Posted by Edward J. Butterworth on November 06, 2002 at 16:06:40:
In Reply to: Re: Importing registered data back into afni posted by Ziad Saad on November 06, 2002 at 15:12:58:
Here's the general plan. If you need more information, or the source code, I can supply it.
I use 3dDeconvolve to process the data, and show the function with the R^2 correlation as intensity and the F-test as a threshold. The base images are 256 x 256 T1 images, and the EPIs are 128 x 128. (From what I gather, that's a preferred pattern among afni users.) Then I import it into Matlab using BrikLoad. Matlab, of course, writes the matrices in double precision by default. Once in Matlab, I expand the EPIs out to 256 x 256, then use the Image Processing Toolbox to perform the registration. At present the fiducial points are still picked by hand. I isolate a single slice, usually one in which the tumor is visible, and use the "cpselect" function to pick about 12 pairs of corresponding points in the two images. Then I use the "cp2tform" function with the "affine" option to generate the transform. (Later I intend to develop a custom transformation.) The data is displayed, and it appears that this method gives very good registration. Next I convert the overlays to RGB and display them in color.
At this point I'd like to get the transformed data back into afni. I have not been able to get WriteBrick to work, and using to3d directly on the transformed data has its own problems. What I'd really like is to get the R^2 and F-test files back into afni, and then display them overlaid onto the T1 images, using the thresholding capacities in the "Function" GUI.
Do you see that as feasible?