Request for corrected HCP tractography atlases to perform RecoBundles
Dear DIPY experts and Dr. Yeh, I am working on a doctoral project that involves tractography and intend to use RecoBundles (as described in https://www.ncbi.nlm.nih.gov/pubmed/28712994) with an HCP atlas in my streamline generation approach. Lately, I encountered several inconsistencies in various tractography atlases (coming from here: https://www.ncbi.nlm.nih.gov/pubmed/29758339). While implementing my code, I spotted that the DIPY-recommended HCP atlas (HCP842) lacks its reference template image. Here is the link to this atlas: https://figshare.com/articles/Advanced_Atlas_of_80_Bundles_in_MNI_space/7375.... In its README file, it is described that the bundles are in the MNI ICBM 2009a space, yet this is not the case when I overlay the TRK files on the 'mni_icbm152_t1_tal_nlin_asym_09a.nii' MNI template image (see Figure 1 attached). As you can see, the affine matrices are inconsistent as well. In this forum entry (https://github.com/dipy/dipy/issues/2007), I later found out that Dr. Francois Rheault had recommended using the 'corrected version' of the HCP842 series, registered to the 'mni_icbm152_t1_tal_nlin_asym_09c.nii' MNI ICBM 2009c space, available here: https://drive.google.com/open?id=1pObyKLGJNPFZhMIncYuLn9utJDKb-Uq9. Again, when I visualized the TRK tracks on the T1 reference, the tracks ended up lying outside of the reference geometry (see Figure 2). Moreover, I learnt there is another tractography atlas (HCP1065, available here: http://brain.labsolver.org/diffusion-mri-templates/tractography), which is coincidentally most ideal for my project, but does not come in the TRK format. Instead, it is available in the TT.GZ (https://pitt.app.box.com/v/HCP1065-TRACT) or NII.GZ (https://pitt.app.box.com/v/HCP1065-NIFTI) formats, which cannot be loaded using DIPY, can it? If you have the TRK-formatted tracks of this atlas, can you please share them with me? It appears that tractography atlas misalignments are an open issue that might not have been resolved yet: https://github.com/dipy/dipy/issues/2166. Also, it seems to me that the code example on the official DIPY website (https://dipy.org/documentation/1.3.0./examples_built/bundle_extraction/) contains a couple of confusing / buggy steps (target header creation, reference images for input TRK bundles in function load_trk(), etc.). As such, the TRK bundles of the HCP842 atlas suffice to perform virtual dissection, however, to be able to overlay the bundle models in any subject space (not in MNI) to compare the dissected bundles with these models, I need the right reference image ensuring the headers / coordinate systems correspond. I would appreciate your help / provision of a corrected and consistent atlas along with its reference template image in the coming days. Thank you very much in advance. Best regards, Daniel Krahulec, MSc (Tech.) PhD candidate Eindhoven University of Technology Eindhoven, The Netherlands
Hi Daniel, Sorry about all the trouble. I am going to convert the new 1065 atlas from TT to TRK and share it in this thread. Best, Frank On Fri, Dec 4, 2020 at 10:39 AM Krahulec, D. <d.krahulec@tue.nl> wrote:
-- Fang-Cheng (Frank) Yeh Assistant Professor Department of Neurological Surgery UPMC Presbyterian, Suite B-400 200 Lothrop Street Pittsburgh, PA 15213 TEL: 412-383-0871
I provide the TRK format for HCP1065 at http://brain.labsolver.org/diffusion-mri-templates/tractography If there is any question, please feel free to contact me. Frank On Fri, Dec 4, 2020 at 10:54 AM Fang-Cheng Yeh <frank.yeh@pitt.edu> wrote:
-- Fang-Cheng (Frank) Yeh Assistant Professor Department of Neurological Surgery UPMC Presbyterian, Suite B-400 200 Lothrop Street Pittsburgh, PA 15213 TEL: 412-383-0871
Dear Dr. Yeh, thank you for the fast supply of the tracks in TRK. Which exact reference space do they come from? Can they be viewed on their reference template image solely in DSI Studio? I tried another TRK viewer (http://users.loni.ucla.edu/~pipeline/viewer/), and used the 'mni_icbm152_t1_tal_nlin_asym_09a.nii' template, yet the tracks were again incorrectly aligned. What I would still need is the following: 1. correct NIfTI (.nii.gz) MNI T1 reference image (I can see the template image is available in FIB) 2. fused tractogram of all these tracks in TRK Your help is extremely valuable. Best regards, Daniel Krahulec, MSc (Tech.) PhD candidate Eindhoven University of Technology Eindhoven, The Netherlands ________________________________ From: Fang-Cheng Yeh <frank.yeh@pitt.edu> Sent: Friday, December 4, 2020 6:47 PM To: Krahulec, D. <d.krahulec@tue.nl> Cc: dipy@python.org <dipy@python.org>; timothyv@andrew.cmu.edu <timothyv@andrew.cmu.edu> Subject: Re: Request for corrected HCP tractography atlases to perform RecoBundles I provide the TRK format for HCP1065 at https://eur02.safelinks.protection.outlook.com/?url=http%3A%2F%2Fbrain.labso... If there is any question, please feel free to contact me. Frank On Fri, Dec 4, 2020 at 10:54 AM Fang-Cheng Yeh <frank.yeh@pitt.edu> wrote:
-- Fang-Cheng (Frank) Yeh Assistant Professor Department of Neurological Surgery UPMC Presbyterian, Suite B-400 200 Lothrop Street Pittsburgh, PA 15213 TEL: 412-383-0871
On Sat, Dec 5, 2020 at 10:30 AM Krahulec, D. <d.krahulec@tue.nl> wrote:
The coordinates are in the RAS orientation with the origin (0,0,0) at the LPI corner
Can they be viewed on their reference template image solely in DSI Studio?
Yes, you may download HCP1065 template (*.fib.gz) from brain.labsolver.org, open it in Step T3 at DSI Studio, and then load all of the tracks.
What I would still need is the following:
correct NIfTI (.nii.gz) MNI T1 reference image (I can see the template image is available in FIB)
It is the ICBM152 2009a image from https://www.bic.mni.mcgill.ca/ServicesAtlases/ICBM152NLin2009 Best, Frank
-- Fang-Cheng (Frank) Yeh Assistant Professor Department of Neurological Surgery UPMC Presbyterian, Suite B-400 200 Lothrop Street Pittsburgh, PA 15213 TEL: 412-383-0871
Hi Daniel and all, The problem appeared because you used a viewer outside DIPY that we do not use ourselves. You should not have any issues extracting bundles. We assume that streamlines are always in MNI space. We can change this and as you said. Francois, Serge and others have been looking into this small issue. This is a small issue because we already have the tools to update/correct the affines of tractograms. Daniel the Atlas with the 80 bundles has many cranial nerves etc that are not so useful for most analyses as they do not exist in the data. I suggest using the smaller 30 bundles atlas first. https://figshare.com/articles/dataset/Atlas_of_30_Human_Brain_Bundles_in_MNI... We should clarify that in our documentation. Bramsh can you help Daniel with his project and make sure all files are consistent? Make sure you report progress back here in this thread so that other people can benefit from the discussion? Also have in mind that the atlas is in MNI space 2009a but it is roughly aligned. This is due to how the template was created. The good news is that RecoBundles does not require perfect registration. Bramsh let's also look at the new atlas that Frank shared. Daniel, apologies for having some issues with this. Be sure these are things that can be corrected easily. We have deployed RecoBundles in projects with more than 4000 subjects with no issues. So Daniel, you should be happy RecoBundling very soon! Thank you for providing feedback and for pushing to improve consistency across neuroimaging tools which is an important issue that we do care about too. Best regards, Eleftherios <https://figshare.com/articles/dataset/Atlas_of_30_Human_Brain_Bundles_in_MNI...> On Fri, Dec 4, 2020 at 10:41 AM Krahulec, D. <d.krahulec@tue.nl> wrote:
Hi Eleftherios, thank you for your clear explanation. Today, I connected with Bramsh and we are going to meet in the week of Dec 14. I suppose looking into the latest atlas (HCP1065) would be beneficial for most users. It is my pleasure to provide feedback and get problems solved. Regards, Daniel ________________________________ From: Eleftherios Garyfallidis <elef@indiana.edu> Sent: Saturday, December 5, 2020 5:37 PM To: Krahulec, D. <d.krahulec@tue.nl>; Bramsh Qamar <bramshq@gmail.com>; François Rheault <Francois.rheault@gmail.com> Cc: dipy@python.org <dipy@python.org>; frank.yeh@pitt.edu <frank.yeh@pitt.edu>; timothyv@andrew.cmu.edu <timothyv@andrew.cmu.edu> Subject: Re: [Dipy] Request for corrected HCP tractography atlases to perform RecoBundles Hi Daniel and all, The problem appeared because you used a viewer outside DIPY that we do not use ourselves. You should not have any issues extracting bundles. We assume that streamlines are always in MNI space. We can change this and as you said. Francois, Serge and others have been looking into this small issue. This is a small issue because we already have the tools to update/correct the affines of tractograms. Daniel the Atlas with the 80 bundles has many cranial nerves etc that are not so useful for most analyses as they do not exist in the data. I suggest using the smaller 30 bundles atlas first. https://figshare.com/articles/dataset/Atlas_of_30_Human_Brain_Bundles_in_MNI_space/12089652<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Ffigshare.com%2Farticles%2Fdataset%2FAtlas_of_30_Human_Brain_Bundles_in_MNI_space%2F12089652&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616120744%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=EhPwiVZDlxRgpKIVHt5UN9ak0W17HzmsB3SGs3XsKbQ%3D&reserved=0> We should clarify that in our documentation. Bramsh can you help Daniel with his project and make sure all files are consistent? Make sure you report progress back here in this thread so that other people can benefit from the discussion? Also have in mind that the atlas is in MNI space 2009a but it is roughly aligned. This is due to how the template was created. The good news is that RecoBundles does not require perfect registration. Bramsh let's also look at the new atlas that Frank shared. Daniel, apologies for having some issues with this. Be sure these are things that can be corrected easily. We have deployed RecoBundles in projects with more than 4000 subjects with no issues. So Daniel, you should be happy RecoBundling very soon! Thank you for providing feedback and for pushing to improve consistency across neuroimaging tools which is an important issue that we do care about too. Best regards, Eleftherios <https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Ffigshare.com%2Farticles%2Fdataset%2FAtlas_of_30_Human_Brain_Bundles_in_MNI_space%2F12089652&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616130736%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=PBk4dlk9Yh%2FI%2Fv5AZ%2F7j6O7PabTGWKz7FOiXNsmkUW8%3D&reserved=0> On Fri, Dec 4, 2020 at 10:41 AM Krahulec, D. <d.krahulec@tue.nl<mailto:d.krahulec@tue.nl>> wrote: Dear DIPY experts and Dr. Yeh, I am working on a doctoral project that involves tractography and intend to use RecoBundles (as described in https://www.ncbi.nlm.nih.gov/pubmed/28712994<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ncbi.nlm.nih.gov%2Fpubmed%2F28712994&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616130736%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=jPvnsTh2%2F5PCKUWFUCr3Buei5qShRFAsybkBG9aA6Ao%3D&reserved=0>) with an HCP atlas in my streamline generation approach. Lately, I encountered several inconsistencies in various tractography atlases (coming from here: https://www.ncbi.nlm.nih.gov/pubmed/29758339<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ncbi.nlm.nih.gov%2Fpubmed%2F29758339&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616140726%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=Poc6Ht1%2FHoON%2B3a4cJQaHJYGe5HqdJDe0TT%2BdX8eUQY%3D&reserved=0>). While implementing my code, I spotted that the DIPY-recommended HCP atlas (HCP842) lacks its reference template image. Here is the link to this atlas: https://figshare.com/articles/Advanced_Atlas_of_80_Bundles_in_MNI_space/7375883<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Ffigshare.com%2Farticles%2FAdvanced_Atlas_of_80_Bundles_in_MNI_space%2F7375883&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616150721%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=1o%2FiMbjXoXPAocGacKiXCzHUW9QlEK9At%2FymJZzy1bs%3D&reserved=0>. In its README file, it is described that the bundles are in the MNI ICBM 2009a space, yet this is not the case when I overlay the TRK files on the 'mni_icbm152_t1_tal_nlin_asym_09a.nii' MNI template image (see Figure 1 attached). As you can see, the affine matrices are inconsistent as well. In this forum entry (https://github.com/dipy/dipy/issues/2007<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fdipy%2Fdipy%2Fissues%2F2007&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616150721%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=a%2BRO0dLO7ic333MImbWQbdJxOd06hp8MEblTI2xJs2Y%3D&reserved=0>), I later found out that Dr. Francois Rheault had recommended using the 'corrected version' of the HCP842 series, registered to the 'mni_icbm152_t1_tal_nlin_asym_09c.nii' MNI ICBM 2009c space, available here: https://drive.google.com/open?id=1pObyKLGJNPFZhMIncYuLn9utJDKb-Uq9<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdrive.google.com%2Fopen%3Fid%3D1pObyKLGJNPFZhMIncYuLn9utJDKb-Uq9&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616160716%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=BRcsVtMFmvjuvq4%2FA4dlGEXcDQ0MyX%2FRvXezdjXJ8v0%3D&reserved=0>. Again, when I visualized the TRK tracks on the T1 reference, the tracks ended up lying outside of the reference geometry (see Figure 2). Moreover, I learnt there is another tractography atlas (HCP1065, available here: http://brain.labsolver.org/diffusion-mri-templates/tractography<https://eur02.safelinks.protection.outlook.com/?url=http%3A%2F%2Fbrain.labsolver.org%2Fdiffusion-mri-templates%2Ftractography&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616170716%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=XaruxThLjveG%2FdtU%2Fcus2kAruGIWVTIXvhwXP%2B2AQrQ%3D&reserved=0>), which is coincidentally most ideal for my project, but does not come in the TRK format. Instead, it is available in the TT.GZ (https://pitt.app.box.com/v/HCP1065-TRACT<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpitt.app.box.com%2Fv%2FHCP1065-TRACT&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616170716%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=olQ%2FqzMRDyBJVJUYz%2FgXbjXGUtiJQQxX%2FEJ4rc%2FAyZk%3D&reserved=0>) or NII.GZ (https://pitt.app.box.com/v/HCP1065-NIFTI<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpitt.app.box.com%2Fv%2FHCP1065-NIFTI&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637427830616180711%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=D4LDjs3l4oG9ZNEgj24ndQ3zIQAroZvM9nI%2BtDojiQc%3D&reserved=0>) formats, which cannot be loaded using DIPY, can it? If you have the TRK-formatted tracks of this atlas, can you please share them with me? It appears that tractography atlas misalignments are an open issue that might not have been resolved yet: https://github.com/dipy/dipy/issues/2166<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fdipy%2Fdipy%2Fissues%2F2166&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616190710%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=q6vhlUmdD0XLmKwE8tctVcejHBcpb13xUg%2BUaA0pWpU%3D&reserved=0>. Also, it seems to me that the code example on the official DIPY website (https://dipy.org/documentation/1.3.0./examples_built/bundle_extraction/<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdipy.org%2Fdocumentation%2F1.3.0.%2Fexamples_built%2Fbundle_extraction%2F&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616190710%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=qvOP66feo11%2FsKd2ALBT%2F0fuFLVfjbDeC1BeVmMiBl8%3D&reserved=0>) contains a couple of confusing / buggy steps (target header creation, reference images for input TRK bundles in function load_trk(), etc.). As such, the TRK bundles of the HCP842 atlas suffice to perform virtual dissection, however, to be able to overlay the bundle models in any subject space (not in MNI) to compare the dissected bundles with these models, I need the right reference image ensuring the headers / coordinate systems correspond. I would appreciate your help / provision of a corrected and consistent atlas along with its reference template image in the coming days. Thank you very much in advance. Best regards, Daniel Krahulec, MSc (Tech.) PhD candidate Eindhoven University of Technology Eindhoven, The Netherlands _______________________________________________ Dipy mailing list -- dipy@python.org<mailto:dipy@python.org> To unsubscribe send an email to dipy-leave@python.org<mailto:dipy-leave@python.org> https://mail.python.org/mailman3/lists/dipy.python.org/<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmail.python.org%2Fmailman3%2Flists%2Fdipy.python.org%2F&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637427830616200699%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=oF77r5qMT9I%2Fd7EqPljS3rLchoG3ikmYTP7S%2Bf2JxPM%3D&reserved=0> Member address: garyfallidis@gmail.com<mailto:garyfallidis@gmail.com>
Hi Daniel, Sorry about all the trouble. I am going to convert the new 1065 atlas from TT to TRK and share it in this thread. Best, Frank On Fri, Dec 4, 2020 at 10:39 AM Krahulec, D. <d.krahulec@tue.nl> wrote:
-- Fang-Cheng (Frank) Yeh Assistant Professor Department of Neurological Surgery UPMC Presbyterian, Suite B-400 200 Lothrop Street Pittsburgh, PA 15213 TEL: 412-383-0871
I provide the TRK format for HCP1065 at http://brain.labsolver.org/diffusion-mri-templates/tractography If there is any question, please feel free to contact me. Frank On Fri, Dec 4, 2020 at 10:54 AM Fang-Cheng Yeh <frank.yeh@pitt.edu> wrote:
-- Fang-Cheng (Frank) Yeh Assistant Professor Department of Neurological Surgery UPMC Presbyterian, Suite B-400 200 Lothrop Street Pittsburgh, PA 15213 TEL: 412-383-0871
Dear Dr. Yeh, thank you for the fast supply of the tracks in TRK. Which exact reference space do they come from? Can they be viewed on their reference template image solely in DSI Studio? I tried another TRK viewer (http://users.loni.ucla.edu/~pipeline/viewer/), and used the 'mni_icbm152_t1_tal_nlin_asym_09a.nii' template, yet the tracks were again incorrectly aligned. What I would still need is the following: 1. correct NIfTI (.nii.gz) MNI T1 reference image (I can see the template image is available in FIB) 2. fused tractogram of all these tracks in TRK Your help is extremely valuable. Best regards, Daniel Krahulec, MSc (Tech.) PhD candidate Eindhoven University of Technology Eindhoven, The Netherlands ________________________________ From: Fang-Cheng Yeh <frank.yeh@pitt.edu> Sent: Friday, December 4, 2020 6:47 PM To: Krahulec, D. <d.krahulec@tue.nl> Cc: dipy@python.org <dipy@python.org>; timothyv@andrew.cmu.edu <timothyv@andrew.cmu.edu> Subject: Re: Request for corrected HCP tractography atlases to perform RecoBundles I provide the TRK format for HCP1065 at https://eur02.safelinks.protection.outlook.com/?url=http%3A%2F%2Fbrain.labso... If there is any question, please feel free to contact me. Frank On Fri, Dec 4, 2020 at 10:54 AM Fang-Cheng Yeh <frank.yeh@pitt.edu> wrote:
-- Fang-Cheng (Frank) Yeh Assistant Professor Department of Neurological Surgery UPMC Presbyterian, Suite B-400 200 Lothrop Street Pittsburgh, PA 15213 TEL: 412-383-0871
On Sat, Dec 5, 2020 at 10:30 AM Krahulec, D. <d.krahulec@tue.nl> wrote:
The coordinates are in the RAS orientation with the origin (0,0,0) at the LPI corner
Can they be viewed on their reference template image solely in DSI Studio?
Yes, you may download HCP1065 template (*.fib.gz) from brain.labsolver.org, open it in Step T3 at DSI Studio, and then load all of the tracks.
What I would still need is the following:
correct NIfTI (.nii.gz) MNI T1 reference image (I can see the template image is available in FIB)
It is the ICBM152 2009a image from https://www.bic.mni.mcgill.ca/ServicesAtlases/ICBM152NLin2009 Best, Frank
-- Fang-Cheng (Frank) Yeh Assistant Professor Department of Neurological Surgery UPMC Presbyterian, Suite B-400 200 Lothrop Street Pittsburgh, PA 15213 TEL: 412-383-0871
Hi Daniel and all, The problem appeared because you used a viewer outside DIPY that we do not use ourselves. You should not have any issues extracting bundles. We assume that streamlines are always in MNI space. We can change this and as you said. Francois, Serge and others have been looking into this small issue. This is a small issue because we already have the tools to update/correct the affines of tractograms. Daniel the Atlas with the 80 bundles has many cranial nerves etc that are not so useful for most analyses as they do not exist in the data. I suggest using the smaller 30 bundles atlas first. https://figshare.com/articles/dataset/Atlas_of_30_Human_Brain_Bundles_in_MNI... We should clarify that in our documentation. Bramsh can you help Daniel with his project and make sure all files are consistent? Make sure you report progress back here in this thread so that other people can benefit from the discussion? Also have in mind that the atlas is in MNI space 2009a but it is roughly aligned. This is due to how the template was created. The good news is that RecoBundles does not require perfect registration. Bramsh let's also look at the new atlas that Frank shared. Daniel, apologies for having some issues with this. Be sure these are things that can be corrected easily. We have deployed RecoBundles in projects with more than 4000 subjects with no issues. So Daniel, you should be happy RecoBundling very soon! Thank you for providing feedback and for pushing to improve consistency across neuroimaging tools which is an important issue that we do care about too. Best regards, Eleftherios <https://figshare.com/articles/dataset/Atlas_of_30_Human_Brain_Bundles_in_MNI...> On Fri, Dec 4, 2020 at 10:41 AM Krahulec, D. <d.krahulec@tue.nl> wrote:
Hi Eleftherios, thank you for your clear explanation. Today, I connected with Bramsh and we are going to meet in the week of Dec 14. I suppose looking into the latest atlas (HCP1065) would be beneficial for most users. It is my pleasure to provide feedback and get problems solved. Regards, Daniel ________________________________ From: Eleftherios Garyfallidis <elef@indiana.edu> Sent: Saturday, December 5, 2020 5:37 PM To: Krahulec, D. <d.krahulec@tue.nl>; Bramsh Qamar <bramshq@gmail.com>; François Rheault <Francois.rheault@gmail.com> Cc: dipy@python.org <dipy@python.org>; frank.yeh@pitt.edu <frank.yeh@pitt.edu>; timothyv@andrew.cmu.edu <timothyv@andrew.cmu.edu> Subject: Re: [Dipy] Request for corrected HCP tractography atlases to perform RecoBundles Hi Daniel and all, The problem appeared because you used a viewer outside DIPY that we do not use ourselves. You should not have any issues extracting bundles. We assume that streamlines are always in MNI space. We can change this and as you said. Francois, Serge and others have been looking into this small issue. This is a small issue because we already have the tools to update/correct the affines of tractograms. Daniel the Atlas with the 80 bundles has many cranial nerves etc that are not so useful for most analyses as they do not exist in the data. I suggest using the smaller 30 bundles atlas first. https://figshare.com/articles/dataset/Atlas_of_30_Human_Brain_Bundles_in_MNI_space/12089652<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Ffigshare.com%2Farticles%2Fdataset%2FAtlas_of_30_Human_Brain_Bundles_in_MNI_space%2F12089652&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616120744%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=EhPwiVZDlxRgpKIVHt5UN9ak0W17HzmsB3SGs3XsKbQ%3D&reserved=0> We should clarify that in our documentation. Bramsh can you help Daniel with his project and make sure all files are consistent? Make sure you report progress back here in this thread so that other people can benefit from the discussion? Also have in mind that the atlas is in MNI space 2009a but it is roughly aligned. This is due to how the template was created. The good news is that RecoBundles does not require perfect registration. Bramsh let's also look at the new atlas that Frank shared. Daniel, apologies for having some issues with this. Be sure these are things that can be corrected easily. We have deployed RecoBundles in projects with more than 4000 subjects with no issues. So Daniel, you should be happy RecoBundling very soon! Thank you for providing feedback and for pushing to improve consistency across neuroimaging tools which is an important issue that we do care about too. Best regards, Eleftherios <https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Ffigshare.com%2Farticles%2Fdataset%2FAtlas_of_30_Human_Brain_Bundles_in_MNI_space%2F12089652&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616130736%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=PBk4dlk9Yh%2FI%2Fv5AZ%2F7j6O7PabTGWKz7FOiXNsmkUW8%3D&reserved=0> On Fri, Dec 4, 2020 at 10:41 AM Krahulec, D. <d.krahulec@tue.nl<mailto:d.krahulec@tue.nl>> wrote: Dear DIPY experts and Dr. Yeh, I am working on a doctoral project that involves tractography and intend to use RecoBundles (as described in https://www.ncbi.nlm.nih.gov/pubmed/28712994<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ncbi.nlm.nih.gov%2Fpubmed%2F28712994&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616130736%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=jPvnsTh2%2F5PCKUWFUCr3Buei5qShRFAsybkBG9aA6Ao%3D&reserved=0>) with an HCP atlas in my streamline generation approach. Lately, I encountered several inconsistencies in various tractography atlases (coming from here: https://www.ncbi.nlm.nih.gov/pubmed/29758339<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ncbi.nlm.nih.gov%2Fpubmed%2F29758339&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616140726%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=Poc6Ht1%2FHoON%2B3a4cJQaHJYGe5HqdJDe0TT%2BdX8eUQY%3D&reserved=0>). While implementing my code, I spotted that the DIPY-recommended HCP atlas (HCP842) lacks its reference template image. Here is the link to this atlas: https://figshare.com/articles/Advanced_Atlas_of_80_Bundles_in_MNI_space/7375883<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Ffigshare.com%2Farticles%2FAdvanced_Atlas_of_80_Bundles_in_MNI_space%2F7375883&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616150721%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=1o%2FiMbjXoXPAocGacKiXCzHUW9QlEK9At%2FymJZzy1bs%3D&reserved=0>. In its README file, it is described that the bundles are in the MNI ICBM 2009a space, yet this is not the case when I overlay the TRK files on the 'mni_icbm152_t1_tal_nlin_asym_09a.nii' MNI template image (see Figure 1 attached). As you can see, the affine matrices are inconsistent as well. In this forum entry (https://github.com/dipy/dipy/issues/2007<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fdipy%2Fdipy%2Fissues%2F2007&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616150721%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=a%2BRO0dLO7ic333MImbWQbdJxOd06hp8MEblTI2xJs2Y%3D&reserved=0>), I later found out that Dr. Francois Rheault had recommended using the 'corrected version' of the HCP842 series, registered to the 'mni_icbm152_t1_tal_nlin_asym_09c.nii' MNI ICBM 2009c space, available here: https://drive.google.com/open?id=1pObyKLGJNPFZhMIncYuLn9utJDKb-Uq9<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdrive.google.com%2Fopen%3Fid%3D1pObyKLGJNPFZhMIncYuLn9utJDKb-Uq9&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616160716%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=BRcsVtMFmvjuvq4%2FA4dlGEXcDQ0MyX%2FRvXezdjXJ8v0%3D&reserved=0>. Again, when I visualized the TRK tracks on the T1 reference, the tracks ended up lying outside of the reference geometry (see Figure 2). Moreover, I learnt there is another tractography atlas (HCP1065, available here: http://brain.labsolver.org/diffusion-mri-templates/tractography<https://eur02.safelinks.protection.outlook.com/?url=http%3A%2F%2Fbrain.labsolver.org%2Fdiffusion-mri-templates%2Ftractography&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616170716%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=XaruxThLjveG%2FdtU%2Fcus2kAruGIWVTIXvhwXP%2B2AQrQ%3D&reserved=0>), which is coincidentally most ideal for my project, but does not come in the TRK format. Instead, it is available in the TT.GZ (https://pitt.app.box.com/v/HCP1065-TRACT<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpitt.app.box.com%2Fv%2FHCP1065-TRACT&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616170716%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=olQ%2FqzMRDyBJVJUYz%2FgXbjXGUtiJQQxX%2FEJ4rc%2FAyZk%3D&reserved=0>) or NII.GZ (https://pitt.app.box.com/v/HCP1065-NIFTI<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpitt.app.box.com%2Fv%2FHCP1065-NIFTI&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637427830616180711%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=D4LDjs3l4oG9ZNEgj24ndQ3zIQAroZvM9nI%2BtDojiQc%3D&reserved=0>) formats, which cannot be loaded using DIPY, can it? If you have the TRK-formatted tracks of this atlas, can you please share them with me? It appears that tractography atlas misalignments are an open issue that might not have been resolved yet: https://github.com/dipy/dipy/issues/2166<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fdipy%2Fdipy%2Fissues%2F2166&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616190710%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=q6vhlUmdD0XLmKwE8tctVcejHBcpb13xUg%2BUaA0pWpU%3D&reserved=0>. Also, it seems to me that the code example on the official DIPY website (https://dipy.org/documentation/1.3.0./examples_built/bundle_extraction/<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdipy.org%2Fdocumentation%2F1.3.0.%2Fexamples_built%2Fbundle_extraction%2F&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C1%7C637427830616190710%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=qvOP66feo11%2FsKd2ALBT%2F0fuFLVfjbDeC1BeVmMiBl8%3D&reserved=0>) contains a couple of confusing / buggy steps (target header creation, reference images for input TRK bundles in function load_trk(), etc.). As such, the TRK bundles of the HCP842 atlas suffice to perform virtual dissection, however, to be able to overlay the bundle models in any subject space (not in MNI) to compare the dissected bundles with these models, I need the right reference image ensuring the headers / coordinate systems correspond. I would appreciate your help / provision of a corrected and consistent atlas along with its reference template image in the coming days. Thank you very much in advance. Best regards, Daniel Krahulec, MSc (Tech.) PhD candidate Eindhoven University of Technology Eindhoven, The Netherlands _______________________________________________ Dipy mailing list -- dipy@python.org<mailto:dipy@python.org> To unsubscribe send an email to dipy-leave@python.org<mailto:dipy-leave@python.org> https://mail.python.org/mailman3/lists/dipy.python.org/<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmail.python.org%2Fmailman3%2Flists%2Fdipy.python.org%2F&data=04%7C01%7C%7Ca204151e310c4eaf3c8008d8993c1392%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637427830616200699%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=oF77r5qMT9I%2Fd7EqPljS3rLchoG3ikmYTP7S%2Bf2JxPM%3D&reserved=0> Member address: garyfallidis@gmail.com<mailto:garyfallidis@gmail.com>
participants (3)
-
Eleftherios Garyfallidis -
Fang-Cheng Yeh -
Krahulec, D.