i am tring to do the resting-state fmri analysis. i did the preprocessing step using afni_proc.py then i did the first level analysis to detect the DMN network but when i underlay the resulted anatomic dataset and overlay the resulted the correlation maps i get a correlatin outside the brain borders(The areas outlined in blue).
Here is the afni proc py used in the preprocessing:
-blocks despike tshift align tlrc volreg blur mask
-align_opts_aea -cost lpc+ZZ -giant_move
-mask_import Tvent template_ventricle_2.5mm+tlrc
-mask_intersect Svent CSFe Tvent
-regress_ROI_PC Svent 3
-regress_make_corr_vols WMe Svent
-regress_apply_mot_types demean deriv
That is a good thing to check. There is even an automatically generated image of seedbased correlation for the DMN in the afni_proc.py generated QC HTML. From your output/results directory, copy+paste:
afni_open -aw QC_*/index.html
… and jump to the “vstat” section. How does that look?
The BOLD effect can be observed outside the brain (meaning here, the final anatomical volume or template brain used for alignment) for various reasons:
- B0 inhomogeneity geometrically stretches/compresses/distorts the EPI volume, so it doesn’t perfectly match the subject anatomical
→ check this in the ‘vorig’ section of the QC HTML
- alignment can be imperfect between EPI-anatomical, and/or anatomical-template
→ check this in the ‘ve2a’ and ‘va2t’ sections of the QC HTML, mentioned above
- the BOLD information measured with FMRI does not measure neuronal activity, but instead blood oxygenation affecting the magnetic field. Blood vessels around the brain can affect the magnetic field greatly (because there some big ones there), so one can see some effects occurring outside the brain at times
- there can be noise—well, there is noise in the data
- if the EPI data is blurred during processing (as in your case), some signal from inside the brain can get smeared outside a bit
That all being noted, the amount of correlation leaking/overlapping the brain there doesn’t seem too terrible to me. But it would be good for you to verify features of your data and processing, noted above.
And blood vessels can carry BOLD signal from inside the brain to outside of it. So any blood vessel departing area of activity will likely show activation along the vessel, including as it goes outside of the brain.
thank you for you valuable information. i attached the “vstat” section. how can we comment this incident?
To me, this looks like a reasonable single-subject seedbased-correlation pattern for a PCC-placed seed (default mode network), with standard resting state data.
The QC HTML there will show piece-by-piece useful aspects of processing to verify–e.g., the alignment images just preceding the vstat–so check everything make sure that all is well (or at least that there aren’t glaring issues…).