Releases: RIKEN-BCIL/HCPstyle-BOLDLagMappingAndCleaning
Lag mapping & deperfusioning for HARP protocol
niimath is recommended over fslmaths/fsl5
https://github.com/rordenlab/niimath
Runs on MATLAB
FSL needs to be installed in addition to SPM12 for flexible volume smoothing
Lag mapping for non-human fMRI
Email me for help if you encounter any issues: aso.toshihiko@gmail.com
Usage:
drLag4Drev7_longTR_monkey( 'test', 2.0,'fMRI2mm.nii.gz',5,0,1,4)
No cerebrum masking for extracting the initial seed timeseries
Tracking resolution is set to 0.5 sec irrespective of the TR of the acquision.
Make sure you have sufficient voxel-wise SNR in the high freq component. Please do:
niimath fMRI1mm -subsamp2offc fMRI2mm
for example, because you don't want to 1) be running out of memory and 2) fail to fill the voxels with reliable lag values.
If you want highly realible lag maps with a short scanning time (say 10 min), make it low resolution!
Revision 7
Email me for help: aso.toshihiko@gmail.com
Now the scripts use niimath if it exists. If not, you must specify the path to fslmaths/FSL5.
Updated to work without any previous subfunctions.
Example usage:
- Modify Line 56 to capture your fMRI directories and run
Einsteining_v05( 'Sbj0001', '0.72', '1200', '9', '0.3', '0', '8', '0')
or Einsteining_v05( 'Sbj0001', 0.72, 1200, 9, 0.3, 0, 8, 0)
function Einsteining_v05( Sdir, TR, Nvols, MaxLag, MinR, Fixed, Sm, onlyLag)
% Sdir: Subject directory name
% TR = in seconds
% Number of volumes per run
% MaxLag = 8 -> lag mapping up to -8TR / +8TR
% MinR = 0.2 -> Cross-correlogram peak < 0.2 will not be used (leaving such voxels as NaN)
% Fixed = 0 -> Recursive (flexible) LFO instead of Fixed-sLFO algorithm
% Spatial smoothing at 8mm FWHM
% onlyLag = 0 -> Lag mapping + deperfusioning (creating clean 4D dataset without perfusion lag structure)
Revision 6
Report errors / corrections to aso.toshihiko@gmail.com.
Lag mapping & deperfusioning (Einsteining) Revision 6, 2023.
Improvements based on test-retest analyses for better reproducibility including smoothing with preservation of NaN voxels and skipping variance normalization to minimize the contribution of small-amplitude runs, etc.
Use fsl5 version if you have problems with niimath. You only need fslmaths from FSL5 because newer versions doesn't support subsamp2 on 4D images.