Is it just those few vertices that are appear to be on the sphere? Can you send me the terminal output from the mri_label2label with the --close operation?
On 05/10/2017 12:16 PM, Hengda He wrote:
Hi Freesurfer Developers,
I'm trying to map the vertices of caudal middle frontal in one subject into the space of another subject by using mri_label2label. And I used the method mentioned in this https://mail.nmr.mgh.harvard.edu/pipermail/freesurfer/2015-November/042298.h...
We are using freesurfer-Linux-centos4_x86_64-stable-pub-v5.1.0, and I noticed the mri_label2label in v5.1 has the bug of getting holes in the mapped labels(http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg42600.html). So I am using the /usr/local/freesurfer/freesurfer_dev/bin/mri_label2label.bin from freesurfer-x86_64-unknown-linux-gnu-dev5-20160513.
For example, we have source Labels(aparc-lh-003_5thCol.label), which are all in SurfaceRAS coordinates: #!ascii label , from subject FreeSurfer vox2ras=TkReg 8 104996 -35.679 16.281 25.156 1 104997 -36.480 16.439 25.362 2 104998 -37.443 16.702 25.369 3 105705 -41.554 16.840 28.930 4 105716 -40.445 17.153 28.046 5 105720 -39.570 17.200 26.959 6 131017 -29.315 12.915 45.422 7 131030 -31.304 12.514 46.455 8
By running this: /usr/local/freesurfer/freesurfer_dev/bin/mri_label2label.bin --srclabel aparc-lh-003_5thCol.label --srcsubject FreeSurfer --trglabel aparc-lh-003_5thCol_Reg2MNI.label --trgsubject FreeSurferMNI152 --regmethod surface --hemi lh
We got mapped Labels(aparc-lh-003_5thCol_Reg2MNI.label), which are all in SurfaceRAS coordinates: 119096 -39.670 37.222 12.833 3.0000000000 117433 -44.642 35.253 14.575 4.0000000000 118250 -43.206 35.844 14.273 5.0000000000 119091 -42.114 36.692 14.044 6.0000000000 120517 -26.436 38.614 21.780 7.0000000000 121236 -25.959 39.757 24.739 8.0000000000 1321 -22.346 -82.436 -9.530 0.0000000000 121253 -26.213 39.504 23.711 8.0000000000
But the mapped labels I got from the mri_label2label in v5.1 and that from using mri_label2label.bin in dev are quite similar to each other when applying the method to around 3500 labels.
So I am wondering if the bug of getting holes was fixed by providing morphological operations in the updated mri_label2label like close operation. If it is true, I am confused about the mapped labels from mri_label2label with close operation.
For example, with the same source labels above, I ran
/usr/local/freesurfer/freesurfer_dev/bin/mri_label2label.bin --srclabel aparc-lh-003_5thCol.label --srcsubject FreeSurfer --trglabel aparc-lh-003_5thCol_Reg2MNI_close2.label --trgsubject FreeSurferMNI152 --regmethod surface --hemi lh --close 2
And we got mapped Labels(aparc-lh-003_5thCol_Reg2MNI_close2.label): 119096 -39.670 37.222 12.833 3.0000000000 117433 -44.642 35.253 14.575 4.0000000000 118250 -43.206 35.844 14.273 5.0000000000 119091 -42.114 36.692 14.044 6.0000000000 120517 -26.436 38.614 21.780 7.0000000000 121236 -25.959 39.757 24.739 8.0000000000 1321 -22.346 -82.436 -9.530 0.0000000000 121253 -26.213 39.504 23.711 8.0000000000 1045 -1.690 -91.572 -40.145 0.0000000000 118251 -18.413 75.394 63.061 0.0000000000 118258 -21.026 75.715 61.847 0.0000000000 119092 -19.797 75.786 62.165 0.0000000000
The problem is that four additional labels 1045, 118251, 118258 and 119092 are not actually on the white matter surface but seem to be right on a sphere(ex. label 1045, 1.690^2+91.572^2+40.145^2=100^2). But I checked all other mapped labels like 119096, they are on the WM surface in the SurfaceRAS coordinates.
I am using CentOS release 6.9.
Thanks!
All the Best, Hengda
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer