Stimulation-related modifications of evolving functional brain networks in unresponsive wakefulness

Abstract

Recent advances in neurophysiological brain network analysis have demonstrated novel potential for diagnosis and prognosis of disorders of consciousness. While most progress has been achieved on the population-sample level, time-economic and easy-to-apply personalized solutions are missing. This prospective controlled study combined EEG recordings, basal stimulation, and daily behavioral assessment as applied routinely during complex early rehabilitation treatment. We investigated global characteristics of EEG-derived evolving functional brain networks during the repeated (3-6 weeks apart) evaluation of brain dynamics at rest as well as during and after multisensory stimulation in ten patients who were diagnosed with an unresponsive wakefulness syndrome (UWS). The age-corrected average clustering coefficient C* allowed to discriminate between individual patients at first (three patients) and second assessment (all patients). Clinically, only two patients changed from UWS to minimally conscious state. Of note, most patients presented with significant changes of C* due to stimulations, along with treatment, and with an increasing temporal distance to injury. These changes tended towards the levels of nine healthy controls. Our approach allowed to monitor both, short-term effects of individual therapy sessions and possibly long-term recovery. Future studies will need to assess its full potential for disease monitoring and control of individualized treatment decisions.

Figure 1

Top: examination schedule comprising two baseline conditions (6 min. each) prior to and after a multisensory stimulation block (18 min.) probing the somatosensory, auditory, and visual modalities. Bottom: exemplary temporal evolution of the clustering coefficient along the examination schedule. A 7-point sliding average is indicated by the black line.

Figure 2

Sample distributions of average clustering coefficient C (left) and average shortest path length L (middle) of evolving functional brain networks from patients during period B1 at first (A1) and second assessment (A2) and from healthy controls. Bottom and top of a box are the first and third quartiles, and the red band and the black triangle are the median and the mean of the distribution, respectively. The ends of the whiskers represent the interquartile range of the data. Outliers are marked by an o sign. Right: Scatterplot of average clustering coefficients C and average shortest path lengths L during period B1 from patients (both assessments; black dots) and healthy controls (single assessment; blue diamonds). Linear regression (all subjects) is represented with a solid black line (patients: Spearman’s ρ = − 0.80, p < 0.01; healthy controls: ρ = − 0.88, p < 0.01).

Figure 3

Scatterplot of average clustering coefficients C (all conditions: B1, S, B2) and age from patients at both assessments (A1: grey triangle; A2: black dot) and healthy controls (blue). Solid lines represent linear regressions (Spearman’s correlation coefficient ρ).

Figure 4

Results of between-patient comparisons (Scheffé tests; top) and distributions of age-corrected average clustering coefficients C* (bottom) at assessments (A1,A2). White rectangles indicate significant differences (p < 0.05). Regarding the distributions of C*, bottom and top of a box are the first and third quartiles, and the black band is the median of the distribution, respectively. The ends of the whiskers represent the interquartile range of the data. Outliers are marked by an o-sign. Blue dotted lines and shaded areas indicate median and range of age-corrected average clustering coefficients C* of healthy controls. At assessment (A1), C* of patient 3 (orange) differs significantly from the values of patients 6 and 12 (green). At (A2), C* of patients 1, 4, 6, and 8 (green) differ significantly from the values of patients 2, 3, 9, 10, 11, and 12 (orange). The EEG recording of patient 1 during B1 at A1 had to be excluded from the analyses because of too many artefacts.

Figure 5

Top: Sample distributions of age-corrected average clustering coefficients C* of evolving functional brain networks during periods B1, S, and B2 (left) as well as of stimulation-induced changes ΔC* between periods (right) at first (A1) and second assessment (A2). Properties of boxplots as in Fig. 2. Bottom: Individual stimulation-induced changes ΔC* between periods B1 and S at first (A1: grey triangle) and second assessment (A2: black dot) on a time line starting with injury.

Figure 6

Proposed model of short-term and long-term stimulation-induced modifications of evolving functional brain networks in patients with an unresponsive wakefulness syndrome. The black arrow indicates a long-term treatment effect due to continued rehabilitation complex treatment as monitored with age-corrected average clustering coefficient C* at different assessments (A1,A2). Short-term modifications can be monitored with changes of C* (ΔC*) while evolving functional brain networks transit from a baseline period at rest (B1) to a period of sensory stimulation (S) and back to a baseline period at rest (B2). A negative change (ΔC* < 0 for B1 → S) is considered as favorable. It can be observed in an increasing number of patients at (A2), while negative changes decrease. The general tendency towards a more segregated network (decreasing C*), as seen for healthy controls, indicates success of therapy and thereby recovery from injury.