We are reaching out for guidance regarding motion correlation warnings in our block design fMRI study, and we hope to benefit from your expertise.
We are conducting a study using the Cyberball game, a block design paradigm consisting of 4 main blocks. Block lengths are 150 s, and between the blocks we have a 20s rating and about 10 s baseline measurement. The blocks are:
Explicit exclusion - the participant is just allowed to watch a full game
Inclusion - participant takes part in the game and tosses the ball
Explicit exclusion – the participant initially is included, but after a while excluded and just allowed to watch
Re-inclusion – the participant is again allowed to enter the full game
After conducting the first-level analysis, we reviewed the index.html output files. We observed that many subjects have severe warnings about high correlations between motion parameters and some of the task blocks, particularly the last “Re-inclusion” block. Some participants have one or two such warnings, while others have warnings on additional blocks, though this is less common. Example screenshot:
We don’t have much experience from block design tasks at out lab, especially with such long block durations.
Our questions:
Interpretation: How should we interpret these warnings about high correlations between motion parameters and task blocks?
Impact on Data Analysis: Do these correlations significantly disrupt the data analysis to the extent that we must exclude these participants?
Recommended Actions: What steps should we take to address these warnings? Are there adjustments or preprocessing techniques we can apply to mitigate these issues?
We appreciate any insights or recommendations you can provide to help us navigate this challenge.
Thank you for your time and assistance.
Best,
Andreas Löfberg
PhD student
CSAN/BKV Linköpings University
Sweden
Handling head motion in fMRI data analysis is particularly challenging. While conventional ad hoc methods in preprocessing and modeling can offer some relief, their effectiveness often varies significantly.
In your case, part of the warnings appear to be attributed to the experimental design—such extraordinary long blocks of 150 seconds make it harder to distinguish between the BOLD signal and potential slow drifts.
For the warnings related to head motion, could you share the time series plots for the head motion estimates along with the block regressors? That may help in diagnosing the issue more clearly.
Thank you for taking a look at this. Here are screenshots of the motion regressors from one subject, hope that helps!
We are experimenting with extracting only the first 30s or so from the blocks. Hopefully that may capture the part where the reaction to the task is the most intense while reducing the total amount of time for motion. But, that also reduces the number of volumes, and we have to await more participants to know if it reduces noise more than we loose signal. That may not, however, avoid the problem with slow drift between early and late blocks.
Based on the plots you shared, the amount of head motion is quite severe, and more critically, it appears to be task-induced. A more effective and proactive solution would involve improving the experimental design to minimize these issues from the outset. Given the current situation, I recommend sticking to your original plan of modeling the entire block rather than focusing on a smaller portion of each block. The high correlations among some of the regressors could result in reduced precision when estimating the BOLD response. However, if you have a sufficient number of participants, this precision issue might be mitigated to some extent.
Gang Chen
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.
