Impact of moderate aerobic exercise on small-world topology and characteristics of brain networks after sport-related concussion: an exploratory study

Abstract

Sport-related concussion (SRC) is a complex brain injury. By applying graph-theoretical analysis to networks derived from neuroimaging techniques, studies have shown that despite an overall retention of small-world topology, changes in small-world properties occur after brain injury. Less is known about how exercise during athletes' return to sport (RTS) influences these brain network properties. Therefore, in the present study dense electroencephalography (EEG) datasets were collected pre- and post-moderate aerobic exercise. Small-world properties of whole brain (WB) and the default mode network (DMN) were extracted from the EEG datasets of 21 concussed athletes and 21 healthy matched controls. More specifically, path length (LP), clustering coefficient (CP), and small-world index (SWI) in binary and weighted graphs were calculated in the alpha frequency band (7-13 Hz). Pre-exercise, SRC athletes had higher DMN-CP values compared to controls, while post-exercise SRC athletes had higher WB-LP compared to controls. Weighted WB analysis revealed a significant association between SRC and the absence of small-world topology (SWI ≤ 1) post-exercise. This explorative study provides preliminary evidence that moderate aerobic exercise during athletes' RTS induces an altered network response. Furthermore, this altered response may be related to the clinical characteristics of the SRC athlete.

Keywords: Concussion; Graph theory; Resting-state EEG; Small-world topology; Sports injury.

Figure 1

Standardized moderate aerobic bike exercise test.

Figure 2

Data processing pipeline for the DMN network. Before the resting-state EEG measurement, the electrode locations on the scalp were obtained using Captrak. After the resting-state EEG measurement, the collected dataset was pre-processed in BrainAnalyzer software. In preparation for functional connectivity analysis, a head model was computed in Brainstorm, run through MATLAB. Next, the ROIs corresponding to the DMN network were exclusively selected (according to Kabbara et al.). PLV between the regions of the network was calculated as a measure of functional connectivity. The resulting matrix was exported and graph (/small world) characteristics such as SWI were computed using BCT. EEG: electroencephalography, ROIs: regions of interest, DMN: default mode network, PLV: Phase Locking Value, SWI: small-world index, BCT: Brain Connectivity Toolbox. Some of the above images were generated with the help of BrainNet Viewer.