Hello, I'm going back and running the jacobian analyses (using v4 recon-all) so that I can look at distortion arising from the spherical registration.
Couple questions about what is going on "under the hood":
1) When using the -jacobian_white flag, the command executed is: mris_jacobian ?h.white ?h.sphere.reg ?h.jacobian_white
Why the use of ?h.white rather than ?h.sphere as the "original surface"?
2) When using the -jacobian_dist0 flag (which appears to be analogous to the old -jacobian flag in the recon-all from v3) the input surface used is ?h.sphere.reg. If I understand correctly, this means that the resulting ?h.jacobian_dist0 file reflects the ADDITIONAL distortion that ensues when registering the already partially registered ?h.sphere.reg to the average convexity of the target atlas with the distance constraints turned offf. Is that concept of "additional" distortion the correct way to think about the ?h.jacobian_dist0 data?
3) In terms of Eq (5) from Fischl et al 1999 ("High-Resolution Intersubject Averaging..."): J = J_p + A*J_a + D*J_d Are BOTH constants A and D set to zero by the -dist0 flag in mris_register (or is it just D that is set to zero)?
thanks, Mike H.
Hi Mike,
1. This will include distortions in the spherical mapping as well as the spherical registration (note that global area scaling is taken out, unless you specify -n). I think this is probably what you want, since otherwise the spherical transformation could wash out interesting things or introduce artifactual ones.
2. Here the thought was that you don't want to look at the residual distortion, but all of it, by morphing without any metric constraints (just topology).
3. Just the D is set to 0. A is actually a topology preserving term and won't change the distortion much (it only affects negative definite or very close to singular areas)
cheers, Bruce
On Fri, 21 Dec 2007, Michael Harms wrote:
Hello, I'm going back and running the jacobian analyses (using v4 recon-all) so that I can look at distortion arising from the spherical registration.
Couple questions about what is going on "under the hood":
- When using the -jacobian_white flag, the command executed is:
mris_jacobian ?h.white ?h.sphere.reg ?h.jacobian_white
Why the use of ?h.white rather than ?h.sphere as the "original surface"?
- When using the -jacobian_dist0 flag (which appears to be analogous to
the old -jacobian flag in the recon-all from v3) the input surface used is ?h.sphere.reg. If I understand correctly, this means that the resulting ?h.jacobian_dist0 file reflects the ADDITIONAL distortion that ensues when registering the already partially registered ?h.sphere.reg to the average convexity of the target atlas with the distance constraints turned offf. Is that concept of "additional" distortion the correct way to think about the ?h.jacobian_dist0 data?
- In terms of Eq (5) from Fischl et al 1999 ("High-Resolution
Intersubject Averaging..."): J = J_p + A*J_a + D*J_d Are BOTH constants A and D set to zero by the -dist0 flag in mris_register (or is it just D that is set to zero)?
thanks, Mike H.
A followup on issue (2):
If I'm understanding you correctly, the -jacobian_dist0 flag and associated ?h.jacobian_dist0 file are supposed to get at the TOTAL distortion, starting at ?h.white, when you morph without any distance constraints.
If that is the intent, then I'm confused about the syntax of 'mris_register. Namely, the "input surface" used currently as part of the -jacobian_dist0 flag is ?h.sphere.reg, whereas it would seem that you would want to use ?h.white as the input surface if you are trying to capture the TOTAL distortion.
e.g., here is the actual command used (with the part of my confusion changed to CAPS):
#@# Jacobian dist0 lh Thu Dec 20 13:43:15 CST 2007 /disk/conte_raid1/freesurfer/conte_mprage/010921_6038/scripts
mris_register -curv -norot -jacobian ../surf/lh.jacobian_dist0 -dist 0 ../SURF/LH.SPHERE.REG /pkg/brainmap/freesurfer- v4.0.1/average/lh.average.curvature.filled.buckner40.tif ../surf/lh.sphere.dist0.jacobian.reg
thanks for clarifying, Mike
On Fri, 2007-12-21 at 13:46 -0500, Bruce Fischl wrote:
Hi Mike,
- This will include distortions in the spherical mapping as well as the
spherical registration (note that global area scaling is taken out, unless you specify -n). I think this is probably what you want, since otherwise the spherical transformation could wash out interesting things or introduce artifactual ones.
- Here the thought was that you don't want to look at the residual
distortion, but all of it, by morphing without any metric constraints (just topology).
- Just the D is set to 0. A is actually a topology preserving term and
won't change the distortion much (it only affects negative definite or very close to singular areas)
cheers, Bruce
On Fri, 21 Dec 2007, Michael Harms wrote:
Hello, I'm going back and running the jacobian analyses (using v4 recon-all) so that I can look at distortion arising from the spherical registration.
Couple questions about what is going on "under the hood":
- When using the -jacobian_white flag, the command executed is:
mris_jacobian ?h.white ?h.sphere.reg ?h.jacobian_white
Why the use of ?h.white rather than ?h.sphere as the "original surface"?
- When using the -jacobian_dist0 flag (which appears to be analogous to
the old -jacobian flag in the recon-all from v3) the input surface used is ?h.sphere.reg. If I understand correctly, this means that the resulting ?h.jacobian_dist0 file reflects the ADDITIONAL distortion that ensues when registering the already partially registered ?h.sphere.reg to the average convexity of the target atlas with the distance constraints turned offf. Is that concept of "additional" distortion the correct way to think about the ?h.jacobian_dist0 data?
- In terms of Eq (5) from Fischl et al 1999 ("High-Resolution
Intersubject Averaging..."): J = J_p + A*J_a + D*J_d Are BOTH constants A and D set to zero by the -dist0 flag in mris_register (or is it just D that is set to zero)?
thanks, Mike H.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Mike,
we should probably rerun mris_jacobian at the end with ?h.white as the orig and ?h.sphere.dist0.jacobian.reg as the target for consistency On Fri, 21 Dec 2007, Michael Harms wrote:
A followup on issue (2):
If I'm understanding you correctly, the -jacobian_dist0 flag and associated ?h.jacobian_dist0 file are supposed to get at the TOTAL distortion, starting at ?h.white, when you morph without any distance constraints.
If that is the intent, then I'm confused about the syntax of 'mris_register. Namely, the "input surface" used currently as part of the -jacobian_dist0 flag is ?h.sphere.reg, whereas it would seem that you would want to use ?h.white as the input surface if you are trying to capture the TOTAL distortion.
e.g., here is the actual command used (with the part of my confusion changed to CAPS):
#@# Jacobian dist0 lh Thu Dec 20 13:43:15 CST 2007 /disk/conte_raid1/freesurfer/conte_mprage/010921_6038/scripts
mris_register -curv -norot -jacobian ../surf/lh.jacobian_dist0 -dist 0 ../SURF/LH.SPHERE.REG /pkg/brainmap/freesurfer- v4.0.1/average/lh.average.curvature.filled.buckner40.tif ../surf/lh.sphere.dist0.jacobian.reg
thanks for clarifying, Mike
On Fri, 2007-12-21 at 13:46 -0500, Bruce Fischl wrote:
Hi Mike,
- This will include distortions in the spherical mapping as well as the
spherical registration (note that global area scaling is taken out, unless you specify -n). I think this is probably what you want, since otherwise the spherical transformation could wash out interesting things or introduce artifactual ones.
- Here the thought was that you don't want to look at the residual
distortion, but all of it, by morphing without any metric constraints (just topology).
- Just the D is set to 0. A is actually a topology preserving term and
won't change the distortion much (it only affects negative definite or very close to singular areas)
cheers, Bruce
On Fri, 21 Dec 2007, Michael Harms wrote:
Hello, I'm going back and running the jacobian analyses (using v4 recon-all) so that I can look at distortion arising from the spherical registration.
Couple questions about what is going on "under the hood":
- When using the -jacobian_white flag, the command executed is:
mris_jacobian ?h.white ?h.sphere.reg ?h.jacobian_white
Why the use of ?h.white rather than ?h.sphere as the "original surface"?
- When using the -jacobian_dist0 flag (which appears to be analogous to
the old -jacobian flag in the recon-all from v3) the input surface used is ?h.sphere.reg. If I understand correctly, this means that the resulting ?h.jacobian_dist0 file reflects the ADDITIONAL distortion that ensues when registering the already partially registered ?h.sphere.reg to the average convexity of the target atlas with the distance constraints turned offf. Is that concept of "additional" distortion the correct way to think about the ?h.jacobian_dist0 data?
- In terms of Eq (5) from Fischl et al 1999 ("High-Resolution
Intersubject Averaging..."): J = J_p + A*J_a + D*J_d Are BOTH constants A and D set to zero by the -dist0 flag in mris_register (or is it just D that is set to zero)?
thanks, Mike H.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Michael Harms