Re: [Freesurfer] question about MRF
Dear Bruce, Thank you very much for your suggestions. I have fixed the reslicing problem and I am still trying to use mri_ca_label to relabel sub-cortical structures in my data. Now the program output a new seg, but the display of tkmedit mysubj norm.mgz -segmentation new_seg.mgz $FREESURFER_HOME/FreeSurferColorLUT.txt showed the orientation wrong (upside down and maybe left-right fliped). This is very weird since I have used tkmedit mysubj norm.mgz -segmentation old_seg.mgz $FREESURFER_HOME/FreeSurferColorLUT.txt to confirm that old_seg.mgz has the same orientation with norm.mgz. Do you have any idea about this? What's the function of transforms/talairach.m3z in mri_ca_label? Could you please send me a simple example T1, init_seg and relabeled result archived by mri_ca_label? Thanks! Best, Zhennan Zhennan Yan Department of Computer Science Rutgers University On 2/1/2013 7:44 PM, Bruce Fischl wrote: that might be the problem. Specify nearest nbr resampling with -rt nearest on the mri_convert cmd line cheers Bruce On Fri, 1 Feb 2013, Zhennan Yan wrote: Dear Bruce, I have mapped my labels to the Freesurfer labels. But when I tried the mri_ca_label -R init_label.mgz norm.mgz transforms/talairach.m3z $FREESURFER_HOME/average/RB_all_2008-03-26.gca new_label.mgz, I got a Segmentation fault after 0 singular and 1812 ill-conditioned covariance matrices regularized. I noticed that my initial segmentation image is not 255 cubed and has only 30+ different subcortical labels (for example labels 0,2,3,4,5,7,8,10,...,60,62). Then I tried to use the command mri_convert init_label.mgz init_label_c.mgz --conform to conform it. But the output init_label_c.mgz is not as I expected. I guess the reslicing interpolation might introduce some new labels to my initial segmentation? My question is do I need to conform the initial seg into 255 cubed image? How can I do this in my case? And what do you think is the problem cause the failure of the mri_ca_label? Thanks! Zhennan Zhennan Yan Department of Computer Science Rutgers University On 2/1/2013 10:28 AM, Bruce Fischl wrote: thanks Zhennan good luck and let us know if we can help Bruce On Fri, 1 Feb 2013, Zhennan Yan wrote: Dear Bruce, Thank you for your help and the great work on Freesurfer! Best, Zhennan Zhennan Yan Department of Computer Science Rutgers University On 1/31/2013 1:43 PM, Bruce Fischl wrote: 1. 40 subjects, 10 young, 10 middle ages, 10 elderly (healthy) and 10 with AD. 2. Many small things, the big one (described in our 2004 paper) is that we now use a nonlinear warp 3. Yes, hydrocephalus for example is hard, although we have made recent improvements in the warping that handle enlarged ventricles much better cheers Bruce On Thu, 31 Jan 2013, Zhennan Yan wrote: Dear Bruce, Thank you very much for your reply. I can convert my labels to Freesurfer labels and try the mri_ca_label now. May I ask some more questions? 1) What subjects were used to train the atlas in the current version? Were they healthy or diseased subjects? 2) How the current implementation differs from the one in your Whole Brain Segmentation (2002) paper? 3) Is there any cases, e.g. on which clinical diseases, FreeSurfer will produce less than ideal results? Thanks! Best, Zhennan Zhennan Yan Department of Computer Science Rutgers University On 1/30/2013 3:10 PM, Bruce Fischl wrote: Hi Zhennan there is an option to mri_ca_label to read in a prior segmentation and iterate the MRF minimization, but unless your labels and conventions are identical to our's it's unlikely to work. You can use mri_ca_train to build the MRF and likelihoods and such if you want. cheers Bruce On Wed, 30 Jan 2013, Zhennan Yan wrote: Dear freesurfer experts, I have questions about the MRF in Freesurfer: (1) I have a T1 image and sub-cortical segmentation for it (nifti format). But the segmentation is not perfect and I would like to relabel it using the ICM algorithm as shown in fig.8 in paper: Whole Brain Segmentation: Automated Labeling of Neuroanatomical Structures in the Human Brain, Fischl et al., (2002). How can I do this using Freesurfer? (2) If I have several subjects with labels, how can I train the anisotropic nonstationary MRF prior as equation (9) in the same paper? Can I use that MRF prior elsewhere in my own program? Many thanks! Best, Zhennan The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail. --
Re: [Freesurfer] question about MRF
Hi Zhennan if you run recon-all on any dataset it will create an example for you. The talairach.m3z maps the volume into the atlas coords. If it is oriented incorrectly then that would explain your problem. You can also use mri_info to see the direction cosines of the various volumes cheers Bruce On Sat, 16 Feb 2013, Zhennan Yan wrote: Dear Bruce, Thank you very much for your suggestions. I have fixed the reslicing problem and I am still trying to use mri_ca_label to relabel sub-cortical structures in my data. Now the program output a new seg, but the display of tkmedit mysubj norm.mgz -segmentation new_seg.mgz $FREESURFER_HOME/FreeSurferColorLUT.txt showed the orientation wrong (upside down and maybe left-right fliped). This is very weird since I have used tkmedit mysubj norm.mgz -segmentation old_seg.mgz $FREESURFER_HOME/FreeSurferColorLUT.txt to confirm that old_seg.mgz has the same orientation with norm.mgz. Do you have any idea about this? What's the function of transforms/talairach.m3z in mri_ca_label? Could you please send me a simple example T1, init_seg and relabeled result archived by mri_ca_label? Thanks! Best, Zhennan Zhennan Yan Department of Computer Science Rutgers University On 2/1/2013 7:44 PM, Bruce Fischl wrote: that might be the problem. Specify nearest nbr resampling with -rt nearest on the mri_convert cmd line cheers Bruce On Fri, 1 Feb 2013, Zhennan Yan wrote: Dear Bruce, I have mapped my labels to the Freesurfer labels. But when I tried the mri_ca_label -R init_label.mgz norm.mgz transforms/talairach.m3z $FREESURFER_HOME/average/RB_all_2008-03-26.gca new_label.mgz, I got a Segmentation fault after 0 singular and 1812 ill-conditioned covariance matrices regularized. I noticed that my initial segmentation image is not 255 cubed and has only 30+ different subcortical labels (for example labels 0,2,3,4,5,7,8,10,...,60,62). Then I tried to use the command mri_convert init_label.mgz init_label_c.mgz --conform to conform it. But the output init_label_c.mgz is not as I expected. I guess the reslicing interpolation might introduce some new labels to my initial segmentation? My question is do I need to conform the initial seg into 255 cubed image? How can I do this in my case? And what do you think is the problem cause the failure of the mri_ca_label? Thanks! Zhennan Zhennan Yan Department of Computer Science Rutgers University On 2/1/2013 10:28 AM, Bruce Fischl wrote: thanks Zhennan good luck and let us know if we can help Bruce On Fri, 1 Feb 2013, Zhennan Yan wrote: Dear Bruce, Thank you for your help and the great work on Freesurfer! Best, Zhennan Zhennan Yan Department of Computer Science Rutgers University On 1/31/2013 1:43 PM, Bruce Fischl wrote: 1. 40 subjects, 10 young, 10 middle ages, 10 elderly (healthy) and 10 with AD. 2. Many small things, the big one (described in our 2004 paper) is that we now use a nonlinear warp 3. Yes, hydrocephalus for example is hard, although we have made recent improvements in the warping that handle enlarged ventricles much better cheers Bruce On Thu, 31 Jan 2013, Zhennan Yan wrote: Dear Bruce, Thank you very much for your reply. I can convert my labels to Freesurfer labels and try the mri_ca_label now. May I ask some more questions? 1) What subjects were used to train the atlas in the current version? Were they healthy or diseased subjects? 2) How the current implementation differs from the one in your Whole Brain Segmentation (2002) paper? 3) Is there any cases, e.g. on which clinical diseases, FreeSurfer will produce less than ideal results? Thanks! Best, Zhennan Zhennan Yan Department of Computer Science Rutgers University On 1/30/2013 3:10 PM, Bruce Fischl wrote: Hi Zhennan there is an option to mri_ca_label to read in a prior segmentation and iterate the MRF minimization, but unless your labels and conventions are identical to our's it's unlikely to work. You can use mri_ca_train to build the MRF and likelihoods and such if you want. cheers Bruce On Wed, 30 Jan 2013, Zhennan Yan wrote: Dear freesurfer experts, I have questions about the MRF in Freesurfer: (1) I have a T1 image and sub-cortical segmentation for it (nifti format). But the segmentation is not perfect and I would like to relabel it using the ICM algorithm as shown in fig.8 in paper: Whole Brain Segmentation: Automated Labeling of Neuroanatomical Structures in the Human Brain, Fischl et al., (2002). How can I do this using Freesurfer? (2) If I have several subjects with labels, how can I train the anisotropic nonstationary MRF prior as equation (9) in the same paper? Can I use that MRF prior elsewhere in my own program? Many thanks! Best, Zhennan The information in this e-mail is intended only for the person to whom it is
Re: [Freesurfer] question about MRF
that might be the problem. Specify nearest nbr resampling with -rt nearest on the mri_convert cmd line cheers Bruce On Fri, 1 Feb 2013, Zhennan Yan wrote: Dear Bruce, I have mapped my labels to the Freesurfer labels. But when I tried the mri_ca_label -R init_label.mgz norm.mgz transforms/talairach.m3z $FREESURFER_HOME/average/RB_all_2008-03-26.gca new_label.mgz, I got a Segmentation fault after 0 singular and 1812 ill-conditioned covariance matrices regularized. I noticed that my initial segmentation image is not 255 cubed and has only 30+ different subcortical labels (for example labels 0,2,3,4,5,7,8,10,...,60,62). Then I tried to use the command mri_convert init_label.mgz init_label_c.mgz --conform to conform it. But the output init_label_c.mgz is not as I expected. I guess the reslicing interpolation might introduce some new labels to my initial segmentation? My question is do I need to conform the initial seg into 255 cubed image? How can I do this in my case? And what do you think is the problem cause the failure of the mri_ca_label? Thanks! Zhennan Zhennan Yan Department of Computer Science Rutgers University On 2/1/2013 10:28 AM, Bruce Fischl wrote: thanks Zhennan good luck and let us know if we can help Bruce On Fri, 1 Feb 2013, Zhennan Yan wrote: Dear Bruce, Thank you for your help and the great work on Freesurfer! Best, Zhennan Zhennan Yan Department of Computer Science Rutgers University On 1/31/2013 1:43 PM, Bruce Fischl wrote: 1. 40 subjects, 10 young, 10 middle ages, 10 elderly (healthy) and 10 with AD. 2. Many small things, the big one (described in our 2004 paper) is that we now use a nonlinear warp 3. Yes, hydrocephalus for example is hard, although we have made recent improvements in the warping that handle enlarged ventricles much better cheers Bruce On Thu, 31 Jan 2013, Zhennan Yan wrote: Dear Bruce, Thank you very much for your reply. I can convert my labels to Freesurfer labels and try the mri_ca_label now. May I ask some more questions? 1) What subjects were used to train the atlas in the current version? Were they healthy or diseased subjects? 2) How the current implementation differs from the one in your Whole Brain Segmentation (2002) paper? 3) Is there any cases, e.g. on which clinical diseases, FreeSurfer will produce less than ideal results? Thanks! Best, Zhennan Zhennan Yan Department of Computer Science Rutgers University On 1/30/2013 3:10 PM, Bruce Fischl wrote: Hi Zhennan there is an option to mri_ca_label to read in a prior segmentation and iterate the MRF minimization, but unless your labels and conventions are identical to our's it's unlikely to work. You can use mri_ca_train to build the MRF and likelihoods and such if you want. cheers Bruce On Wed, 30 Jan 2013, Zhennan Yan wrote: Dear freesurfer experts, I have questions about the MRF in Freesurfer: (1) I have a T1 image and sub-cortical segmentation for it (nifti format). But the segmentation is not perfect and I would like to relabel it using the ICM algorithm as shown in fig.8 in paper: Whole Brain Segmentation: Automated Labeling of Neuroanatomical Structures in the Human Brain, Fischl et al., (2002). How can I do this using Freesurfer? (2) If I have several subjects with labels, how can I train the anisotropic nonstationary MRF prior as equation (9) in the same paper? Can I use that MRF prior elsewhere in my own program? Many thanks! Best, Zhennan The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail. ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Re: [Freesurfer] question about MRF
Dear Bruce, Thank you very much for your reply. I can convert my labels to Freesurfer labels and try the mri_ca_label now. May I ask some more questions? 1) What subjects were used to train the atlas in the current version? Were they healthy or diseased subjects? 2) How the current implementation differs from the one in your Whole Brain Segmentation (2002) paper? 3) Is there any cases, e.g. on which clinical diseases, FreeSurfer will produce less than ideal results? Thanks! Best, Zhennan Zhennan Yan Department of Computer Science Rutgers University On 1/30/2013 3:10 PM, Bruce Fischl wrote: Hi Zhennan there is an option to mri_ca_label to read in a prior segmentation and iterate the MRF minimization, but unless your labels and conventions are identical to our's it's unlikely to work. You can use mri_ca_train to build the MRF and likelihoods and such if you want. cheers Bruce On Wed, 30 Jan 2013, Zhennan Yan wrote: Dear freesurfer experts, I have questions about the MRF in Freesurfer: (1) I have a T1 image and sub-cortical segmentation for it (nifti format). But the segmentation is not perfect and I would like to relabel it using the ICM algorithm as shown in fig.8 in paper: Whole Brain Segmentation: Automated Labeling of Neuroanatomical Structures in the Human Brain, Fischl et al., (2002). How can I do this using Freesurfer? (2) If I have several subjects with labels, how can I train the anisotropic nonstationary MRF prior as equation (9) in the same paper? Can I use that MRF prior elsewhere in my own program? Many thanks! Best, Zhennan The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail. ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Re: [Freesurfer] question about MRF
1. 40 subjects, 10 young, 10 middle ages, 10 elderly (healthy) and 10 with AD. 2. Many small things, the big one (described in our 2004 paper) is that we now use a nonlinear warp 3. Yes, hydrocephalus for example is hard, although we have made recent improvements in the warping that handle enlarged ventricles much better cheers Bruce On Thu, 31 Jan 2013, Zhennan Yan wrote: Dear Bruce, Thank you very much for your reply. I can convert my labels to Freesurfer labels and try the mri_ca_label now. May I ask some more questions? 1) What subjects were used to train the atlas in the current version? Were they healthy or diseased subjects? 2) How the current implementation differs from the one in your Whole Brain Segmentation (2002) paper? 3) Is there any cases, e.g. on which clinical diseases, FreeSurfer will produce less than ideal results? Thanks! Best, Zhennan Zhennan Yan Department of Computer Science Rutgers University On 1/30/2013 3:10 PM, Bruce Fischl wrote: Hi Zhennan there is an option to mri_ca_label to read in a prior segmentation and iterate the MRF minimization, but unless your labels and conventions are identical to our's it's unlikely to work. You can use mri_ca_train to build the MRF and likelihoods and such if you want. cheers Bruce On Wed, 30 Jan 2013, Zhennan Yan wrote: Dear freesurfer experts, I have questions about the MRF in Freesurfer: (1) I have a T1 image and sub-cortical segmentation for it (nifti format). But the segmentation is not perfect and I would like to relabel it using the ICM algorithm as shown in fig.8 in paper: Whole Brain Segmentation: Automated Labeling of Neuroanatomical Structures in the Human Brain, Fischl et al., (2002). How can I do this using Freesurfer? (2) If I have several subjects with labels, how can I train the anisotropic nonstationary MRF prior as equation (9) in the same paper? Can I use that MRF prior elsewhere in my own program? Many thanks! Best, Zhennan The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail. ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
[Freesurfer] question about MRF
Dear freesurfer experts, I have questions about the MRF in Freesurfer: (1) I have a T1 image and sub-cortical segmentation for it (nifti format). But the segmentation is not perfect and I would like to relabel it using the ICM algorithm as shown in fig.8 in paper: Whole Brain Segmentation: Automated Labeling of Neuroanatomical Structures in the Human Brain, Fischl et al., (2002). How can I do this using Freesurfer? (2) If I have several subjects with labels, how can I train the anisotropic nonstationary MRF prior as equation (9) in the same paper? Can I use that MRF prior elsewhere in my own program? Many thanks! Best, Zhennan -- Zhennan Yan Department of Computer Science Rutgers University ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
Re: [Freesurfer] question about MRF
Hi Zhennan there is an option to mri_ca_label to read in a prior segmentation and iterate the MRF minimization, but unless your labels and conventions are identical to our's it's unlikely to work. You can use mri_ca_train to build the MRF and likelihoods and such if you want. cheers Bruce On Wed, 30 Jan 2013, Zhennan Yan wrote: Dear freesurfer experts, I have questions about the MRF in Freesurfer: (1) I have a T1 image and sub-cortical segmentation for it (nifti format). But the segmentation is not perfect and I would like to relabel it using the ICM algorithm as shown in fig.8 in paper: Whole Brain Segmentation: Automated Labeling of Neuroanatomical Structures in the Human Brain, Fischl et al., (2002). How can I do this using Freesurfer? (2) If I have several subjects with labels, how can I train the anisotropic nonstationary MRF prior as equation (9) in the same paper? Can I use that MRF prior elsewhere in my own program? Many thanks! Best, Zhennan ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.