Department of Clinical Neurosciences

University of Cambridge
{imc31,sc672}@cam.ac.uk

Spectral signatures of brain networks

in disorders of consciousness and sedation
Iulia M. Coma and Srivas Chennu

INTRODUCTION
OVERVIEW CLINICAL SIGNIFICANCE A BETTER SOLUTION? SUMMARY
Impaired states of consciousness are Understanding the mechanics of Graph theory can quantify key We present three studies that
characterised by distinct signatures of brain consciousness impairment would have properties of connectivity use electroencephalographic
networks. These signatures often reflect the multifaceted clinical benefits. Up to 42% of networks of the brain at various (EEG) data recorded at the
level of behavioural responsiveness, thus patients with disorders of consciousness are organization scales. Thus, it can bedside to investigate patterns
giving us insight into how the brain misclassified1, 1 in 1000 individuals remain potentially inform about brain and dynamics of functional
subserves consciousness. Furthermore, inadvertently aware during anaesthesia2, states of patients with disorders brain networks during three
brain network patterns can reveal covert and current behavioural measures for of consciousness or under conditions of pharmacologically
processes that differentiate between prognosis after severe brain injury are sedation, patients, thus aiding in and pathologically impaired
conditions of impaired consciousness. insufficiently informative3. clinical diagnostics and prognosis. consciousness.

METHODS
Samples used from populations with EEG ANALYSIS ASSESSING
CONNECTIVITY
RESPONSIVENESS
impaired consciousness Graph theory analysis NETWORKS
High-density Connectivity During sedation: drug level in blood
networks The Weighted Phase Lag Index
Sedation EEG (91 channels) and hit rate during a simple button
(WPLI)4 estimates functional press auditory discrimination task.
WPLI matrix

20 healthy subjects; connectivity between pairs of

Sedative: propofol;
Disorders of In disorders of consciousness: the
consciousness nodes (here, scalp electrodes).
7 min recordings at baseline, CRS-R (Coma Recovery Scale
mild sedation, moderate sedation It uses cross-spectral analysis
Revised)5, which measures the degree
and recovery. to compute phase differences
of visual, auditory and motor
between signals and it corrects
Threshold response, communication ability, and
for volume conduction.
Traumatic coma Vegetative Minimally Keep 10-50% Flattened 3D wakefulness.
No wakefulness, state conscious state Cross-spectral strongest topograph topograph
no awareness analysis connections What network properties can graph theory discover?6
Wakefulness Wakefulness with
16 patients; without awareness partial awareness Canonical frequencies: Segregation Integration Node centrality
Bedside overnight Delta: 1-4 Hz The presence of local, The facility of long-range Nodes involved in
recordings early after Theta: 4-8 Hz specialised groups of nodes information exchange across information exchange
32 chronic patients (13 VS, 19 MCS); Alpha: 8-13 Hz
injury (acute phase); at micro- and mesoscale nodes in the network at between modules:
64 healthy controls;
CRS-R outcome scores level: node clustering macroscale level: characteristic betweenness,
10 min of resting state data.
recorded after 2 months. coefficient, modularity. path length, global efficiency. participation coefficient.

RESULTS
SEDATION7 COMA
Subjects were split into 2 groups: Patients were split into 2 groups:
Responsive: Subjects who remained Good outcome: CRS-R score >= 10
responsive during moderate sedation. Poor outcome: CRS-R score < 10.
Drowsy: Subjects who stopped Network topography
responding during moderate sedation, discriminates between good
at the same blood level of drug. and poor long-term outcome
Alpha band connectivity networks in acute comatose patients.
predict behaviour under similar At the same levels of behavioural responsiveness
early after traumatic brain injury, patients with
anaesthetic blood level. good eventual outcome showed robust fronto-
The responsive group showed a stable fronto-centro- parietal connectivity in alpha and theta networks,
occipital pattern of connectivity, whereas drowsy group whereas patients with poor outcome showed
networks altered significantly. stronger connectivity in delta networks.

Alpha band power changes
as a function of sedation.
Alpha network connectivity changes mirror
alpha power changes during sedation.
Alpha network properties CHRONIC DISORDERS OF CONSCIOUSNESS8
before sedation predict
Connectivity network topology varies Alpha network metrics predict CRS-R
susceptibility to propofol.
across healthy controls and patients. scores and suggest covert awareness.
Despite no differences in alpha power
strength or topography before sedation, Alpha networks in healthy brains display a balance between Good scores in MCS patients are predicted by alpha network
small-world-ness properties of alpha strong local interactions and robust interconnectivity. Patients metrics. Alpha network topography also discriminates between
networks predicted whether subjects with impaired consciousness show stronger connectivity at patients with the same clinical classification who are able or
stopped responding during moderate lower frequencies. unable to imagine playing tennis.
sedation.

SUMMARY
Functional brain connectivity, conveniently assessed using EEG, is helpful in
explicating and predicting behaviour in conditions of impaired consciousness.
The presence of a robust frontoparietal module in alpha networks predicts
preserved alertness during moderate sedation, higher CRS-R scores in chronic
disorders of consciousness, and better long-term clinical outcomes from coma.
On the contrary, stronger structured connectivity in lower frequency delta
networks suggests states of impaired consciousness.
Our findings inform the development of clinically valuable tools for bedside REFERENCES
1Schnakers, C. et al. Brain Injury 22, 786-792 (2008) | 2Sebel, P. et al. Anesth Analg 99, 833-839 (2004) | 3Grote, S., et al. J Neurotraum 28, 527-534 (2011)
4Vinck, M., et al. NeuroImage 55, 1548-1565 (2011) | 5Giacino, J., et al. Arch Phys Med Rehab 85, 2020-2029 (2004) | 6Bullmore, E. & Sporns, O. Nat Rev
diagnostics and monitoring in intensive care units and in the operating theatre. Neurosci 10, 186-198 (2009) | 7Chennu, S. et al. PLoS Comp Biol (in press) | 8Chennu, S. et al. PLoS Comp Biol 10, e1003887 (2014).