External Email - Use Caution
Dear list members,
As a PhD student working on longitudinal child and adolescent T1-weighted data (.75x.75x.8 mm3) acquired at 3T (age range: 8 - 18 at baseline; 11 - 24 at follow-up), with the aim of doing cortical laminar analyses eventually, I am interested in getting the most of our lab's submillimeter resolution data. Therefore, in addition to a default recon-all run (with the -3T option enabled), I ran a separate recon-all with both the -3T and the -hires option enabled (with mris_inflate -n 50 added to the expert file), resulting in the following two runs: recon-all -all -i ${subj} -subjid ${SUB_ID} -3T and recon-all -all -i ${subj} -subjid ${SUB_ID} -3T -hires -expert $EXPERT_FILE Now, the surfaces (i.e., lh.pial, lh.white, rh.pial, rh.white) seem to be calculated more accurately for most brain regions, indeed, which looks very promising. However, the brainmask.mgz turns out to be of poor quality for a number of our subjects, cutting away a substantial portion of the brain. See the image below for an example of the same subject, on which both runs have been performed. The same slices are shown for each run. As can be seen in the image, a part of the the temporal region of the brain has been excluded in the brainmask.mgz of the -hires run (left in the image).
[cid:b4d26f40-11bb-4d8f-888c-ab4da9fea062]
Is there anything I can do, such as adjusting certain parameters, to prevent this from occurring? It would be a shame if I couldn't make use of the -hires option due to this, as it does seem to be a promising option in all other aspects.
Thank you for considering my question. I look forward to hearing from you.
Good holidays to you!
Kind regards, Simon Poortman (Erasmus MC, Rotterdam, the Netherlands)
External Email - Use Caution
Hi Simon,
in case your surfaces (and other outputs) look fine, there is no need to worry about the skull stripping too much. However, in case you what to do something about this you can do either of the following.
Which version are you using? I assume v6? In v7 the hires option basically has changed to what was known as the HCP pipeline before. Here, only the surface related parts of FreeSurfer are handled in native space. The other stages are handled in resampled 1mm space. Therefore, you will benefit from your higher resolution, while not having any of these issues. There are also some other stages included which may help with high resolution 3T data (e.g. denoising prior to surface placement).
In case you want to stick with v6 and what to run everything in native space, you can get better skull stripping following the first few steps from this guide I put together few years ago (and is still valid today):
https://secure-web.cisco.com/1cP9GZTFwQIDCR-WcWJPQA-tzd6FKvDGnFtvVRx4GcsgT8p...
What you basically do is to run the 1mm stream to create a nice brainmask (or any other tool you like), resample that to the resolution of your high resolution stream and apply the mask after the normalization stage. Then continue with autorecon2 and autrecon3.
Best,
Falk
http://secure-web.cisco.com/14B0CCxSnSk8LrteEbfghlzh-feG3u2xtCGi78x6DXWU9bSNZtxp8vwjgpFqaI4NxVFgO1Zp6xDGtfTImpLuiQrHGEmugVlg5jZPTtlr-ohLjOW89ll8xlm2Jr8I61Zr-vCQTOHR168GzDltq-PrIhvdgkqQDtRsuHmJx1a1kMwm_GddhhyFRZ5FHjhX9lfap0EzM_ZkWFB4-zmcrPCrxAeovnJ60OGefIvqrfJrU4r3Hy4JYnkhRUCRlWhNAWcRKTj7ann3pS3eNjhzGZe1pVA/http%3A%2F%2Fwww.kneu.ovgu.de%2Fkneu.html ________________________________ Von: freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edu im Auftrag von S.R. Poortman s.poortman@erasmusmc.nl Gesendet: Montag, 4. Januar 2021 09:13:11 An: freesurfer@nmr.mgh.harvard.edu Betreff: [Freesurfer] Fw: recon-all -hires option
External Email - Use Caution
Dear list members,
As a PhD student working on longitudinal child and adolescent T1-weighted data (.75x.75x.8 mm3) acquired at 3T (age range: 8 - 18 at baseline; 11 - 24 at follow-up), with the aim of doing cortical laminar analyses eventually, I am interested in getting the most of our lab's submillimeter resolution data. Therefore, in addition to a default recon-all run (with the -3T option enabled), I ran a separate recon-all with both the -3T and the -hires option enabled (with mris_inflate -n 50 added to the expert file), resulting in the following two runs: recon-all -all -i ${subj} -subjid ${SUB_ID} -3T and recon-all -all -i ${subj} -subjid ${SUB_ID} -3T -hires -expert $EXPERT_FILE Now, the surfaces (i.e., lh.pial, lh.white, rh.pial, rh.white) seem to be calculated more accurately for most brain regions, indeed, which looks very promising. However, the brainmask.mgz turns out to be of poor quality for a number of our subjects, cutting away a substantial portion of the brain. See the image below for an example of the same subject, on which both runs have been performed. The same slices are shown for each run. As can be seen in the image, a part of the the temporal region of the brain has been excluded in the brainmask.mgz of the -hires run (left in the image).
[cid:b4d26f40-11bb-4d8f-888c-ab4da9fea062]
Is there anything I can do, such as adjusting certain parameters, to prevent this from occurring? It would be a shame if I couldn't make use of the -hires option due to this, as it does seem to be a promising option in all other aspects.
Thank you for considering my question. I look forward to hearing from you.
Good holidays to you!
Kind regards, Simon Poortman (Erasmus MC, Rotterdam, the Netherlands)
If you want to try to do it systematically, you can add -wsless (should reduce the agressiveness of the skull strip)
On 1/4/2021 3:52 AM, falk.luesebrink@med.ovgu.demailto:falk.luesebrink@med.ovgu.de wrote:
External Email - Use Caution
Hi Simon,
in case your surfaces (and other outputs) look fine, there is no need to worry about the skull stripping too much. However, in case you what to do something about this you can do either of the following.
Which version are you using? I assume v6? In v7 the hires option basically has changed to what was known as the HCP pipeline before. Here, only the surface related parts of FreeSurfer are handled in native space. The other stages are handled in resampled 1mm space. Therefore, you will benefit from your higher resolution, while not having any of these issues. There are also some other stages included which may help with high resolution 3T data (e.g. denoising prior to surface placement).
In case you want to stick with v6 and what to run everything in native space, you can get better skull stripping following the first few steps from this guide I put together few years ago (and is still valid today):
https://secure-web.cisco.com/1cP9GZTFwQIDCR-WcWJPQA-tzd6FKvDGnFtvVRx4GcsgT8p...
What you basically do is to run the 1mm stream to create a nice brainmask (or any other tool you like), resample that to the resolution of your high resolution stream and apply the mask after the normalization stage. Then continue with autorecon2 and autrecon3.
Best,
Falk
http://secure-web.cisco.com/14B0CCxSnSk8LrteEbfghlzh-feG3u2xtCGi78x6DXWU9bSNZtxp8vwjgpFqaI4NxVFgO1Zp6xDGtfTImpLuiQrHGEmugVlg5jZPTtlr-ohLjOW89ll8xlm2Jr8I61Zr-vCQTOHR168GzDltq-PrIhvdgkqQDtRsuHmJx1a1kMwm_GddhhyFRZ5FHjhX9lfap0EzM_ZkWFB4-zmcrPCrxAeovnJ60OGefIvqrfJrU4r3Hy4JYnkhRUCRlWhNAWcRKTj7ann3pS3eNjhzGZe1pVA/http%3A%2F%2Fwww.kneu.ovgu.de%2Fkneu.htmlhttp://secure-web.cisco.com/14B0CCxSnSk8LrteEbfghlzh-feG3u2xtCGi78x6DXWU9bSNZtxp8vwjgpFqaI4NxVFgO1Zp6xDGtfTImpLuiQrHGEmugVlg5jZPTtlr-ohLjOW89ll8xlm2Jr8I61Zr-vCQTOHR168GzDltq-PrIhvdgkqQDtRsuHmJx1a1kMwm_GddhhyFRZ5FHjhX9lfap0EzM_ZkWFB4-zmcrPCrxAeovnJ60OGefIvqrfJrU4r3Hy4JYnkhRUCRlWhNAWcRKTj7ann3pS3eNjhzGZe1pVA/http%3A%2F%2Fwww.kneu.ovgu.de%2Fkneu.html ________________________________ Von: freesurfer-bounces@nmr.mgh.harvard.edumailto:freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edumailto:freesurfer-bounces@nmr.mgh.harvard.edu im Auftrag von S.R. Poortman s.poortman@erasmusmc.nlmailto:s.poortman@erasmusmc.nl Gesendet: Montag, 4. Januar 2021 09:13:11 An: freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu Betreff: [Freesurfer] Fw: recon-all -hires option
External Email - Use Caution
Dear list members,
As a PhD student working on longitudinal child and adolescent T1-weighted data (.75x.75x.8 mm3) acquired at 3T (age range: 8 - 18 at baseline; 11 - 24 at follow-up), with the aim of doing cortical laminar analyses eventually, I am interested in getting the most of our lab's submillimeter resolution data. Therefore, in addition to a default recon-all run (with the -3T option enabled), I ran a separate recon-all with both the -3T and the -hires option enabled (with mris_inflate -n 50 added to the expert file), resulting in the following two runs: recon-all -all -i ${subj} -subjid ${SUB_ID} -3T and recon-all -all -i ${subj} -subjid ${SUB_ID} -3T -hires -expert $EXPERT_FILE Now, the surfaces (i.e., lh.pial, lh.white, rh.pial, rh.white) seem to be calculated more accurately for most brain regions, indeed, which looks very promising. However, the brainmask.mgz turns out to be of poor quality for a number of our subjects, cutting away a substantial portion of the brain. See the image below for an example of the same subject, on which both runs have been performed. The same slices are shown for each run. As can be seen in the image, a part of the the temporal region of the brain has been excluded in the brainmask.mgz of the -hires run (left in the image).
[cid:b4d26f40-11bb-4d8f-888c-ab4da9fea062]
Is there anything I can do, such as adjusting certain parameters, to prevent this from occurring? It would be a shame if I couldn't make use of the -hires option due to this, as it does seem to be a promising option in all other aspects.
Thank you for considering my question. I look forward to hearing from you.
Good holidays to you!
Kind regards, Simon Poortman (Erasmus MC, Rotterdam, the Netherlands)
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
falk.luesebrink＠med.ovgu.de -
Greve, Douglas N.,Ph.D. -
S.R. Poortman