Denis Schluppeck, 2025-03-20, help for DAFNI students
... on how to get from coordinates reported in MM (often with negative numbers, because images are centered on 0,0,0) to VOXEL coordinates in SPM.
The matlab script test_coords.m included below shows you how you can go about this. It should work with minor modifications. See also comment on this gist for where in SPM you can find info about locations of voxels.
| %% Some hints on getting timeseries information | |
| % with SPM data | |
| % | |
| % ds 2025-03-20 ... wrote it for DAFNI questions... | |
| %% the following shows code in principle. | |
| % you may have to adjust paths a bit to reflect | |
| % where your data files, spm.mat, etc are stored. | |
| %% Load the transformation matrix | |
| % info is stored in this MAT file | |
| load('spm.mat'); | |
| SPM | |
| %% load timeseries data... | |
| % load in the NIFTI file with the timeseries | |
| % you can check what is was for you here... | |
| SPM.xY.VY(1).fname | |
| % for me on my machine that file was up one level and | |
| % still had this long ugly filename... | |
| fname = '../swrMSC_COG_NEURO_03_WIP_visual_fmri_20250206110123_401.nii'; | |
| % read into matlab workspace | |
| yData = niftiread(fname); | |
| %% Get the transformation matrix | |
| % say the coordinates come from the spm maps... | |
| % e.g. the third contrast (faces > objects) is stored here.. | |
| % descrption | |
| SPM.xCon(3).Vspm.descrip | |
| % the spatial transform | |
| transform = SPM.xCon(3).Vspm.mat | |
| %% Define the coordinates / in MNI space ... | |
| % THIS IS where you'd put the info from your actual coordinates | |
| mni_coords = [-16.3; -74.9; -25.7]; % in units of MM | |
| % e.g. -16.3 -74.9 -25.7 (MNI) | |
| % should go to [48.2 19.5 23.1] in voxels... | |
| % read off the GUI in SPM window | |
| vox_coords = [48.2; 19.5; 23.1]; | |
| % BTW .. the statistical value for Faces > Objects | |
| % is - (MINUS!) 3.437... which means stronger response to | |
| % OBJECTS than faces...it's DARK on the gray scale image. | |
| % Apply the inverse transformation | |
| % this is a bit of linear algebra | |
| % to work out how we can find them... | |
| % they should be the same as reported by SPM! | |
| native_coords = inv(transform) * [mni_coords(:);1] | |
| %% now we can use our function to get timecourse | |
| % one gotcha is that we need to round the voxel coordinates | |
| % to the neearest integer: eg voxel coord x=23.6 is not cool! it needs to | |
| % be 27 | |
| rounded_vox_coords = round(native_coords(1:3)); | |
| % make clear what we need for returnTimecourse function | |
| x = rounded_vox_coords(1); | |
| y = rounded_vox_coords(2); | |
| z = rounded_vox_coords(3); | |
| timeseriesData = returnTimecourse(yData, x,y, z ) | |
| %% now plot - should look like faces > objects | |
| figure() | |
| plot(timeseriesData) | |
| %% we can also get the slice from the z slice? | |
| % it was the 3rd contrast on the list... so | |
| conImage = niftiread('con_0003.nii'); | |
| % get the z'th slice | |
| figure() | |
| theSlice = returnSlice(conImage, z); | |
| imagesc(theSlice) | |
| colormap(gray) | |
| caxis([-5 5]) | |
| % plot the point we got: | |
| hold on | |
| % gotcha - images use Y, X ordering, plots... use X, Y | |
| plot(y,x, 'ro', 'MarkerSize',12, 'MarkerFaceColor','red') | |
| title('location of the voxel in this slice') |
... and the time series plot using our returnTimeseries() function
... and even the slice of the contrast / statistical map applying returnSlice()
SPM window of one of the contrast images... showing where VOXEL and other coordinate information is located: