Re: [caret-users] posting
Hi Jason, See the last paragraph in this post: http://brainvis.wustl.edu/pipermail/caret-users/2006-April/000725.html If you run into problems/questions, let us know. Dear Donna or David, We are now attempting to generate surface statistics. As a first step we need to generate a single multi subject metric file that has many columns (with each individual column representing a single subject's values). Can you help us out as to how to go about generating such a multi-subject metric file? Many thanks, Jason. Jason D. Connolly, PhD Center for Neural Science, New York University 6 Washington Place Room 875, New York, NY 10003 cell:646.417.2937 lab:212.998.8347 fax:212.995.4562 http://www.psych.nyu.edu/curtislab/people/jasonconnolly.html ___ caret-users mailing list caret-users@brainvis.wustl.edu http://pulvinar.wustl.edu/mailman/listinfo/caret-users
Re: [caret-users] posting reply for Donna
Jason: See inline replies below. On 04/19/2006 04:40 PM, Jason D Connolly wrote: Dear Donna, I have uploaded the .spec file containing the black inflated, fiducial and spherical left hemisphere datasets. The file is a tar.gz format labelled fordonna. Thank you again for your help. We are trying to deform the sulcal depth maps of subjects left hemispheres to the PALS atlas. Again, following spherical deformation, the left - but not right - fiducial, inflated and spherical hemispheres appear black, as shown in the screen captures that are appended in jpg format. A picture is worth a thousand words. Try this when viewing these same surfaces: Surface: Normals: Flip Normals. I'll bet the problem disappears. How your normals got flipped is another question; I don't know the answer, but it happens now and then. These were Caret-generated surfaces, right? I can't replicate the problem on my end; when I select only these files, the surface looks fine, other than the topology issue (see green ID node in attached capture): deformed_Human.6.L.Fiducial.2006-02-13.64283.coord Human.sphere_6.73730.topo deformed_Human.6.L.surface_shape_file_24.2006-02-13.64283.surface_shape One important point is that when you click on the deformed lateral view, you get a medial view, so something is wrong here. Solve this problem by entering this command in the directory where the spec is located: echo hem_flag left deformed_Human.6.L.spec It would be nice if Caret inserted the hem_flag into the deformed spec file, but to be honest, we don't use these deformed spec files much. Instead, we create composite files of the deformed*shape and deformed*metric, and then use an analysis/visualization spec. There is a dorsal view of the deformed depth fiducial map; a dorsal view of the deformed sphere, and the medial view of the deformed inflated surface. As you will see, the appended and deformed flat maps look fine. Indeed, only one CLOSED topo file is loaded. The CLOSED topo file is: CLOSED Human.sphere_6.73730.topo The fiducial, inflated and spherical coord files are: deformed_Human.6.L.Fiducial.2006-02-13.64283.coord deformed_Human.6.L.Inflated.2006-02-13.64283.coord deformed_Human.6.L.SPHERE_ALIGN.2006-02-13.64283.coord We have tried troubleshooting this by performing this deform on other subjects left hemispheres, and the results are the same (so it cannot be due to a bad subjects dataset. We have also tried this on the right hemisphere using the same target atlas, so the atlas files are not corrupted. Finally, we have tried this on both the Linux and Mac platforms and we get the same results, i.e. bad left - but not right - hemisphere images. It is almost certainly a case of flipped normals, and my guess is that the impact is 0, other than the visualization anomaly that is easily corrected. If you are using the atlas target dataset linked here, then you're using what we're using, and your deformed shape files are fine: http://brainvis.wustl.edu/help/landmarks_core6/landmarks_core6.html What I need to add, though, are links to the visualization and analysis specs (e.g., including the latest/greatest average fiducials and the distortion metric files needed to run processes like Attributes: Surface Shape: Find Significant Cluster). Best, Jason. Jason D. Connolly, PhD Center for Neural Science, New York University 6 Washington Place Room 875, New York, NY 10003 cell:646.417.2937 lab:212.998.8347 fax:212.995.4562 http://www.psych.nyu.edu/curtislab/people/jasonconnolly.html ___ caret-users mailing list caret-users@brainvis.wustl.edu http://pulvinar.wustl.edu/mailman/listinfo/caret-users -- Donna L. Dierker (Formerly Donna Hanlon; no change in marital status -- see http://home.att.net/~donna.hanlon for details.) inline: fiducial.jpg
Re: [caret-users] posting - precentral sulcus as a landmark?
Hi Jason, Donna's email nicely laid out our lab's perspective on several important issues regarding cortical landmarks and registration. I will amplify on a few aspects here. The issues that arise in considering whether the precentral sulcus may be a useful landmark are linked to the nature of individual variability and the relationship between cortical function and cortical folding. We wrestle with these issues regularly. While we don't claim to have definitive answers we do think that it is important to promote a stronger dialog among individuals and groups interested in the issues. The challenges regarding the precentral sulcus can indeed be illustrated by inspection of individual sulcal depth maps from the PALS_B12 atlas. This can be done by downloading the datasets and viewing in Caret, as Donna described, but it can also be done using WebCaret without downloading data, as shown below. The attached figure (if it gets past the filter) makes several important points, using dorsal views of the first four cases (B1.R - B4.R) from the atlas. A precentral sulcus is discernible in all 4 cases. However, in Cases 2 and 3 (middle 2) there is a big gap between superior and inferior precentral sulci. Moreover, the individual maps differ markedly in how the precentral sulcus terminates on the medial side (blue arrows). You can make educated guesses as to how the landmarks should be drawn in each case in order to maximize the correspondence of geographical and/or functional domains, but in our view it is very difficult to establish clear and objective criteria that have a good chance at improving the alignment of functional subdivisions across individuals. Similar to what Donna proposed, you might consider starting by drawing contours for the fundus of the precentral sulcus in a group of hemispheres from our original PALS_B12 dataset (e.g., the four cases below). These landmarks from individual hemispheres can then be registered to the PALS atlas without doing a complete new registration - by using the appropriate deformation_map file (accessible from SumsDB; we can provide more detailed advice on how to do apply the deformation). Once you have done this, you can inspect how variable the position of the precentral fundus is by the standard PALS_B12 registration (i.e., using the Core6 landmarks). If, on closer review, you still suspect that imposing the precentral sulcus as an additional landmark would reduce individual variability in maps of fMRI data (or other spatial data), then it is worth doing a full-blown analysis using the approach Donna outlined. If you do go down this road, we definitely urge that you do the side-by-side comparisons of the outcome of Core6 vs Connolly7 landmarks. It won't be a whole lot of extra work, and it will be the only way to ascertain whether the extra landmark helps or hinders in terms of reducing overall variance. Without parallel evaluations of this type, it will be difficult for us and others to assess the relative merits of different landmark-selection options. I remain curious as to what types of experimental data you are trying to analyze and why you think that the precentral landmark might be a useful one to add. Hope this helps. David attachment: PALS_Cases1-4_Dorsal_SulcDepth.jpg inline: PALS_Cases1-4_Dorsal_SulcDepth.jpg On Apr 7, 2006, at 2:52 PM, Donna Dierker wrote: Hi Jason, Dr. Van Essen and I just discussed your question, and we'd like more information about what you're trying to do. Often, there are easier ways to accomplish goals than posters envision. But to answer your question directly, no -- there are no published materials for adapting the core6 landmarks. (We typically use more landmarks for nonhuman primates, for whom variability is much less of an issue. I often refer users to the old Caret 4.6 Tutorial Part II, tutorial 8, Register Flat Maps, because I think it's a useful reference -- even though we don't use the flat map registration as a method anymore.) In principle, it's possible to add landmarks. In general, you would need to do this for each of the 24 PALS_B12 hemispheres (12 left, 12 right): 1. Identify the landmark on the fiducial surface. 2. Draw the border (probably easiest on the PALS flat surface, using the deformed folding as a shape underlay; we don't believe a composite folding file is available in sumsdb, but we have them). 3. Project the border. 4. Load the PALS_B12 spherical surface, and save the border relative to that surface. 5. Average the resulting borders to create an average PreCes target border for each hemisphere; add this border to the Core6 landmarks to generate a new landmark set -- let's call it Connolly7. Before going down this path, we think you would benefit from a warm- up exercise that might help you understand why the PreCes is not a core
