I have a question when using PPI analysis. I'm interested in two ROIs (for example, A and B and they have been defined in ROI analysis). When I used A as the seed of PPI analysis, I found A more correlated with B in stimA than stimB. But when I used B as the seed of PPI analysis, I didn't find the same pattern. Or even worse, B was more correlated with A in stimulus 2 than in stimulus 1.
Is this asymmetric result presented after exchanging seed and ROI normal? Does this indicate that there is a problem with my PPI analysis? Thank you for help.
PPI examines the interaction between a physiological signal (i.e., the seed region’s time series) and psychological variables (such as task conditions). Since the model typically assumes a standard hemodynamic response function (HRF) to infer neuronal activity, it can be somewhat crude and limited in accurately capturing the context-dependent association between brain regions.
Is this asymmetric result presented after exchanging seed and ROI normal? Does this indicate that there is a problem with my PPI analysis?
This kind of asymmetry is quite common and does not necessarily indicate a problem with your analysis. When Region A is used as the seed, you're modeling how its activity interacts with task conditions in relation to Region B’s signal. Swapping the seed (i.e., using Region B instead) reverses that direction of interaction. The results are not expected to be conceptually identical for several reasons:
Functional asymmetry: Region A and Region B may play different roles in the task, leading to asymmetric patterns of context-dependent connectivity.
Association patterns: Even if the two regions are reciprocally related, their association profiles across the brain (and their task-modulated interactions) may differ.
So, the observed asymmetry is not necessarily unexpected. This asymmetry also reflects the way PPI is modeled. It’s important to remember:
An interaction between Region A and Region B doesn’t imply that A is driving B; it simply indicates that the relationship between A and B is modulated by the task context.
Reciprocal interactions may exist: Region B may also influence Region A, but modeling the PPI from B as the seed might reveal different aspects of this interaction, especially depending on regional processing dynamics and noise characteristics.
Thanks a lot for your detailed explanation. It's great to know this asymmetry is common. Since we focus on the two regions of interest, A and B, I don't seem to have a good reason to only take one as the seed and not look at the other. Perhaps it is the different functions of the two regions that lead to the differences in contex-dependent connectivity. All in all, I appreciate that you have inspired me with some ideas to explain the asymmetrical results.
Shulei
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.