I’m having an issue applying 3dDespike to multi-echo fMRI data. For parts of the data with small signal intensity - later echoes in regions of signal dropout - 3dDespike is zero-ing out the data, even when I use the -nomask option. This leads to later issues in T2* estimation in these areas. Any thoughts as to why this might be happening, and how to prevent it? I’m running on Mac OS 10.15, AFNI version info: Precompiled binary macos_10.12_local: Oct 19 2020 (Version AFNI_20.3.01 ‘Vespasian’).
I have not seen this before. Was there any processing done before 3dDespike in this case?
How does it look when you plot one time series on top of the other?
For this, you can look at the Regression Hands-On video, 5 of 5, at time 3:20 for how to plot both time series at once. Load the input to 3dDeconvolve as the Underlay, and use Dataset #N with the output from 3dDespike, and go to a voxel with 0 in the output (if you know where one is).
If you find such a voxel, I might want to get the time series for the input to 3dDespike (you can “write” it out from the afni GUI, by clicking ‘w’ with the mouse over that voxel’s time series in a Graph window).
Are these data split into separate files? In other words, if you have 3 echoes, I assume you have 3 different .nii (or .BRIK) files, each one corresponding do a different echo time, is that right? I know some conversions produce a single time series with echo times interleaved.
I haven’t applied any preprocessing to the data before 3dDespike, and I’m inputting separate files for each echo. I’ve attached an image of a timecourse from a zero-d out voxel (this is from fsleyes), and uploaded examples of input data (sub-01_task-rest_run-01_echo-5_bold.nii.gz), and outputs of 3dDespike (despike+orig.BRIK) and a subsequent 3dAFNItoNIFTI command (despike.nii), at: Dropbox - RestEcho5Data - Simplify your life
I ran the following command:
3dDespike -nomask -prefix ltddespike sub-01_task-rest_run-01_echo-5_bold.nii.gz
And there were no zeroed out voxels - this is with a version of AFNI updated yesterday. The zeroed out voxels are clear in your data, but not sure how they came about. What is the full command that was used to despike the data?
Looks like I had the -nomask flag after the input, so it wasn’t being applied. Moved the flag and it’s working now. Thanks for your help!