Hi,

In order to change the atlas (35 entries) by subdividing the mediotemporal into N rois, I tried first to manipulate the atlas annotation file, I constructed the gcs file and apply the mris_ca_label, the parcellation (plotted on inflated) is good but quite different from the one obtained using the original rh.curvature.buckner40.filled.desikan_killiany.gcs:

These are the command I used to construct and use my gcs (theoretically the same that rh.curvature.buckner40.filled.desikan_killiany.gcs)

mris_ca_train -t ./average/colortable_desikan_killiany.txt rh /usr/local/freesurfer/subjects/fsaverage/surf/rh.sphere.reg /usr/local/freesurfer/subjects/fsaverage/label/rh.aparc.annot fsaverage mydesik.gcs

mris_ca_label chuv44 rh ../surf/rh.sphere.reg /usr/local/freesurfer/mydesik.gcs ../label/rh.saaparcssmat.annot

In fact the output lines are :

$Id: mris_ca_label.c,v 1.16.2.1 2006/05/02 21:41:01 nicks Exp $

$Id: mrisurf.c,v 1.441.2.5 2006/11/29 20:28:16 nicks Exp $

reading atlas from /usr/local/freesurfer/mydesik.gcs...

reading color table from GCSA file....

average std = 0.1   using min determinant for regularization = 0.000

0 singular and 167 ill-conditioned covariance matrices regularized

input 0: MEAN CURVATURE, flags 0, avgs 5, name mean_curvature

labeling surface...

relabeling using gibbs priors...

000:      0 changed, 155533 examined...

000: 87 total segments, 52 labels (341 vertices) changed

001: 37 total segments, 2 labels (61 vertices) changed

002: 35 total segments, 0 labels (0 vertices) changed

10 filter iterations complete (10 requested, 126 changed)

writing output to ../label/rh.saaparcssmat.annot...

writing colortable into annotation file...

classification took 0 minutes and 56 seconds.

While when I use directly the available gsc file it seems that there are much more iterations skipped in the command above:

mris_ca_label chuv44 rh ../surf/rh.sphere.reg /usr/local/freesurfer/average/rh.curvature.buckner40.filled.desikan_killiany.gcs ../label/rh.saaparc.annot

$Id: mris_ca_label.c,v 1.16.2.1 2006/05/02 21:41:01 nicks Exp $

$Id: mrisurf.c,v 1.441.2.5 2006/11/29 20:28:16 nicks Exp $

reading atlas from /usr/local/freesurfer/average/rh.curvature.buckner40.filled.desikan_killiany.gcs...

reading color table from GCSA file....

average std = 35.0 0.3   using min determinant for regularization = 0.011

0 singular and 829 ill-conditioned covariance matrices regularized

input 1: CURVATURE FILE, flags 0, avgs 0, name sulc

input 0: MEAN CURVATURE, flags 0, avgs 5, name mean_curvature

labeling surface...

relabeling using gibbs priors...

000:   3991 changed, 155533 examined...

001:    950 changed, 16590 examined...

002:    204 changed, 5127 examined...

003:     49 changed, 1234 examined...

004:     15 changed, 308 examined...

005:     11 changed, 96 examined...

006:      5 changed, 54 examined...

007:      3 changed, 32 examined...

008:      1 changed, 15 examined...

009:      1 changed, 8 examined...

010:      0 changed, 6 examined...

000: 111 total segments, 76 labels (659 vertices) changed

001: 39 total segments, 4 labels (51 vertices) changed

002: 35 total segments, 0 labels (0 vertices) changed

10 filter iterations complete (10 requested, 32 changed)

writing output to ../label/rh.saaparc.annot...

writing colortable into annotation file...

classification took 2 minutes and 53 seconds.

Do you know why the parcellations are quite different (between the original gcs file and the constructed gcs file from the original annotation file)? Why there are some steps skipped in mris_ca_label with my gcs file? And why there are ‘0 changed’ in gibbs step?

Thank you very much for your help

-----Original Message-----
From: Bruce Fischl [mailto:fischl@nmr.mgh.harvard.edu]
Sent: lundi, 5. mars 2007 21:22
To: Leila BH
Cc: freesurfer@nmr.mgh.harvard.edu
Subject: Re: [Freesurfer] use modified annotation file for parcellation

Hi Liela,

the .gcs file is constructed by training a classifier on multiple

individual .annot files using mris_ca_train. And it is definitely needed

for the

auto-parcellation.

cheers,

Bruce

On Mon, 5 Mar 2007, Leila BH wrote:

> Hello,

>

> I modified the annotation file rh.aparc.annot to 'rh.myaparc.annot' I can

> read it with read_annotation.m and it seems correct. (The modification is

> simply a subdivision of the middletemporal to 32 subregion named :

> middletemporal_1 .middletemporal_32).

>

>

>

> I copied this 'rh.myaparc.annot' file to

> /usr/local/freesurfer/sibjects/fsaverage/label/rh.myaparc.annot (I did this

> because the original 'rh.aparc.annot' file was there.

>

>

>

> When I run the mris_ca_label as done in the log file generated by recon-all

> command:

>

> mris_ca_label chuv44 rh ../surf/rh.sphere.reg

> /usr/local/freesurfer/subjects/fsaverage/label/rh.myaparc.annot

> ../label/rh.myaparc.annot

>

>

>

> I have an error:

>

> GCSAread(/usr/local/freesurfer/subjects/fsaverage/label/rh.myaparc.annot):

> file magic #28002 != GCSA_MAGIC (ababcdcd)

>

> mris_ca_label: could not read classifier from

> /usr/local/freesurfer/subjects/fsaverage/label/rh.myaparc.annot

>

>

>

> How can I do to use this annotation file? Is it necessary to have GCS file

> for parcellation? If yes how can I create it with this annotation file. I

> don't want to create my own atlas, I want to use the freesurfer atlas (35

> entries) but just divide some of its regions.

>

> If needed, I can send you my annotation file.

>

>

>

> Thanks a lot

>

>

>

>

--

No virus found in this incoming message.

Checked by AVG Free Edition.

Version: 7.1.413 / Virus Database: 268.18.7/710 - Release Date: 04.03.2007