COALIA: A Computational Model of Human EEG for Consciousness Research

Abstract

Understanding the origin of the main physiological processes involved in consciousness is a major challenge of contemporary neuroscience, with crucial implications for the study of Disorders of Consciousness (DOC). The difficulties in achieving this task include the considerable quantity of experimental data in this field, along with the non-intuitive, nonlinear nature of neuronal dynamics. One possibility of integrating the main results from the experimental literature into a cohesive framework, while accounting for nonlinear brain dynamics, is the use of physiologically-inspired computational models. In this study, we present a physiologically-grounded computational model, attempting to account for the main micro-circuits identified in the human cortex, while including the specificities of each neuronal type. More specifically, the model accounts for thalamo-cortical (vertical) regulation of cortico-cortical (horizontal) connectivity, which is a central mechanism for brain information integration and processing. The distinct neuronal assemblies communicate through feedforward and feedback excitatory and inhibitory synaptic connections implemented in a template brain accounting for long-range connectome. The EEG generated by this physiologically-based simulated brain is validated through comparison with brain rhythms recorded in humans in two states of consciousness (wakefulness, sleep). Using the model, it is possible to reproduce the local disynaptic disinhibition of basket cells (fast GABAergic inhibition) and glutamatergic pyramidal neurons through long-range activation of vasoactive intestinal-peptide (VIP) interneurons that induced inhibition of somatostatin positive (SST) interneurons. The model (COALIA) predicts that the strength and dynamics of the thalamic output on the cortex control the local and long-range cortical processing of information. Furthermore, the model reproduces and explains clinical results regarding the complexity of transcranial magnetic stimulation TMS-evoked EEG responses in DOC patients and healthy volunteers, through a modulation of thalamo-cortical connectivity that governs the level of cortico-cortical communication. This new model provides a quantitative framework to accelerate the study of the physiological mechanisms involved in the emergence, maintenance and disruption (sleep, anesthesia, DOC) of consciousness.

Keywords: GABA; TMS-EEG; brain connectivity; computational modeling; disinhibition; disorders of consciousness (DOC); feedforward inhibition.

Figure 1

(A) Illustration of “horizontal” and “vertical” connectivity. Left: Horizontal connectivity refers to the cortico-cortical connections which are functionally effective during wakefulness, with a weak level of thalamo-cortical coupling. Right: Vertical connectivity refers to thalamo-cortical projections that functionally impair cortico-cortical connectivity during sleep. (B) General architecture of the micro- and macro-circuits implemented in the computational model. The local neural mass model (NMM) of cortical activity is composed of a PC exciting two GABAergic interneurons, namely, the somatic-projecting BC and the dendritic-projecting SST, responsible for the generation of fast and slow oscillations, respectively; the VIP were introduced as they play a crucial role in cortical column communication through disinhibition of SST. The subcortical module consisted in TC sending excitatory glutamatergic projections to the TRN block composed of fast and slow GABAergic interneurons TRN₁ and TRN₂, respectively. VIP, vasoactive intestinal peptide positive GABAergic interneurons; SST, Somatostatin-positive GABAergic interneurons; BC, Basket-type GABAergic interneurons; PC, Glutamatergic Pyramidal Cells.

Figure 2

Large-scale architecture of the model. Illustration of the synaptic projections between cortical modules, and between thalamic and cortical modules. Note the presence of long-range thalamocortical and cortico-cortical feedforward inhibition. For the sake of clarity, long-range connections between cortical NMM_i and NMM_j are unidirectional, whereas in the model, PCs of NMM_j also project on neurons of NMM_i. The strength of the synaptic input of PC onto distant VIP cells is larger than the input onto the other distant interneurons (SST, Basket).

Figure 3

Full processing pipeline leading to the simulation of scalp EEGs. The pipeline to simulate EEG is a two-step process. (A) First, the forward problem is solved at the level of 257 scalp electrodes from a dipole layer constrained to the surface of a cortical mesh (15,000 vertices) midway between the Gray/White matter interface and the cortex surface. The boundary Element Method is used for the calculation within a realistic head model that accounts for the conductivity properties and the geometry of brain, skull and scalp. This step provides a 257 × 15,000 leadfield matrix A representing the contribution of each individual cortical dipole at each of the 257 scalp electrodes. In this matrix, leadfield vectors belonging to a common region of the Desikan Atlas are added to obtain a simplified 66 × 257 matrix G. Second, the time courses S at the whole brain level are obtained in the mean-field model from a set of 66 cortical and one thalamic coupled NMMs. (B) Coupling between these 67 NMMs is done using combination of connection weight matrices. Pairs of structurally connected cortical NMMs are first defined from a matrix of connection weights representing a density of fibers between all pairs of 66 cortical regions of the Desikan Atlas. This matrix is provided in Hagmann et al. (2008). Using an element-wise multiplication, this matrix is combined with a set of horizontal (i.e., cortico-cortical) functional connectivity matrices that reproduce the coefficients weights used for wakefulness and sleep in the toy model. Vertical (i.e., thalamo-cortical) connectivity matrices are added to each of these products to obtain connectivity weight matrices that account for anatomical connections as well as cortico-cortical and thalamo-cortical connectivity matrices. Cortico- and thalamo-cortical time-delays were similarly organized in the form of matrices where the elements represent the Cartesian distance between cortical NMMs divided by the mean velocity of traveling for action potentials. (C) The mean-field model includes explicitly the contribution of an external stimulus term that represents the effect of TMS. At the output of the pipeline, scalp EEG signals at the level of 257 channels are obtained as the product of leadfield G and source time courses S.

Figure 4

Comparison of real and simulated intracerebral EEG (iEEG; toy model, N = 4). Left column: in the condition of high thalamo-cortical connectivity (i.e., low cortico-cortical connectivity), signals generated by the mean-field model are characterized by delta waves (~4 Hz). These simulated signals are similar (although slightly faster) to delta waves recorded by iEEG during slow wave sleep (SWS) in non-epileptic cortical regions of one patient undergoing an invasive EEG exploration. Right column: in the condition of low thalamo-cortical connectivity, signals generated by the mean-field model are similar to background activity recorded by iEEG in real conditions during wakefulness. Note that these iEEG recordings are performed in patients who are candidates to epilepsy surgery. For the sake of this study, only iEEG signals that do not show epileptic activity were retained.

Figure 5

Simulated vs. Real EEG during wakefulness and SWS. Signals simulated with the whole-brain model using weak thalamo-cortical connectivity parameters (A) display background activity. The morphology and spectral content of these simulated signals are similar to scalp EEG recorded in a human subject during wakefulness in humans (C), except for a higher power spectral density in the beta sub-band. Signals simulated with the whole-brain model using strong thalamo-cortical connectivity parameters (B) display delta waves similar to the activity recorded in real condition during SWS in humans (D). The spectral content of signals as well as the topographical voltage distribution at the peak of delta waves were similar in the simulated and real conditions.

Figure 6

Comparison of TMS-EEG evoked responses in silico and in humans. (A) Time course of a human TMS-EEG response (modified from Casali et al., 2013) following stimulation of the motor cortex during wakefulness. Once that cortical sources have been computed from EEG recordings, a spatio-temporal matrix of significant sources was built and the Lempel-Ziv compression algorithm was used to compute the complexity of the evoked response (Perturbational Complexity Index, PCI). (B) Time course of a simulated TMS-EEG response using our brain-scale, following stimulation of the motor area in the wakefulness mode. Cortical sources were reconstructed from the simulated EEG, and a similar procedure was used to compute PCI. A similar PCI value was obtained in the simulated and experimental TMS-evoked EEG responses in the awake state.

Figure 7

Comparison of TMS-evoked EEG responses in wakefulness and sleep. (A) TMS-evoked EEG responses obtained through TMS of the motor cortex in humans (upper panel, during wakefulness; lower panel, during sleep) with associated PCI values. (B) TMS-evoked EEG responses obtained through simulated TMS of the motor region (upper panel, in the wakefulness mode; lower panel, during the sleep mode) with associated PCI values.