MN 55905 mayoclinic.org On Wed, Apr 22, 2015 at 2:45 PM, Glasser, Matthew glass...@wusm.wustl.edu wrote: Hi Alex, For group data I usually use a gradient pre-smoothing of 1mm sigma, though for some modalities like myelin maps, even this reduces some detail (you might try both ways…). For seed exclusion I use 2mm (the acquisition resolution), basically the goal is to exclude the central spike in correlation. Recently we’ve found that the fisher z transform distorts gradients some (relative to other modalities) and global signal regression/MGTR distorts them more (basically moving gradients towards cortical areas that are more correlated with the global sensory systems signal). This is one of the reasons we’ve stopped doing global signal regression/MGTR, though there remains debate on this issue in some quarters of the HCP. To avoid using things like global signal regression, use of a denoising approach like ICA+FIX is absolutely critical (and is important anyway, as past approaches to denoising did not address spatially specific temporal artifacts in a data driven way, possibly leading to things like distance dependent artifacts that are possibly then exacerbated by using global signal regression). There is also a difference of opinion on edge detection and we don’t use it (preferring to stay closer to the original data and avoid thresholding), though others do. We have not implemented edge detection in Connectome Workbench, but I believe someone has code that uses wb_command outputs and computes this in matlab. This is how you visualize CIFTI dense connectomes .dconn.nii, and so this is a core feature of Connectome Workbench. Note that one uses -cifti-correlation to compute the dense connectome and then -cifti-correlation-gradient to compute the gradients. Peace, Matt. From: Alex Cohen alexco...@gmail.com Reply-To: alexco...@gmail.com alexco...@gmail.com Date: Wednesday, April 22, 2015 at 2:19 PM To: hcp-users@humanconnectome.org hcp-users@humanconnectome.org Subject: [HCP-Users] cifti-correlation-gradient settings and surface-based math? Hi all, Two quick (ok three) fcMRI questions: - What settings have you all been using for calculating surface-based gradients with resting BOLD data for surface presmoothing, seed exclusion and the like? - Is there a a list of methods using wb_command for doing surface-based image processing beyond simple gradient (e.g., like edge detection, etc...) I know there is extrema detection, but just wondering what else is available... and/or how to implement things like this... - As a side-note, back in the Caret days there was a function where you could click on the surface and the associated correlation map would appear on the surface (assuming it was already made)... Did this make it into wb_view? and/or was adapted for CIFTI files containing correlation maps? -Alex
Re: [HCP-Users] cifti-correlation-gradient settings and surface-based math?
Interesting, thanks for the numbers. I am using FEAT and ICA+FIX as my preprocessing as well, so that should be less of an issue. Of note, my impression from the wb_command help was that correlation-gradient was doing BOTH correlation and gradient-average... wrapped up into one tool. is this not correct and one needs to run cifti-correlation prior??? (also, any idea regarding the click-view of correlation maps? Donna?) -Alex *Alexander Li Cohen, M.D., Ph.D.* Resident Physician Child and Adolescent Neurology E-mail: cohen.alexan...@mayo.edu (Medical/Science Email) E-mail: alexco...@gmail.com (Lifetime Email) *Mayo Clinic* 200 First Street SW Rochester, MN 55905 mayoclinic.org On Wed, Apr 22, 2015 at 2:45 PM, Glasser, Matthew glass...@wusm.wustl.edu wrote: Hi Alex, For group data I usually use a gradient pre-smoothing of 1mm sigma, though for some modalities like myelin maps, even this reduces some detail (you might try both ways…). For seed exclusion I use 2mm (the acquisition resolution), basically the goal is to exclude the central spike in correlation. Recently we’ve found that the fisher z transform distorts gradients some (relative to other modalities) and global signal regression/MGTR distorts them more (basically moving gradients towards cortical areas that are more correlated with the global sensory systems signal). This is one of the reasons we’ve stopped doing global signal regression/MGTR, though there remains debate on this issue in some quarters of the HCP. To avoid using things like global signal regression, use of a denoising approach like ICA+FIX is absolutely critical (and is important anyway, as past approaches to denoising did not address spatially specific temporal artifacts in a data driven way, possibly leading to things like distance dependent artifacts that are possibly then exacerbated by using global signal regression). There is also a difference of opinion on edge detection and we don’t use it (preferring to stay closer to the original data and avoid thresholding), though others do. We have not implemented edge detection in Connectome Workbench, but I believe someone has code that uses wb_command outputs and computes this in matlab. This is how you visualize CIFTI dense connectomes .dconn.nii, and so this is a core feature of Connectome Workbench. Note that one uses -cifti-correlation to compute the dense connectome and then -cifti-correlation-gradient to compute the gradients. Peace, Matt. From: Alex Cohen alexco...@gmail.com Reply-To: alexco...@gmail.com alexco...@gmail.com Date: Wednesday, April 22, 2015 at 2:19 PM To: hcp-users@humanconnectome.org hcp-users@humanconnectome.org Subject: [HCP-Users] cifti-correlation-gradient settings and surface-based math? Hi all, Two quick (ok three) fcMRI questions: - What settings have you all been using for calculating surface-based gradients with resting BOLD data for surface presmoothing, seed exclusion and the like? - Is there a a list of methods using wb_command for doing surface-based image processing beyond simple gradient (e.g., like edge detection, etc...) I know there is extrema detection, but just wondering what else is available... and/or how to implement things like this... - As a side-note, back in the Caret days there was a function where you could click on the surface and the associated correlation map would appear on the surface (assuming it was already made)... Did this make it into wb_view? and/or was adapted for CIFTI files containing correlation maps? -Alex *Alexander Li Cohen, M.D., Ph.D.* Resident Physician Child and Adolescent Neurology E-mail: cohen.alexan...@mayo.edu (Medical/Science Email) E-mail: alexco...@gmail.com (Lifetime Email) *Mayo Clinic* 200 First Street SW Rochester, MN 55905 mayoclinic.org ___ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users -- The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail. ___ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users
Re: [HCP-Users] cifti-correlation-gradient settings and surface-based math?
-cifti-correlation-gradient does a second order correlation (where you had used eta^2), taking a .dconn.nii as its input. The first order correlation is done with -cifti-correlation from a .dtseries.nii to a .dconn.nii. I already explained about click view maps for .dconn.nii. ;) Peace, Matt. From: Alex Cohen alexco...@gmail.com Reply-To: alexco...@gmail.com alexco...@gmail.com Date: Wednesday, April 22, 2015 at 3:19 PM To: Matt Glasser glass...@wusm.wustl.edu Cc: hcp-users@humanconnectome.org hcp-users@humanconnectome.org Subject: Re: [HCP-Users] cifti-correlation-gradient settings and surface-based math? Interesting, thanks for the numbers. I am using FEAT and ICAFIX as my preprocessing as well, so that should be less of an issue. Of note, my impression from the wb_command help was that correlation-gradient was doing BOTH correlation and gradient-average... wrapped up into one tool. is this not correct and one needs to run cifti-correlation prior??? (also, any idea regarding the click-view of correlation maps? Donna?) -Alex Alexander Li Cohen, M.D., Ph.D. Resident Physician Child and Adolescent Neurology E-mail: cohen.alexan...@mayo.edu (Medical/Science Email) E-mail:alexco...@gmail.com (Lifetime Email) Mayo Clinic 200 First Street SW Rochester, MN 55905 mayoclinic.org On Wed, Apr 22, 2015 at 2:45 PM, Glasser, Matthew glass...@wusm.wustl.edu wrote: Hi Alex, For group data I usually use a gradient pre-smoothing of 1mm sigma, though for some modalities like myelin maps, even this reduces some detail (you might try both ways…). For seed exclusion I use 2mm (the acquisition resolution), basically the goal is to exclude the central spike in correlation. Recently we’ve found that the fisher z transform distorts gradients some (relative to other modalities) and global signal regression/MGTR distorts them more (basically moving gradients towards cortical areas that are more correlated with the global sensory systems signal). This is one of the reasons we’ve stopped doing global signal regression/MGTR, though there remains debate on this issue in some quarters of the HCP. To avoid using things like global signal regression, use of a denoising approach like ICAFIX is absolutely critical (and is important anyway, as past approaches to denoising did not address spatially specific temporal artifacts in a data driven way, possibly leading to things like distance dependent artifacts that are possibly then exacerbated by using global signal regression). There is also a difference of opinion on edge detection and we don’t use it (preferring to stay closer to the original data and avoid thresholding), though others do. We have not implemented edge detection in Connectome Workbench, but I believe someone has code that uses wb_command outputs and computes this in matlab. This is how you visualize CIFTI dense connectomes .dconn.nii, and so this is a core feature of Connectome Workbench. Note that one uses -cifti-correlation to compute the dense connectome and then -cifti-correlation-gradient to compute the gradients. Peace, Matt. From: Alex Cohen alexco...@gmail.com Reply-To: alexco...@gmail.com alexco...@gmail.com Date: Wednesday, April 22, 2015 at 2:19 PM To: hcp-users@humanconnectome.org hcp-users@humanconnectome.org Subject: [HCP-Users] cifti-correlation-gradient settings and surface-based math? Hi all, Two quick (ok three) fcMRI questions: - What settings have you all been using for calculating surface-based gradients with resting BOLD data for surface presmoothing, seed exclusion and the like? - Is there a a list of methods using wb_command for doing surface-based image processing beyond simple gradient (e.g., like edge detection, etc...) I know there is extrema detection, but just wondering what else is available... and/or how to implement things like this... - As a side-note, back in the Caret days there was a function where you could click on the surface and the associated correlation map would appear on the surface (assuming it was already made)... Did this make it into wb_view? and/or was adapted for CIFTI files containing correlation maps? -Alex Alexander Li Cohen, M.D., Ph.D. Resident Physician Child and Adolescent Neurology E-mail: cohen.alexan...@mayo.edu (Medical/Science Email) E-mail:alexco...@gmail.com (Lifetime Email) Mayo Clinic 200 First Street SW Rochester, MN 55905 mayoclinic.org ___ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users The materials in this message are private and may contain Protected Healthcare Information or other
