Hierarchical communities of high-amplitude co-fluctuations in extremity disuse

1733 

Abstract Submission 

Youngheun Jo¹, Haily Merritt², Joshua Faskowitz¹, Richard Betzel²

¹Indiana University Bloomington, Bloomington, IN, ²Indiana University, Bloomington, IN

Youngheun Jo  
Indiana University Bloomington
Bloomington, IN

Haily Merritt  
Indiana University
Bloomington, IN

Joshua Faskowitz, Ph.D.  
Indiana University Bloomington
Bloomington, IN

Richard Betzel  
Indiana University
Bloomington, IN

Edge time series (eTS) decompose static functional connectivity into its exact time-varying contributions [1, 2]. Studies have shown that global, instantaneous, high-amplitude co-fluctuations in eTS – "events" – originate from modules of structural connectivity [3], synchronize with naturalistic stimuli [4], and improve both brain-behavior correlations [2] and subject identifiability [5].
Recently, high-amplitude region-level pulses in fMRI activation have been reported in motor areas during a disuse period of the upper extremities [6, 7]. However, it is unknown whether such regional pulses and global events are related, and whether clusters of events reflect changes in extremity disuse. Here, we investigate events and their hierarchical community structure in extremity disuse and investigate instantaneous global changes at various organizational scales [8]. We find that pulses in the motor cortices are aligned with, but do not equate to global events. Events have been found to have distinct patterns with local change in disuse and global changes with cast removal.

We detected motor cortex-specific activation peaks (a node-level measurement) and global co-fluctuation events (an edge-level measurement) using the casting dataset [6]. Specifically, we detected motor cortex-specific peaks by calculating the root mean squared (RMS) from nodal time series in only the motor regions and using only the peaks with positive activations. Events were found by using the RMS across all brain regions and thresholding the global peaks against a null circularly shifted surrogate time series [2]. Next, we applied a hierarchical clustering algorithm across all events for each subject, to extract hierarchical communities of events [8]. We then investigated the large (> 100 events) found at the coarsest level of hierarchy (level 2) and derived the community centroid for the pre-casting, casting, and post-casting periods. We compared whether there are significant system-level differences between the casting and pre-casting, casting and post-casting periods compared to a scan randomized and nodal-permutation null. We note that all analyses were performed on each subject's fMRI time series data, individually due to different TRs.

First, we found that motor pulses and events are significantly more temporally aligned than chance (Fig.1a, d) but that their activation patterns are not highly correlated (Fig. 1b). Also, events were found to be significantly more variable than motor pulses (Fig. 1c). We then identified the brainwide activation patterns of global events and positive motor pulses (Fig.1e, f). After establishing that motor pulses and events are not identical, and that events are more variable than motor pulses, we identified the hierarchical organization of event clusters for each subject (Fig.2a, f, k). In each subject there were 4 ~ 6 large communities at level 2 (Fig.2b, g, l). Our results show that the largest community which constitutes ~50 percent of all events in level 2 are similar upon visual inspection across subjects and when creating brain plots using the cluster's first principal component (Fig.2c, h, m). Lastly, we investigated whether this cluster changes between paradigms - pre-casting, casting, and post-casting periods (Fig.2d, e, i, j, m, o). Overall, these results suggest that there are local and global changes in events throughout the disuse paradigm.

Our results demonstrate that peaks of motor cortex-specific activations are not identical to that of global events which change systematically with the casting paradigm. Shifting the focus to global events of various organizational scales may allow further understanding of changes occurring in the disuse paradigm. Also, the dataset includes multiple scans of each subject, allowing for precise, individualized investigation of disuse-induced changes in the brain's functional organization [9, 10].

Connectivity (eg. functional, effective, structural) ²

fMRI Connectivity and Network Modeling ¹

Motor Behavior Other

BOLD fMRI

Computational Neuroscience

FUNCTIONAL MRI

Motor

Plasticity