Autorecon2 - Revision history

Chris: /* White Matter Voxels Missing From wm.mgz */

2018-02-09T01:50:54Z

White Matter Voxels Missing From wm.mgz

← Older revision		Revision as of 21:50, 8 February 2018
Line 94:		Line 94:
	There is no prescribed formula for this, but the information from the previous step can be used to guide what sort of operations you might want to do in 3dcalc to generate a fixed wm.mgz or enhanced T1.mgz file. Below are a few plausible scenarios:		There is no prescribed formula for this, but the information from the previous step can be used to guide what sort of operations you might want to do in 3dcalc to generate a fixed wm.mgz or enhanced T1.mgz file. Below are a few plausible scenarios:
	====White Matter Voxels Missing From wm.mgz====		====White Matter Voxels Missing From wm.mgz====
	If the SPM WM segmentation looks reasonable, you might consider simply using that file in place of wm.mgz. I will note that SPM includes the cerebellum in its WM segmentation, whereas wm.mgz does not. I'm not certain if it's important to remove it, but you might simply convert a rescaled version of c2001.nii into filled.mgz. The voxels in c2001.nii range from 0.0 to 1.00, and are thus SPM's probabalistic determination that voxel v contains white matter, whereas filled.mgz contains values ranging from 0 to 255, so we rescale accordingly. This file can be converted to a .mgz file and replace the one generated by FreeSurfer:		If the SPM WM segmentation looks reasonable, you might consider simply using that file in place of wm.mgz and/or filled.mgz (wm.mgz is used to generate filled.mgz). I will note that SPM includes the cerebellum in its WM segmentation, whereas wm.mgz does not. I'm not certain if it's important to remove it, but you might simply convert a rescaled version of c2001.nii into filled.mgz. The voxels in c2001.nii range from 0.0 to 1.00, and are thus SPM's probabalistic determination that voxel v contains white matter, whereas filled.mgz contains values ranging from 0 to 255, so we rescale accordingly. This file can be converted to a .mgz file and replace the one generated by FreeSurfer:
	3dcalc -a c2001.nii -expr 'a*255' -prefix rescaled_spm_wm.nii		3dcalc -a c2001.nii -expr 'a*255' -prefix rescaled_spm_wm.nii
	mri_convert rescaled_spm_wm.nii filled.mgz		mri_convert rescaled_spm_wm.nii filled.mgz

Chris: /* Retesselation */

2018-02-09T01:42:27Z

Retesselation

← Older revision		Revision as of 21:42, 8 February 2018
Line 137:		Line 137:
	And then a few more steps all at once:		And then a few more steps all at once:
	recon-all -s $SUBJID -fix -white -smooth2 -inflate2		recon-all -s $SUBJID -fix -white -smooth2 -inflate2
	(~~awhile~~)		(~4.5 hours)

128.205.173.202: /* The Mysterious Case of the Missing Chunks */

2018-02-08T20:30:41Z

The Mysterious Case of the Missing Chunks

← Older revision		Revision as of 16:30, 8 February 2018
Line 127:		Line 127:
	The next step would be to copy any new enhanced mgz files back to the FreeSurfer subject directory and restart the tesselation process		The next step would be to copy any new enhanced mgz files back to the FreeSurfer subject directory and restart the tesselation process

			===Retesselation===
			After you have regenerated filled.mgz and/or norm.mgz, you can see if these modified files will tesselate any better. Recon-all can be started at any point in the process, so you will be picking up at the -tesselate step:
			$SUBJID=FS_0238
			recon-all -s $SUBJID -tesselate
			(<2 minutes)
			It then looks like you can do the next few steps in one fell swoop:
			recon-all -s $SUBJID -smooth1 -inflate1 -qsphere
			(~10 minutes)
			And then a few more steps all at once:
			recon-all -s $SUBJID -fix -white -smooth2 -inflate2
			(awhile)

128.205.173.202: /* Gray Matter Too Noisy/Dark */

2018-02-08T20:08:10Z

Gray Matter Too Noisy/Dark

@@ Line 100: / Line 100: @@
 ====Gray Matter Too Noisy/Dark====
 In a problem volume I'm looking at now, the FreeSurfer WM segmentation seemed to get *most* of what SPM found, so that may not entirely explain the missing anterior temporal tissue. What else might be the problem? I noted that the brainmask.mgz volume is pretty noisy in this area, so perhaps the apparently random distribution of higher/lower values within (what should be) the same tissue is causing a problem. Perhaps making the GM tissue a bit more homogeneous would help. I might start by selecting GM voxels below a certain value and increasing them to a threshold minimum (perhaps the average value of the rest of the GM volume). According to the [[https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllDevTable|Recon-All wiki]] the tesselation step (the part that turns voxels into triangles) takes filled.mgz and norm.mgz as input volumes. This tells us that we can try using the SPM gray matter mask to fix norm.mgz and then re-tesselate.
 [[Category: FreeSurfer]]

128.205.173.202: /* Gray Matter Too Noisy/Dark */

2018-02-08T19:15:50Z

Gray Matter Too Noisy/Dark

← Older revision		Revision as of 15:15, 8 February 2018
Line 99:		Line 99:

	====Gray Matter Too Noisy/Dark====		====Gray Matter Too Noisy/Dark====
	In a problem volume I'm looking at now, the FreeSurfer WM segmentation seemed to get most of what SPM found, so that may not entirely explain the missing anterior temporal tissue. What else might be the problem? I noted that the brainmask.mgz volume is pretty noisy in this area, so perhaps the apparently random distribution of higher/lower values within (what should be) the same tissue is causing a problem. Perhaps making the GM tissue a bit more homogeneous would help. I might start by selecting GM voxels below a certain value and increasing them to a threshold minimum (perhaps the average value of the rest of the GM volume). According to the [[https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllDevTable\|Recon-All wiki]] the tesselation step (the part that turns voxels into triangles) takes filled.mgz and norm.mgz as input volumes. ~~(darnit, I think we needed to fix filled.mgz in the above step, rather than wm.mgz)~~		In a problem volume I'm looking at now, the FreeSurfer WM segmentation seemed to get most of what SPM found, so that may not entirely explain the missing anterior temporal tissue. What else might be the problem? I noted that the brainmask.mgz volume is pretty noisy in this area, so perhaps the apparently random distribution of higher/lower values within (what should be) the same tissue is causing a problem. Perhaps making the GM tissue a bit more homogeneous would help. I might start by selecting GM voxels below a certain value and increasing them to a threshold minimum (perhaps the average value of the rest of the GM volume). According to the [[https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllDevTable\|Recon-All wiki]] the tesselation step (the part that turns voxels into triangles) takes filled.mgz and norm.mgz as input volumes. This tells us that we can try using the SPM gray matter mask to fix norm.mgz and then re-tesselate.
	This tells us that we can try using the SPM gray matter mask to fix norm.mgz and then re-tesselate.


	[[Category: FreeSurfer]]		[[Category: FreeSurfer]]

128.205.173.202: /* White Matter Voxels Missing From wm.mgz */

2018-02-08T19:14:42Z

White Matter Voxels Missing From wm.mgz

← Older revision		Revision as of 15:14, 8 February 2018
Line 94:		Line 94:
	There is no prescribed formula for this, but the information from the previous step can be used to guide what sort of operations you might want to do in 3dcalc to generate a fixed wm.mgz or enhanced T1.mgz file. Below are a few plausible scenarios:		There is no prescribed formula for this, but the information from the previous step can be used to guide what sort of operations you might want to do in 3dcalc to generate a fixed wm.mgz or enhanced T1.mgz file. Below are a few plausible scenarios:
	====White Matter Voxels Missing From wm.mgz====		====White Matter Voxels Missing From wm.mgz====
	If the SPM WM segmentation looks reasonable, you might consider simply using that file in place of wm.mgz. I will note that SPM includes the cerebellum in its WM segmentation, whereas wm.mgz does not. I'm not certain if it's important to remove it, but you might simply convert a rescaled version of c2001.nii into wm.mgz. ~~FreeSurfer expects the mean WM voxel value to be 110, however the~~ voxels in c2001.nii range from 0.0 to 1.00, and are thus SPM's probabalistic determination that voxel v contains white matter. ~~We can use 0.5 as a threshold using astep to set v<0.5 to 0.0 and v>~~0.5 to ~~1.0, then multiply those values by 110. The resulting volume will have 110 in all wm voxels~~, ~~and 0.0 in the remainder~~. This file can be converted to a .mgz file and replace the one generated by FreeSurfer:		If the SPM WM segmentation looks reasonable, you might consider simply using that file in place of wm.mgz. I will note that SPM includes the cerebellum in its WM segmentation, whereas wm.mgz does not. I'm not certain if it's important to remove it, but you might simply convert a rescaled version of c2001.nii into filled.mgz. The voxels in c2001.nii range from 0.0 to 1.00, and are thus SPM's probabalistic determination that voxel v contains white matter, whereas filled.mgz contains values ranging from 0 to 255, so we rescale accordingly. This file can be converted to a .mgz file and replace the one generated by FreeSurfer:
	3dcalc -a c2001.nii -expr '~~astep(~~a~~,0.5)~~*~~110~~' -prefix rescaled_spm_wm.nii		3dcalc -a c2001.nii -expr 'a*255' -prefix rescaled_spm_wm.nii
	mri_convert rescaled_spm_wm.nii wm.mgz		mri_convert rescaled_spm_wm.nii filled.mgz

	====Gray Matter Too Noisy/Dark====		====Gray Matter Too Noisy/Dark====
	In a problem volume I'm looking at now, the FreeSurfer WM segmentation seemed to get most of what SPM found, so that may not entirely explain the missing anterior temporal tissue. What else might be the problem? I noted that the brainmask.mgz volume is pretty noisy in this area, so perhaps the apparently random distribution of higher/lower values within (what should be) the same tissue is causing a problem. Perhaps making the GM tissue a bit more homogeneous would help. I might start by selecting GM voxels below a certain value and increasing them to a threshold minimum (perhaps the average value of the rest of the GM volume). According to the [[https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllDevTable\|Recon-All wiki]] the tesselation step (the part that turns voxels into triangles) takes filled.mgz and norm.mgz as input volumes. (darnit, I think we needed to fix filled.mgz in the above step, rather than wm.mgz)		In a problem volume I'm looking at now, the FreeSurfer WM segmentation seemed to get most of what SPM found, so that may not entirely explain the missing anterior temporal tissue. What else might be the problem? I noted that the brainmask.mgz volume is pretty noisy in this area, so perhaps the apparently random distribution of higher/lower values within (what should be) the same tissue is causing a problem. Perhaps making the GM tissue a bit more homogeneous would help. I might start by selecting GM voxels below a certain value and increasing them to a threshold minimum (perhaps the average value of the rest of the GM volume). According to the [[https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllDevTable\|Recon-All wiki]] the tesselation step (the part that turns voxels into triangles) takes filled.mgz and norm.mgz as input volumes. (darnit, I think we needed to fix filled.mgz in the above step, rather than wm.mgz)

128.205.173.202: /* =White Matter Voxels Missing From wm.mgz */

2018-02-08T19:08:42Z

=White Matter Voxels Missing From wm.mgz

← Older revision		Revision as of 15:08, 8 February 2018
Line 93:		Line 93:
	===3dcalc===		===3dcalc===
	There is no prescribed formula for this, but the information from the previous step can be used to guide what sort of operations you might want to do in 3dcalc to generate a fixed wm.mgz or enhanced T1.mgz file. Below are a few plausible scenarios:		There is no prescribed formula for this, but the information from the previous step can be used to guide what sort of operations you might want to do in 3dcalc to generate a fixed wm.mgz or enhanced T1.mgz file. Below are a few plausible scenarios:
	====White Matter Voxels Missing From wm.mgz===		====White Matter Voxels Missing From wm.mgz====
	If the SPM WM segmentation looks reasonable, you might consider simply using that file in place of wm.mgz. I will note that SPM includes the cerebellum in its WM segmentation, whereas wm.mgz does not. I'm not certain if it's important to remove it, but you might simply convert a rescaled version of c2001.nii into wm.mgz. FreeSurfer expects the mean WM voxel value to be 110, however the voxels in c2001.nii range from 0.0 to 1.00, and are thus SPM's probabalistic determination that voxel v contains white matter. We can use 0.5 as a threshold using astep to set v<0.5 to 0.0 and v>0.5 to 1.0, then multiply those values by 110. The resulting volume will have 110 in all wm voxels, and 0.0 in the remainder. This file can be converted to a .mgz file and replace the one generated by FreeSurfer:		If the SPM WM segmentation looks reasonable, you might consider simply using that file in place of wm.mgz. I will note that SPM includes the cerebellum in its WM segmentation, whereas wm.mgz does not. I'm not certain if it's important to remove it, but you might simply convert a rescaled version of c2001.nii into wm.mgz. FreeSurfer expects the mean WM voxel value to be 110, however the voxels in c2001.nii range from 0.0 to 1.00, and are thus SPM's probabalistic determination that voxel v contains white matter. We can use 0.5 as a threshold using astep to set v<0.5 to 0.0 and v>0.5 to 1.0, then multiply those values by 110. The resulting volume will have 110 in all wm voxels, and 0.0 in the remainder. This file can be converted to a .mgz file and replace the one generated by FreeSurfer:
	3dcalc -a c2001.nii -expr 'astep(a,0.5)*110' -prefix rescaled_spm_wm.nii		3dcalc -a c2001.nii -expr 'astep(a,0.5)*110' -prefix rescaled_spm_wm.nii

128.205.173.202: /* The Mysterious Case of the Missing Chunks */

2018-02-08T19:08:17Z

The Mysterious Case of the Missing Chunks

← Older revision		Revision as of 15:08, 8 February 2018
Line 83:		Line 83:
	At the end of this process, you will have a series of .nii files, prefixed with C1 to C5. We will be recombining these files and using MRICron to modify the WM.mgz file output by recon-all during the autorecon2 step.		At the end of this process, you will have a series of .nii files, prefixed with C1 to C5. We will be recombining these files and using MRICron to modify the WM.mgz file output by recon-all during the autorecon2 step.

	'''~~WORK IN PROGRESS~~'''		===Copy Files to Fix Directory===
			The next bunch of work will take place in the fix directory that contains all your segmented files. We will need the wm.mgz generated by the autorecon2 step. You can find that in $SUBJECTS_DIR/$SUBJID/mri/wm.mgz. We can use mri_convert to copy it over as a .nii, which is the format we need the file in anyways:
			SUBJID=FS_0238
			FIXDIR=~/Fix/0238/mri
			mri_convert ${SUBJECTS_DIR}/${SUBJID}/mri/wm.mgz ${FIXDIR}/wm.nii
			mri_convert ${SUBJECTS_DIR}/${SUBJID}/mri/brainmask.mgz ${FIXDIR}/brainmask.nii
			===Explore Masks in MRICron===
			At this point, we can load the brainmask.nii into MRICron, and then load the various segmentation masks as overlays to see whether the FreeSurfer white matter segmentation missed some critical tissue in the missing regions. In MRICron, select '''Overlay>Add''' and select wm.nii, which should then appear as a semitransparent red mask. Do likewise for the c2001.nii (SPM white matter segmentation) and c1001.nii (SPM gray matter segmentation) files. As you load these overlays, you should get an idea of what tissues SPM identified as WM that do not appear in the wm.mgz file. Likewise, the SPM gray matter volume should help identify the extent of the pial surface.
			===3dcalc===
			There is no prescribed formula for this, but the information from the previous step can be used to guide what sort of operations you might want to do in 3dcalc to generate a fixed wm.mgz or enhanced T1.mgz file. Below are a few plausible scenarios:
			====White Matter Voxels Missing From wm.mgz===
			If the SPM WM segmentation looks reasonable, you might consider simply using that file in place of wm.mgz. I will note that SPM includes the cerebellum in its WM segmentation, whereas wm.mgz does not. I'm not certain if it's important to remove it, but you might simply convert a rescaled version of c2001.nii into wm.mgz. FreeSurfer expects the mean WM voxel value to be 110, however the voxels in c2001.nii range from 0.0 to 1.00, and are thus SPM's probabalistic determination that voxel v contains white matter. We can use 0.5 as a threshold using astep to set v<0.5 to 0.0 and v>0.5 to 1.0, then multiply those values by 110. The resulting volume will have 110 in all wm voxels, and 0.0 in the remainder. This file can be converted to a .mgz file and replace the one generated by FreeSurfer:
			3dcalc -a c2001.nii -expr 'astep(a,0.5)*110' -prefix rescaled_spm_wm.nii
			mri_convert rescaled_spm_wm.nii wm.mgz
			====Gray Matter Too Noisy/Dark====
			In a problem volume I'm looking at now, the FreeSurfer WM segmentation seemed to get most of what SPM found, so that may not entirely explain the missing anterior temporal tissue. What else might be the problem? I noted that the brainmask.mgz volume is pretty noisy in this area, so perhaps the apparently random distribution of higher/lower values within (what should be) the same tissue is causing a problem. Perhaps making the GM tissue a bit more homogeneous would help. I might start by selecting GM voxels below a certain value and increasing them to a threshold minimum (perhaps the average value of the rest of the GM volume). According to the [[https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllDevTable\|Recon-All wiki]] the tesselation step (the part that turns voxels into triangles) takes filled.mgz and norm.mgz as input volumes. (darnit, I think we needed to fix filled.mgz in the above step, rather than wm.mgz)
			This tells us that we can try using the SPM gray matter mask to fix norm.mgz and then re-tesselate.


	[[Category: FreeSurfer]]		[[Category: FreeSurfer]]

128.205.173.202: /* The Mysterious Case of the Missing Chunks */

2018-02-08T18:21:45Z

The Mysterious Case of the Missing Chunks

← Older revision		Revision as of 14:21, 8 February 2018
Line 76:		Line 76:
	After running autorecon2, you may notice that your surfaces don't exclude some of the anatomical voxels. This really only can be seen when checking quality via tkmedit, so be sure to do that. We've been primarily seeing this around the anterior temporal lobes. The cause is unknown, but we suspect it is related to the larger head coil used with the Semantic Imagery experiment.		After running autorecon2, you may notice that your surfaces don't exclude some of the anatomical voxels. This really only can be seen when checking quality via tkmedit, so be sure to do that. We've been primarily seeing this around the anterior temporal lobes. The cause is unknown, but we suspect it is related to the larger head coil used with the Semantic Imagery experiment.

	We have been experimenting with different methods to enhance the T1 MPRAGE image so that autorecon2 is better able to detect the white matter tracts and the gray matter boundaries and avoid missing large chunks of the anterior temporal lobes. The following technique seems to work, and it makes use of the segmentation files that SPM generates, and the intensity thresholding provided by MRICron. As of Feb 8, 2018, only ''Brocasarea'', ''Accumbens'' and ''Wernickesarea'' have SPM12 configured correctly, so you will need to copy the subject/mri folder to one of those ~~two~~ computers and work there.		We have been experimenting with different methods to enhance the T1 MPRAGE image so that autorecon2 is better able to detect the white matter tracts and the gray matter boundaries and avoid missing large chunks of the anterior temporal lobes. The following technique seems to work, and it makes use of the segmentation files that SPM generates, and the intensity thresholding provided by MRICron. As of Feb 8, 2018, only ''Brocasarea'', ''Accumbens'' and ''Wernickesarea'' have SPM12 configured correctly, so you will need to copy the subject/mri folder to one of those computers and work there.

	===Incorporating SPM Segmentation Files into Recon-All===		===Incorporating SPM Segmentation Files into Recon-All===

128.205.173.202: /* Incorporating SPM Segmentation Files into Recon-All */

2018-02-08T18:20:25Z

Incorporating SPM Segmentation Files into Recon-All

← Older revision		Revision as of 14:20, 8 February 2018
Line 79:		Line 79:

	===Incorporating SPM Segmentation Files into Recon-All===		===Incorporating SPM Segmentation Files into Recon-All===
	The first step will be to segment the original MPRAGE file as described in the section on SPM Segmentation under Brainmask Editing for [[~~autorecon1\|~~ Autorecon1#SPM_Segmentation]]. If you haven't done so already, follow those instructions first for the problematic brain.		The first step will be to segment the original MPRAGE file as described in the section on SPM Segmentation under Brainmask Editing for [[Autorecon1#SPM_Segmentation\|Autorecon1]]. If you haven't done so already, follow those instructions first for the problematic brain.

	At the end of this process, you will have a series of .nii files, prefixed with C1 to C5. We will be recombining these files and using MRICron to modify the WM.mgz file output by recon-all during the autorecon2 step.		At the end of this process, you will have a series of .nii files, prefixed with C1 to C5. We will be recombining these files and using MRICron to modify the WM.mgz file output by recon-all during the autorecon2 step.