Abstract
Functional brain connectivity studies typically rely on the synchronization of resting-state fMRI signals, often assuming stable relationships over time. However, recent evidence suggests that the strength and spatial distribution of these connections are dynamic. This study investigates how a third brain region modulates the connectivity between two other regions using the physiophysiological interaction (PPI) technique. By identifying eight brain networks and conducting voxel-wise PPI analysis, we demonstrate that regions within the same functional system generally exhibit positive modulatory interactions, while competing networks—specifically the Default Mode Network (DMN) and the Executive Network—show mutual negative modulation. These findings validate PPI as a tool for exploring the complex, non-linear dynamics of resting-state functional connectivity.
Introduction
In neuroimaging, functional connectivity (FC) is defined as the temporal correlation between brain regions during rest. While traditionally viewed as a static property, it is now understood that these relationships fluctuate. The mechanism driving these fluctuations remains a key question: is connectivity between two regions (A and B) regulated by a third region (C)? This study explores "modulatory interactions," where one brain region acts as a regulator for the communication between others.
Methods
The researchers adapted the Physiophysiological Interaction (PPI) method for resting-state fMRI (rs-fMRI). Traditionally used in task-based paradigms, PPI here identifies how the activity of one region modulates the connectivity between two others.
The study utilized the Beijing_Zang dataset (1000 Functional Connectomes Project), split into a Discovery Sample (n=64) and a Replication Sample (n=64). After identifying eight intrinsic networks via Spatial Independent Component Analysis (ICA), two representative seed regions (ROIs) were selected for each. A regression model was then applied where voxel activity was predicted by the interaction of the two seed regions' signals. Crucially, a deconvolution algorithm was used to estimate underlying neural activity from the BOLD signal, ensuring more robust results.
Key Results and Insights
The analysis revealed distinct patterns of modulation across the brain's functional architecture, which were consistently observed in both the discovery and replication samples.
Synergy and Competition
Most positive modulatory interactions occurred within the same functional systems. For instance, in the visual (extrastriate) network, additional visual regions enhanced the connectivity between primary seeds, suggesting a cooperative internal dynamic.
Conversely, the study found significant negative modulation between "competing" networks. Specifically, activity in the Executive Network was associated with decreased internal connectivity within the DMN, and vice versa. This provides a mechanistic basis for the well-known anti-correlation between these two systems.
Attention Network Dynamics
Modulatory interactions were also observed in the superior parietal lobule and fusiform gyrus for the dorsal attention network. These findings indicate that third-party regions within the attention system help coordinate the synchronization of its primary nodes.
The Role of the Caudate
An interesting finding was the role of the left caudate body, which showed positive modulatory interactions with the DMN. This suggests that the basal ganglia may play a regulatory role in coordinating DMN activity.
Methodological Validation
The use of deconvolution was validated as it produced more statistically significant and robust clusters compared to raw BOLD signals, despite a high correlation between the two methods ($r \approx 0.6$). This confirms that brain connectivity is fundamentally non-linear and non-additive.
Discussion and Limitations
While PPI effectively identifies modulatory relationships, it does not establish causal directionality—it cannot distinguish which region is the "modulator" versus the "modulated." Future research could expand this work using connectome-wide approaches to avoid ROI selection bias and investigate how these modulatory "volume knobs" shift during cognitive tasks compared to the resting state.
Summary updated