 |
 |
Dear On, > However, the > only caveat to that solution is that I can no longer do a 1-to-1 overlap > comparison with a separate manually segmented file that is in the > positionally-normalized-Talairach space. How about something like this. Reorient the raw T1W to match AC-PC line and make sure you remember exactly how much rotation and translation you had to apply to achieve this. Assuming segmentataion works better after setting AC-PC, and also assume that segmented image shared the same origin information (I know this is another problem for your case) as raw T1W, you can then apply the same amount of rotation and translation in the opposite direction. This step will translate and rotate the segmented image to match with your manually segmented image. Will this work? With regards, Hedok > Any suggestions or insight is quite > welcome and greatly appreciated. Thank you for everyone's help thus far! > > Best Regards, > On > > === > > On Tue, 29 Jan 2008 23:10:39 -0500, Hedok Lee <[log in to unmask]> wrote: > > >> Dear On, >> >> As you pointed out, the origin has altered after segmentation and I >> don't really know why. Could you elaborate your steps in details so >> that someone, or I(very unlikely), may be able to track down what >> happened? Posting your Matlab version and spm_preproc version would >> also help. >> >> I can come up with one idea for poor segmentataion result. Have you >> checked origin of your image using "display" and set position of the >> cursor to 0mm,0mm,0mm? It's recommenced to place the origin of the >> image near AC that is also in the same plane with PC. >> >> Hope this helps, >> >> Hedok >> >> >> On Tsang wrote: >> >>> Dear Hedok, >>> >>> Thank you for your insight. I have tried your suggestion and the results >>> confirm my suspicion: that the Native space option does not maintain the >>> same coordinate space in the output. In the following screenshots please >>> note that the origin has been altered in the segmented output of the gray >>> matter image: >>> >>> http://www.utdallas.edu/~okt061000/anat.png >>> http://www.utdallas.edu/~okt061000/gm.png >>> >>> Would you please verify that for me? Thank you so much! If there is indeed a >>> bug, how should I go about escalating the issue? Thanks again, Hedok. >>> >>> Best Regards, >>> On >>> >>> === >>> >>> On Tue, 29 Jan 2008 16:55:03 -0500, Hedok Lee <[log in to unmask]> wrote: >>> >>> >>> >>>> Dear On, >>>> >>>> Just to add. When I wrote, "These number are typically all the same >>>> between segmented images and raw T1W.", I meant segmented images in >>>> native space. This is no longer true in warped images(modulated and >>>> unmodulated). >>>> >>>> With regards, >>>> >>>> Hedok >>>> >>>> >>>> Hedok Lee wrote: >>>> >>>> >>>>> Dear On, >>>>> >>>>> Somebody who is more knowledgeable about SPM may be able to help you >>>>> better, but I give it a try. >>>>> >>>>> I typically check image tags to ensure raw T1W and native images are >>>>> in the same space. If you right click an image in the "checkreg" >>>>> function, it pulls out the menu with bunch of nice features. When you >>>>> click on the top of the menu, it also displays information such as >>>>> Rotations >>>>> Origin >>>>> Voxel size >>>>> Image dimensions. >>>>> >>>>> These number are typically all the same between segmented images and >>>>> raw T1W. Can you check if they are the same? In my experience, they >>>>> are always the same. If they are the same, it's possible that >>>>> something went wrong during segmentation. See the attached screenshot. >>>>> >>>>> Hope this helps, >>>>> >>>>> >>>>> Hedok >>>>> >>>>> On Tsang wrote: >>>>> >>>>> >>>>>> Hello SPMers, >>>>>> >>>>>> I have another question concerning the SPM segmentation algorithm, >>>>>> spm_preproc. This question is rather basic and quite fundamental. >>>>>> Essentially, I suspect that the algorithm does not actually retain the >>>>>> Native space in the output. Can anyone else confirm this? More >>>>>> specifically, >>>>>> I used the IBSR database which consists of >>>>>> positionally-normalized-Talairach >>>>>> datasets, and the output of spm_preproc is no longer in >>>>>> positionally-normalized-Talairach space. I am attaching a screenshot >>>>>> depicting GM segmentation for one such dataset. The image can also be >>>>>> found >>>>>> at the following location: >>>>>> >>>>>> http://www.utdallas.edu/~okt061000/native.png >>>>>> >>>>>> Please can anyone confirm this behavior? I would really appreciate it. >>>>>> Thanks in advance! >>>>>> >>>>>> Best Regards, >>>>>> On >>>>>> >>>>>> >>>>>> >>>>> ------------------------------------------------------------------------ >>>>> >>>>> >>>>> >>> >>> >> ========================================================================= >> > > >