 |
 |
I see. Thanks for your explanation! It makes more sense to me now. Shiyu Date: Mon, 7 Jul 2014 17:30:01 +0200 Subject: Re: [SPM] SPM error in normalise (estimate &write) From: [log in to unmask] To: [log in to unmask] CC: [log in to unmask] Dear Shiyu, I don't think that 8mm is to much. In our department the PET scan that we use have a max resolution of 1.4mm. Then, for SPM analysis we reslice the images to voxel sizes of 0.2mm. This images are finally scaled x10 to avoid problems with some of the functions in SPM. So, if you apply a kernel of 8mm, is the equivalent of 0.8mm in the "real" brain size, and its about half the resolution of the camera. Also, take in account that this smoothing is just for the normalization procedure, and is not altering you image. But even then, I found many papers using in rats a kernel of 8-16mm (0.8-1.6mm in real brain size) for final smoothing prior to the SPM analysis. I must say that this values may be valid for PET only, depending of your tracer and the resolution of the camera. In MRI studies, if the resolution of the camera is higher you can play with smaller numbers. By default, in human brain, the normalization step uses voxel sizes of about 2mm, and 8mm for kernel in smoothing, so I think the values are fine. Of course, all these values depend on the resolution of your camera, the procedure that you followed (scaling of the image), voxel size (iso-tropic or not), the background noise that you want to "remove", etc. You can always test with few different smoothing values, and check the final images and the results. I hope you are able to solve it. Best regards, David On Mon, Jul 7, 2014 at 3:50 PM, TangShiyu <[log in to unmask]> wrote: Hi David, Oh, I see, 1.52mm is not proper. The 1.52mm is from previous protocol for rat brain. I haven't changed it since I was focusing on the orientation problem. Thanks for pointing it out. But I'm afraid the default 8mm is too large for rats. Do you think I should use smaller value, like 4mm or 5mm? Shiyu Subject: Re: [SPM] SPM error in normalise (estimate &write) From: [log in to unmask] Date: Fri, 4 Jul 2014 08:13:30 +0200 CC: [log in to unmask] To: [log in to unmask] Hi Shiyu, Most probably the error is coming from the parameters used in the smoothing. The voxel size of the Karolinska template are 2mm, but your are using 1.52 as smoothing value. Not sure what is your reasoning for that value, but I will recommend you to try with a value of 8mm. Regards, David El 03/07/2014, a las 21:05, TangShiyu <[log in to unmask]> escribi¨®: Hi Helmut, Thanks for the suggestion! But I coregistered the structural image to the template first, so they seem to have the same voxel size and centre point location when I checked them via "Display". To make the processing time shorter, I just tried to normalise one T2 image to the T2 template. Their orientation look quite similar in Check Reg, but the error still came up. Another thing I've tried is, I first used ImCalc to create a mean image of all my cases after I manual set them in the similar orientation and location. Then I tried to coregister and normalise individual images to the mean image. Still, I got the same error in normalise. I'm so confused about this situation. Shiyu > Date: Thu, 3 Jul 2014 19:45:17 +0100 > From: [log in to unmask] > Subject: Re: [SPM] SPM error in normalise (estimate &write) > To: [log in to unmask] > > Some comments: The voxel size of your data can be different from the voxel size of the templates. As David states, make sure though that 1 mm in template space corresponds to 1 mm in world / "true" space, otherwise you have to scale your raw data via "Display". > > Your error message arises if you load several template images which don't fit well together. Maybe you've selected the rat template plus another file like the SPM T1 template? > > If your data does not correspond well to the template due to different orientations, you should get an error message like below: > > Coarse Affine Registration.. > > * - SPM8: spm_affreg ---------------------------------------------- > > There is not enough overlap in the images > to obtain a solution. > > Please check that your header information is OK. > The Check Reg utility will show you the initial > alignment between the images, which must be > within about 4cm and about 15 degrees in order > for SPM to find the optimal solution. > > ----------------------------------------- ... - .... > > Failed 'Normalise: Estimate & Write' > Error using spm_affreg>error_message (line 474) > insufficient image overlap > In file "C:\Progs\spm8\spm_affreg.m" (v4152), function "error_message" at line 474. > In file "C:\Progs\spm8\spm_affreg.m" (v4152), function "spm_affreg" at line 177. > In file "C:\Progs\spm8\spm_normalise.m" (v4621), function "spm_normalise" at line 172. > In file "C:\Progs\spm8\config\spm_run_normalise_estwrite.m" (v4152), function "spm_run_normalise_estwrite" at line 22. > > > Best, > > Helmut