Hi Jidan, If you have 40 subjects, you should use our scripts.
Avi Snyder computed this affine transform from avg152T1 to 711-2B on 2/26/2008: 0.953903 -0.00387241 -0.0212912 -0.113526 -0.0104025 0.950932 -0.0164124 -3.41943 0.0208524 0.0550674 0.946383 2.56815 0 0 0 1 As for the the core 6 landmark guidelines being too precise, this judgment is in the eyes of the beholder. In fact, Harold Burton (whose lab is one floor up from ours) decided the guidelines were too lax, so he and Donald McLaren created a version that is even more anal (http://brainvis.wustl.edu/help/burton_manuals/LANDMARKS_DGM_revised4.doc). My guess is that if you "eyeballed it" (which is what I would tend to do, left to my own devices), your results would be just fine -- provided you were consistent and avoided any systematic biases that would affect your results (e.g., drew all of group A's medial walls stingily, but all of group B's medial walls too generously). But with 40 subjects, it makes sense to use auto-landmarks, so your "eye-balling" will be limited to touching up the borders that need edits. While I hoped to whip these scripts into shape this week, my attention has been diverted onto something else. For now, I have put interim versions here: http://brainmap.wustl.edu/pub/donna/FREESURFER/SCRIPTS/2009_09/ login pub password download All but freesurfer2caret.Ver2a.preborder.mni2wustl.sh were developed for a project where the segmentations and surfaces were generated in Caret, so they haven't been vetted yet on a Freesurfer project. Alan Anticevic will probably do this soon. The freesurfer2caret.Ver2a.preborder.mni2wustl.sh script is the one that will need most attention, because it depends on the spatial normalization used to get the surface into 711-2B space, which the draw borders script currently requires. This tends to be study-specific, and at least two of the studies here at Wash U using scripts like this fed Freesurfer anatomicals that were already in 711-2B space, so no normalization was necessary. You can either combine your mysubject2MNI matrix with the MNI2711-2B matrix above, or apply each in sequence to your midthickness surface like so: MATRIX=`$CURDIR/MNIlin_T1_to_711-2B.mat | tr "\n" " "` caret_command -surface-apply-transformation-matrix $CASE.$HEM_FLAG.MIDTHICKNESS_MNI.coord $CASE.$HEM_FLAG.topo $CASE.$HEM_FLAG.MIDTHICKNESS_711-2B.coord -matrix $MATRIX Next week, I hope to tweak this script for a more typical Freesurfer to Caret stream (i.e., not already in 711-2B; not registered via FLIRT; use existing talairach.xfm, but then apply MNI2711-2B matrix above). When that script is done, I'll copy it to the ftp site above and post to the list, because Traci is waiting patiently. ;-) Donna On 09/29/2009 01:51 AM, z丹丹 wrote: > Hi Donna, > > Thanks for your detailed interpretation. In fact I have 40 subjects, I > did try to draw the sulci manully, which took me a lot of time for > even one subject.... So I'm thinking I should use the automated way to > generate the sulci. > > At first,I have the freesurface generated surfaces, then I use Flirt > to do rigid transformation with the fsl template (MNI152_T1_1mm, which > is different from what you mentioned.) and apply this transformation > to my surface to put it into the MNI space. > > So, maybe I can try the automated method if I redo the > registration...Could you tell me where the affine transformation to > get from MNI space to 711-2* is? > > Another question...I read your description for the 6 landmarks. Is it > neccessary to follow the rules you defined there? I mean, the > description like "the exact 19 millimeters to the anterior end ", is > really too precise. If we don't do it like that, but we still > delineate the sulcus consistently across subjects, what would happen? > (e.g. we just delineate it 9 mm at the end across subjects? ) > > Thanks for your reply!!! > > Jidan > > > From: [email protected] > > Subject: caret-users Digest, Vol 72, Issue 10 > > To: [email protected] > > Date: Mon, 21 Sep 2009 12:00:01 -0500 > > > > Send caret-users mailing list submissions to > > [email protected] > > > > To subscribe or unsubscribe via the World Wide Web, visit > > http://brainvis.wustl.edu/mailman/listinfo/caret-users > > or, via email, send a message with subject or body 'help' to > > [email protected] > > > > You can reach the person managing the list at > > [email protected] > > > > When replying, please edit your Subject line so it is more specific > > than "Re: Contents of caret-users digest..." > > > > > > Today's Topics: > > > > 1. Re: do spherical registration using freesurfer generated > > surfaces (Donna Dierker) > > > > > > ---------------------------------------------------------------------- > > > > Message: 1 > > Date: Mon, 21 Sep 2009 08:44:18 -0500 > > From: Donna Dierker <[email protected]> > > Subject: Re: [caret-users] do spherical registration using freesurfer > > generated surfaces > > To: "Caret, SureFit, and SuMS software users" > > <[email protected]> > > Message-ID: <[email protected]> > > Content-Type: text/plain; charset=UTF-8; format=flowed > > > > Hi Jidan, > > > > We'll set aside the question of why your flat map is really ugly, > > because the impression I get is that you don't really want/need a flat > > map. You want to register, which is one of the most typical objectives. > > > > Correct: To register a human subject to the PALS_B12 atlas, you need a > > borderproj file with the core 6 landmarks for your source subject. You > > c a n draw them yourself (see > > http://brainvis.wustl.edu/help/landmarks_core6/landmarks_core6.html) or > > use the auto-landmark feature, available via the command line > > (caret_command -surface-border-landmark-identification) or via the GUI > > (Layers: Borders: Border Operations (in Development): Create Flattening > > and Registration Landmark Borders). I recommend a Caret version than > > July, 2009, if possible, because updates were made to the algorithm > this > > summer. > > > > Note that for the auto-landmark process to work, the fiducial surface > > must be in Wash U's 711-2* stereotaxic space. For labs outside Wash U, > > this is unlikely to be the case, but if you can use a volumetric > > registration tool (e.g., FSL's flirt) to register your anatomical > volume > > to avg152T1 (a template provided in both the FSL and SPM > distributions), > > then we can provide an affine transform to get from MNI spa c e to > 711-2*. > > Then you're just a couple of "caret_command > > -surface-apply-transformation-matrix" commands away from having a > > fiducial surface in 711-2* space. > > > > If you only have one or two subjects to register, it might be easier to > > just draw them yourself; however, drawing borders on surfaces generated > > from fiducials that are rotated off AC-PC can be very tricky: > > > > http://brainvis.wustl.edu/help/why_normalize_input > > > > For example, the anterior superior temporal gyrus border (aka > > "SF_STSant" or "the pink one") starts directly under the starting > > Central Sulcus border point; however, if the chin is rotated much up or > > down, you'll need to rotate the inflated surface to roughly AC-PC > before > > mentally dropping the plumb bob down onto the aSTG. Similarly, trimming > > points from the Central Sulcus dorsal termination can be confounded if > > the surfa c e is rotated left or right. > > > > See also my recent comments to Traci Sandoval about drawing landmarks > > without a flat map: > > > > http://www.mail-archive.com/[email protected]/msg01806.html > > > > Keep in touch. These processes are in a state of flux, due to the > > auto-landmark enhancements. > > > > Donna > > > > On 09/19/2009 08:36 AM, z?? wrote: > > > Hi Donna, > > > > > > Thanks for your reply. Acturally I'm a new user of CARET and I'm > > > learning it from the tutorial. I already fix the problem of the space > > > and coordinates, and now I understand I don't need to get flat maps > > > for registration. Also I don't want to use flat map because I tried > > > two subjects, the flat map i got is really ... ugly... > > > > > > But in this way, if I only have the fiducial surface, spherical > > > surface, I need to draw a l l the 6 landmarks manually, right? > > > With te 6 landmarks I draw, I can do the spherical registration. Is > > > this a right way to do it? > > > > > > Thanks. > > > > > > Jidan > > > > > > > > > > > > > > > > > > > > Message: 2 > > > > Date: Thu, 17 Sep 2009 08:43:45 -0500 > > > > From: Donna Dierker <[email protected]> > > > > Subject: Re: [caret-users] do spherical registration using > freesurfer > > > > generated surfaces > > > > To: "Caret, SureFit, and SuMS software users" > > > > <[email protected]> > > > > Message-ID: <[email protected]> > > > > Content-Type: text/plain; charset=GB2312 > > > > > > > > Hi Jidan, > > > > > > > > See inline replies below. > > > > > > > > Donna > > > ; > > > > > On 09/17/2009 02:06 AM, z?? wrote: > > > > > Hi, > > > > > > > > > > I already have fiducial surfaces generated by Freesurfer, > which are > > > > > all in the MNI space. I want to do spherical registration to > an atlas > > > > > spherical surface. The steps i need to do are: > > > > > > > > > > 1, generate the ellipsoid, spherical surfaces, and flat maps. > > > > > 2, create the 6 landmarks on both i ndividual and atlas surface. > > > > > 3, do spherical mapping. > > > > > > > > > > Are the procedure right? Thanks! > > > > Funny you should ask. We (the Van Essen Lab) are using > Freesurfer for no > > > > less than five projects, and we have shell scripts that > streamline much > > > > of the import, border drawing, QA, and registration. There is a > feature > > > > we call "auto- l andmarks" that draws the borders reasonably > well. Many > > > > borders need touch-ups, but a single rater fixed 72 hemispheres' > borders > > > > in 1.5 days. If you have only a handful of subjects, it's not clear > > > > whether it would be worth it for you to use our methods. But if > you have > > > > dozens, then they would save you a lot of time. > > > > > > > > We haven't broadly publicized these features/scripts, because > they're > > > > still in the final stages of development and testing. But they > have been > > > > working well for us. > > > > > > > > Even if you don't use auto landmark s or these scripts, one question > > > > stands out in your steps above: Do you need a flat map? No, not > really. > > > > It certainly is easier to draw landmarks on the flat map than the > > > > sphere; however, I'm not sure it's worth the hassle. Then again, > if < BR>> > > you're not using the scripts, it might be the shortest > path to > > > registration. > > > > > In fact before I move to the 2nd step, I was stuck in the flat map > > > > > generation. I met two problems: > > > > > 1, As my surface is in MNI space, the surface orientation is > not right > > > > > in the CARET software. I found that the anterior and posterior are > > > > > reversed with superior and inferior. Is there any method to > roate it > > > > > into the right orientation in CARET? > > > > Yes, but this is not consistent with my experience of MNI space nor > > > > Freesurfer surface orientation. While MNI volumes are sometimes > stored > > > > using left-handed coordinate systems (what some call radio l ogical > > > > orientation), the coordinates are always reported with -x being > left, +x > > > > being right; -y being posteri o r, +y being anterior; -z being > inferior, > > > > +z being superior. I've never seen a Freesurfer surface with the > Y and Z > > > > axes flipped -- not even ones written in MNI space. > > > > > > > > It's easy to flip a surface from LAS to LPI > > > > (http://brainmap.wustl.edu/OLD/SureFit/orient.html) like this: > > > > > > > > caret_command -surface-apply-transformation-matrix $COORD $TOPO > > > > $FLIPPED_COORD -matrix "1 0 0 0 0 -1 0 0 0 0 -1 0 0 0 0 1" > > > > > > > > But I really would not brush this off. I would get to the bottom > of why > > > > the supposedly MNI surface is not in the expected orientation. The > > > > evidence suggests it is not in MNI space, as asserted. > > > > > 2, The other problem is , I found that the origin of my > surface is not > > > > > AC. So when I flatten the surface, it can't find the correct > media l > > > > & g t; wall to cut. Do you think there is a way to solve this too? > > > > This is more evidence that the surface is not really in MNI > space. How > > > > did it get transformed to MNI? > > > > > > > > Note that the MNI origin is not exactly the AC; see > > > > http://imaging.mrc-cbu.cam.ac.uk/imaging/MniTalairach for > details. But > > > > it is certainly close enough to the AC that the compressed > medial wall > > > > view should be centered over the medial wall. The fact that it > is not > > > > indicates a problem with your transformation to MNI space. > > > > > > > > > > Thanks a lot! > > > > > > > > > > Jidan > > > > > > > > > > > > ------------------------------ > > > > _______________________________________________ > > caret-users mailing list > > [email protected]. e du > > http://brainvis.wustl.edu/mailman/listinfo/caret-users > > > > > > End of caret-users Digest, Vol 72, Issue 10 > > ******************************************* > > ------------------------------------------------------------------------ > 把MSN装进手机，更多聊天乐趣等你挖掘！ 立刻下载！ > <http://mobile.msn.com.cn/> > ------------------------------------------------------------------------ > > _______________________________________________ > caret-users mailing list > [email protected] > http://brainvis.wustl.edu/mailman/listinfo/caret-users > _______________________________________________ caret-users mailing list [email protected] http://brainvis.wustl.edu/mailman/listinfo/caret-users
