open-discussion > ISAS BIS - second level/group analysis
Showing 1-10 of 10 posts
Results per page:
Jun 15, 2016  09:06 PM | Luis Araujo
ISAS BIS - second level/group analysis
Hi ISAS experts,

I am just starting to use your tool. Thanks for the excellent tutorial/instructions you have set up in the website.

My aim is to run a group analysis with six pairs of inter/intra-ictal SPECTS. With the instructions in the website, I easily performed the individual analysis (for each patient), but now I want to proceed to a second-level analysis to identify and mark hyper and hypoperfusion areas for the whole group of patients.

I suppose I should start in the multiple image calculations option on the Data Tree Manager menu, but I am struggling to find instructions on how to do it in you website.

Could someone please point me towards the proper steps in the analysis of the group of images?
Jun 16, 2016  12:06 PM | Xenophon Papademetris
ISAS BIS - second level/group analysis
Let me think about this, it's been a while. It is a little on the messy side.
Jun 16, 2016  05:06 PM | Luis Araujo
RE: ISAS BIS - second level/group analysis
I will try to explain where I am at:

- set all the analysed patients in the tree (attachment);
- pick one of the patients to serve as reference, sent its Ictal SPECT node to space;
- clicked multiple images calculations, selected all ictal images;
- averaged to the reference in space;
- now went to the Interictal SPECT node (same patient), sent it to space;
- clicked multiple images calculations, selected all interctal images;
- averaged to the reference in space;

Now I have an Average.hdr, Stdev.hdr and Tscore.hdr for the Interictal SPECT and Ictal SPECT nodes.

What to do (considering that I had these steps right) now on?
Jun 17, 2016  11:06 AM | Xenophon Papademetris
RE: ISAS BIS - second level/group analysis
You probably want to work on the Diff-Spect images which is the direct comparison between the two images instead of the individual images as those are not normalized and cannot be compared directly easily. You probably want to do a "paired" analysis instead of averaging the individual data. I would "Save Images in Reference Space" for your results and then you can import them into Matlab/SPM etc. and do whatever other analysis you want. Other tools in BioImage Suite can also do some things like this. These aspects of the Data Tree are not the best tested (and have not been used much since it was originally written ~10 years ago so I would not try to do too much within this.)
Nov 10, 2016  06:11 PM | Luis Araujo
RE: ISAS BIS - second level/group analysis
Dear Xenophon,

Thanks for the answer, it really helped.

I found out how to export the diff-spect, however upon importing into SPM the images get rotated. Is this some problem that I have to address in SPM or there is something that I might have done wrong in the export part at ISAS?

One think is that I save one of the images into anatomic space in order to export the images, you mention saving into reference. These are the same, right?

Many thanks,

Nov 11, 2016  08:11 AM | Xenophon Papademetris
RE: ISAS BIS - second level/group analysis
Are you saving them as NIFTI or analyze? If you save them as NIFTI it should be OK. Analyze relies on implicit orientation conventions which causes trouble.
Nov 11, 2016  05:11 PM | Luis Araujo
RE: ISAS BIS - second level/group analysis
Thanks again,

The Multiple Image Calculation tool doesn't really give me the option to choose the filetype. What I can say is that it returns the diff files in .img and .hdr extensions. If I am correct, the position of the images are saved in these hdr and SPM was supposed to import this information as well... Am I wrong? Should I perform an extra step to get the position/rotation right?

Nov 13, 2016  07:11 AM | Larry Olson - Emory University
RE: ISAS BIS - second level/group analysis
I think I know the problem you are referring to. First, I will tell you how I deal with it. Your multiple image issue adds a complexity, but I will tell you also how I think you can deal with that.

I have found that different programs do actually handle all of these images differently, and the only way I have found to work with them consistently is to insist all of my steps work with Analyze images in an axial orientation. Nifty just doesn't work the same in every program even though "it should".
-First, to be clear, most of my images when I convert Dicoms to Analyze with MRIconvert or with ImageJ, are flipped on the y axis in BIS. I usually do not alter this for things like coregistrations, thresholds, etc in BIS because it doesn't matter. However for ISAS it is necessary to flip the images on the y axis to get them in the same orientation as the MNI template in order to do the analysis. So be sure your orientations match the MNI template before you start. But you will need to reorient the resulting diff images/blobs when you finish in order to see them correctly in other programs including SPM. That is tricky because the diff blob ISAS creates is actually not in the same orientation as other images even though it looks correct in BIS.
-The ISAS analysis puts the diff blob in the MNI space.  You may want to leave it at this, but more on that later. I want it in the patient space, so I put the blob in one of the viewers, bring up the Transformations dialog, select the "Inverse" transformation that ISAS made, apply the Inverse transformation to the blobs, copy Result to Image, and save as "patientBlob" or something meaningful. The blob is now an Analyze image in the patient space for the BIS/ISAS purpose.
-Now I need to be sure this new blob matches the original orientation of my starting images. It doesn't. Something BIS does creates these flipped wrongly in other programs, and simply flipping on y doesn't bring them into the orientation I would expect. I haven't fussed with this too much in BIS. The simplest way to match the orientations is to bring both my reference T1 and patientBlob into Amide where you can see how both are oriented easily, and "rotate" (right click) the patientBlob only with "invert axis" on both x and z axis. This gets the image back to the orientation I started with (and which was inverted on y in BIS). Show only the blob in Amide, Export Data Set (all visible slices, analyze format (SPM)) and this newly reoriented image then saves in the orientation I began with (before I flipped my original Analyze image converted from dicoms into BIS MNI orientation) and works fine in other programs.

You have 2 other issues:
-The images for your patient group all need to be coregistered with one another. I would not recommend choosing one of the patient images for this. Why not leave them all coregistered to MNI template? They need to be coregistered to that anyway in SPM. This is the same template image used by SPM analysis. Then this could skip bringing the images back to the patient space as I described in the last step above. You should still check that the group's blobs match the MNI template orientation in an external program like Amide. If not, fix this first. Now the MNI template and all of your blobs should be in the same orientation.
-Verify this in SPM as well. If it is not true you can always specify a different reference image for your SPM analysis. But orientation has to match as a prerequisite.
-Finally, you need to be sure the images you use in SPM have a center specified, typically at the posterior commissure. BIS does not create this or use it, but it is necessary in SPM. You can do much of the reorientation and set the center from within the SPM tools, but I won't go into that since it depends on the analysis you have set up there. If center and orientation match the MNI template used in SPM everything works fine. So, to answer your question (Should I perform an extra step to get the position/rotation right?), yes, you have to, but it depends on which images you begin with. their file format, and how you save them.
Nov 13, 2016  09:11 PM | Luis Araujo
RE: ISAS BIS - second level/group analysis
Thank you Larry,

You answer most of my doubts, the comments were very helpful.

I left them coregistered to the MNI space and this solved the problem with rotation while importing into SPM.

The problem now is how to define a center to these images in SPM. At this point, an analysis would generate valid results (even without centering the images)? I suspect the answer is no, and in this case how could someone center them at the posterior comissure, since these images are diffs that were smoothed during the process of ISAS analysis, hindering the recognition of anatomical landmarks?


Nov 14, 2016  04:11 PM | Larry Olson - Emory University
RE: ISAS BIS - second level/group analysis
> Luis,
> Glad that helped. This is now outside the scope of the bioimagesuite
> forum. I'm afraid I am rusty with SPM, although I will probably have to
> pick it up again within a few weeks. (Obviously rusty: in retrospect I
> believe the origin is supposed to be set at the anterior not posterior
> commissure). If you go to American Epilepsy Society meeting in Houston, I
> could meet you and see if I can help.
> Without more detail, here is a guess about what you can do. Open the MNI
> template in SPM in the Display module. There you can see where the
> coordinates for the origin of the built in reference MNI T1 are. Copy them
> down. If all of your images are correctly coregistered to the same
> reference, then you can use the same Display module to set the origin for
> them there. Just type in the coordinates and set it. I think you have to
> save it after. There are guides for SPM available that can help.
> Good luck.