Mild traumatic brain injury impairs the coordination of intrinsic and motor-related neural dynamics

Abstract

Mild traumatic brain injury (mTBI) poses a considerable burden on healthcare systems. Whilst most patients recover quickly, a significant number suffer from sequelae that are not accompanied by measurable structural damage. Understanding the neural underpinnings of these debilitating effects and developing a means to detect injury, would address an important unmet clinical need. It could inform interventions and help predict prognosis. Magnetoencephalography (MEG) affords excellent sensitivity in probing neural function and presents significant promise for assessing mTBI, with abnormal neural oscillations being a potential specific biomarker. However, growing evidence suggests that neural dynamics are (at least in part) driven by transient, pan-spectral bursting and in this paper, we employ this model to investigate mTBI. We applied a Hidden Markov Model to MEG data recorded during resting state and a motor task and show that previous findings of diminished intrinsic beta amplitude in individuals with mTBI are largely due to the reduced beta band spectral content of bursts, and that diminished beta connectivity results from a loss in the temporal coincidence of burst states. In a motor task, mTBI results in diminished burst amplitude, altered modulation of burst probability during movement, and a loss in connectivity in motor networks. These results suggest that, mechanistically, mTBI disrupts the structural framework underlying neural synchrony, which impairs network function. Whilst the damage may be too subtle for structural imaging to see, the functional consequences are detectable and persist after injury. Our work shows that mTBI impairs the dynamic coordination of neural network activity and proposes a potent new method for understanding mTBI.

Keywords: Beta bursts; Concussion; MEG; Networks; mTBI; mild Traumatic Brain Injury.

Fig. 1

Patient enrolment flow chart.

Fig. 2

Resting-state burst statistics. a) The upper panel shows the spatial distribution of beta amplitude during bursts. The lower panel shows beta amplitude in the non-burst windows. In both cases, patients are shown on the left and controls on the right. b) Spatial signature of the differences in beta amplitude during bursts (top) and non-burst periods (bottom) (i.e. the difference between patients and controls, left and right in a) respectively) c) The left panel shows average burst amplitude, the right panel shows beta amplitude in the non-burst states, demonstrating no significant difference between patients and controls. In both cases, results are collapsed across 78 brain regions and each data point represents an individual subject. Note that burst beta amplitude is significantly (p = 0.016; Wilcoxon sum rank test) diminished in patients. d) Violin plot showing the average time spent in the burst state (no significant difference between groups) e) Scatter plot of global burst amplitude versus SCAT-2 symptom severity. Spearman correlation for the combined data points (patients and controls) gave R = -0.399; p = 0.0088. For the patients only, we found R = -0.189; p = 0.387.

Fig. 3

Resting-state functional connectivity. a) Resting-state connectome matrices for patients (upper panel) and controls (lower panel). In both cases, connectivity is calculated via the assessment of burst coincidence between regions. The glass brains show the spatial structure of the information in the corresponding connectome matrices. The red lines show 5% of connections with the highest functional connectivity value. b) The glass brain plot shows the 2% of connections with the largest difference between patients and controls. c) Whole-brain functional connectivity assessment in patients and controls. Each data point represents the average connectivity across all connections, for a single individual (i.e. the overall sum of all of the matrix elements in the connectomes shown in (a) divided by the number of connections). Whole head connectivity – computed via assessment of burst coincidence – is significantly (p = 0.031; Wilcoxon sum rank test) diminished in patients. d) Glass brain showing the connections selected using recursive Random Forest Feature selection and violin plot showing the average connectivity for those connections in both groups. e) Scatter plot showing the relationship between symptom severity and burst connectivity in the rRF-FS selected connections. Spearman correlation for the combined data points gave R = -0.72; p = 8x10^-8. Spearman correlation for the mTBI group only gave R = -0.45; p = 0.03. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

Burst statistics during a motor task. a) Time courses showing the probability of bursts throughout a button press. The upper panel shows the case for the left primary motor cortex during a left hand (left) and right hand (right) movement. The lower panel shows the case for the right primary motor cortex during a right hand (left) and left hand (right) movement. b) Burst amplitude in the beta band during the post-movement beta rebound. Left-hand plots show contralateral and ipsilateral cortices during a left-hand button press. Right-hand plots show ipsilateral and contralateral cortices during a right-hand button press. * indicates statistical significance (p < 0.05*; Wilcoxon Rank Sum test) following FDR correction for multiple comparisons.

Fig. 5

Functional connectivity during a motor task. a) Time course showing the temporal evolution of burst coincidence throughout finger movement. Note that burst coincidence is significantly (p = 0.021; Wilcoxon rank sum test) less likely during the beta rebound window, in patients relative to controls. b) The spatial distribution of dominant connections at 4 time points during the task. The upper set shows the case for patients. The centre set shows the case for controls, and the lower set shows the dominant differences. In all cases, the 2% of connections with the highest values are shown. Notice that during the rebound window, the dominant differences are between bilateral motor regions. See the supplementary materials for a video showing the whole time evolution of the functional connections.