Max correlation at initial corrout volume results in 0 R2, lagstrength, and lagtime.

[ahmedhumza94]

Using singularity version: rapidtide_1.9.0.sif

I run rapidtide2x while specifying corrmask, refineinclude, and nofitfilt flags. The same mask is used for corrmask, and refineinclude. Full command line options example below:

/usr/local/miniconda/bin/rapidtide2x /scratch/ahmedha/BOLD/rapid_troubleshoot/231225/epi_folder/231225_registered_toMni_epi.nii.gz /scratch/ahmedha/BOLD/rapid_troubleshoot/231225/epi_folder/output --regressor=/scratch/ahmedha/BOLD/rapid_troubleshoot/blockTimes225.txt -N 0 -r -10,140 -f 3 -F 0.005,0.05 --lagmaxthresh=40 --ampthresh=0.2 --pca --noglm --nofitfilt --nprocs=1 --passes=3 --corrmask=/scratch/ahmedha/BOLD/rapid_troubleshoot/average305_t1_tal_lin_mask.nii.gz --refineinclude=/scratch/ahmedha/BOLD/rapid_troubleshoot/average305_t1_tal_lin_mask.nii.gz

The outputs looked to be masked correctly, however there are several voxels within the mask that look to have odd behavior.

In these voxels lagstrengths, lagtimes, and R2 look to have values of 0. When looking at the corrout of these voxels it looks like the maximum cross correlation occurs at the first volume. Example in screenshots below:

Is this expected behavior?

[bbfrederick]

There's two things going on here. The first is that yes, the largest value is at the edge of the lag range. The fitting routine will only fit peaks if it can fully resolve the peak - if the correlation value is still growing at the edge of the range, it will fail to find a peak. The way to resolve this is to open up the lag range on that side - perhaps go to -15 instead of -10.

The other issue is that it looks like there is a very strong, pretty distinct negative amplitude peak at what I'm guessing is 30s delay. That could be a real, arterial peak (since arterial blood has negative correlations with the global mean signal). However that's an awfully long delay for arterial blood. What does the lag map look like?

[ahmedhumza94]

To confirm the first issue won't be resolved with the nofitfilt option? There is also an option softlimit which we haven't tried yet, but looks to be available from the command line help. Would this be a better approach than extending the lagrange?

Here is the lagtimes map with a histogram across the brain. We generally see a peak delay between 25-30sec due to our gas challenge experiment. We have tried looking at this in the past outside of rapidtide since the toolbox focuses. on peak "positive" correlations. Is there a way to explore this within the toolbox?