Characterization of Large Scale Functional Brain Networks During Ketamine-Medetomidine Anesthetic Induction
Several experiments evidence that specialized brain regions functionally interact and reveal that the brain processes and integrates information in a specific and structured manner. Networks can be used to model brain functional activities constituting a way to characterize and quantify this structured form of organization. Reports state that different physiological states or even diseases that affect the central nervous system may be associated to alterations on those networks, that might reflect in graphs of different architectures. However, the relation of their structure to different states or conditions of the organism is not well comprehended. Thus, experiments that involve the estimation of functional neural networks of subjects exposed to different controlled conditions are of great relevance. Within this context, this research has sought to model large scale functional brain networks during an anesthetic induction process. The experiment was based on intra-cranial recordings of neural activities of an old world macaque of the species Macaca fuscata. Neural activity was recorded during a Ketamine-Medetomidine anesthetic induction process. Networks were serially estimated in time intervals of five seconds. Changes were observed in various networks properties within about one and a half minutes after the administration of the anesthetics. These changes reveal the occurrence of a transition on the networks architecture. During general anesthesia a reduction in the functional connectivity and network integration capabilities were verified in both local and global levels. It was also observed that the brain shifted to a highly specific and dynamic state. The results bring empirical evidence and report the relation of the induced state of anesthesia to properties of functional networks, thus, they contribute for the elucidation of some new aspects of neural correlates of consciousness.
One of the major goals of neuroscience is to understand how the brain works, comprehending how cognitive capacities or physiological states of the organism are related to neural processes that involve functional interactions of several brain areas. Although neuroscience is a well developed and consolidated science, an understanding at this level has not been reached, and possibly the scientific community may still need several years in order to comprehend the brain at this level.
In recent times, due to the influence of other fields of science, such as the physics of complex systems, allied with constant realizations that neural activities involve the participation of many distinct brain areas at the same time, a new perspective has been gaining strength among the community of neuroscientists. This perspective aims to understand the brain as a whole, from a system point of view, and considers as fundamental the comprehension of the way how functional interactions occur among distinct brain regions, in order to understand how the brain is able to perform its activities (Bullmore and Sporns, 2009). Based on this perspective, researchers have been using concepts and tools from Modern Network Science (Strogatz, 2001; Newman, 2003) in order to model, characterize and study the brain.
The Modern Network Science is a highly interdisciplinary field of science aimed at understanding the functioning, behavior, and evolution of complex systems, based on properties of their structure, that is, the specific way how elements of the system establish interactions (Mitchell, 2009). One of the main uses of complex networks in neuroscience is carried out through the modeling of functional interactions established among distinct cerebral regions by the use of the mathematical structure of a graph. Assembled those networks in which nodes represent cortical areas and edges functional interactions (Friston et al., 1993; Friston, 1994) between the regions, several procedures and analysis can be performed in order to comprehend the properties those networks. Using complex networks measures (Rubinov and Sporns, 2010), it is possible to estimate how the connectivity is shaped, the organization at local and global levels, the nodes that perform great influence in the integration of several regions and mediation of the flow of information (Bullmore and Sporns, 2009; Sporns, 2011; Stam and Van Straaten, 2012).
It is believed that those networks reflect the specific way of how processing and integration of information is performed among distinct brain areas, and there is a consensus that different cognitive demands, or disorders that affect the central nervous system may be associated with different configurations and properties of the networks (Stam and Reijneveld, 2007; Sporns, 2011).
The understanding about general anesthesia is of great relevance for both medicine and neuroscience. Given its importance and the fact that it is not well comprehended, the Science Magazine has pointed out the elucidation of processes and mechanisms involving general anesthesia as one of 125 most important open questions in science (Kennedy and Norman, 2005).
Anesthetic agents are small molecules that when administrated to patients are able to interact and modulate the activity of specific ionic channel proteins in neurons inducing dramatic physiological alterations in the organism. General anesthesia can be defined as a state physiologically stable, reversible, induced by drugs that lead to analgesia, amnesia, immobility and loss of consciousness (Schwartz et al., 2010). Those pharmacological effects are widely known and described. However, very little is known about the neurophysiological mechanisms that underlie the sedation and loss of consciousness (Schwartz et al., 2010; Lewis et al., 2012).
Besides their undeniable importance and usefulness in clinical medicine, those drugs constitute tools of great value to neuroscience. Anesthetic agents are experimental tools able to induce different levels of consciousness, in a stable and reproductive manner (Uhrig et al., 2014). Thus, they offer an excellent opportunity for the study of consciousness and neural correlates of consciousness, providing possibilities to comprehend fundamental process and phenomena that happen in the brain (Hameroff et al., 1998).
There are several theories about anesthesia (Flohr, 1995; Alkire et al., 2000; Mashour, 2004; John and Prichep, 2005), and also theories of consciousness based on the interface between consciousness and anesthesia (Hameroff, 2006; Mashour, 2006). There are also several hypotheses and reports on how anesthetics agents lead to the loss of consciousness, both are based on the depression of cerebral functions (Alkire et al., 1995; Alkire, 1998), on the reduction of functional interactions between brain areas (White and Alkire, 2003; Imas et al., 2005), fragmentation of neural networks (Lewis et al., 2012), and others.
Within this context of several theories, hypotheses, and reports, the present study has sought to use tools and concepts of the Modern Network Science in order to investigate the alterations on the organization of functional brain activities that occur at the onset of the anesthetic induction. In order to perform those analyses, macaque functional brain networks were estimated serially over time during an anesthetic induction process. Complex networks measures were used in order to characterize and compare the graphs estimated in different instants of time during the experiment.
Experimentally, an old world monkey of the species Macaca fuscata was used as an animal model. This species of macaque has great anatomical and evolutionary similarity with humans, making them an excellent platform for the study of the human brain (Iriki and Sakura, 2008). Being the Macaca fuscata one of the main primates used as experimental model in neuroscience, specially among Japanese neuroscientists (Isa et al., 2009).
A cocktail of Ketamine and Medetomidine was used to induce anesthesia. Ketamine is a drug able to induce an anesthetic state characterized by the dissociation between the thalamocortical and limbic systems (Bergman, 1999). It acts as a non competitive antagonist of the receptor N-metil-D-asparthate (Green et al., 2011). Medetomidine (agonist alfa-2 adrenergic receptor) was combined with Ketamine to promote muscular relaxation (Young et al., 1999). The antagonist of the Medetomidine, Atipamezole, was used to trigger and promote the recovering process (Young et al., 1999).
The technique used to record neural activity was the Multidimensional Recording Electrocorticogram (MDR-ECoG). This technology records neural activity using an ECoG electrode array implanted chronically at subdural space directly over the cortical surface of the monkey. According to the researchers that developed this technology, the MDR-ECoG is the most advanced and balanced technology available to record neural activity, it is able to sample neural signals at temporal resolutions higher than 1KHz, and a spatial resolution up to 3mm (Yanagawa et al., 2013; Fukushima et al., 2014).
Neural Connectivity Estimator
As a neural connectivity estimator, a method based on Granger causality (Granger, 1969) was used in order to infer statistical dependencies between the time series of the electrodes. When applied in neuroscience, Granger causality provides an indication about the information flow from one cortical area to another (Seth and Edelman, 2007; Seth, 2010).
Experimental Procedures - Summary of Steps
The modeling of functional neural activities performed in this study followed these steps:
Neural registers were recorded by using the MDR-ECoG technique. The matrix of ECoG electrodes continuously covered an entire brain hemisphere and parts of the medial walls.
Each one of the electrodes of the array was considered as a vertex of the network, and represented the cortical area in which it was positioned.
A neural connectivity estimator of Granger causality in the frequency domain was used to estimate values of association between the registers (time series) of the electrodes.
An adjacency matrix was assembled, containing all the pairwise association values between the nodes.
The characterization of the topology of the estimated networks was performed by the use of some complex networks measures.
Networks were estimated serially in time intervals of five seconds, in five physiological frequency bands along the experiment. In each frequency band, all the networks in the sequence were estimated through the same procedures and parameters. Thus, the alterations observed between the properties of distinct networks in the sequence came only from differences on the records of neural activity.
Note: All the experimental procedures involving the animal model and data records (Methods Sections II and III), were performed by researchers from the laboratory of adaptive intelligence at the RIKEN BRAIN SCIENCE INSTITUTE, laboratory under the supervision of PhD. Naotaka Fujii.
Ii Subjects and Materials
A multi-channel ECoG electrode array (Unique Medical, Japan) was chronically implanted in the subdural space of an adult male macaque (Macaca fuscata). Each electrode was made of a 3 mm platinum disc, coated with insulating silicone, except for a 0.8mm diameter central area of the disc. One hundred and twenty eight ECoG electrodes with an average inter-electrode distance of 5 mm were implanted in the left brain hemisphere, continuously covering the frontal, parietal and occipital lobes. Some electrodes were also disposed in the medial frontal and parietal walls. Rectangular platinum plates were placed in the subdural space between the ECoG array and the duramater, to serve as reference electrodes. A ground electrode was placed in the epidural space, see (Nagasaka et al., 2011) for a detailed description. Electromyography (EMG) was performed at 1Khz , by a data acquisition system (Blackrock Microsystems, USA). All experimental and surgical procedures were performed according to experimental protocols (No. H24-2-203(4)) approved by the RIKEN ethics committee and the recommendations of the Weatherall report, "The use of non-human primates in research".
Iii Experimental Procedures
The monkey was seated in a proper chair with head and arms restrained. Neural activity started to be recorded with the monkey awake and with open eyes. After that, the eyes were covered with a patch to prevent visual evoked response. After about 10 minutes, a Ketamine-Medetomidine cocktail (5.6mg/Kg of Ketamine + 0.01mg/Kg of Medetomidine) was injected intramuscularly to induce anesthesia. The loss of consciousness point (LOC) was set at the time when the monkey no longer responded to external stimuli (touching the nostrils or opening the hands). After established the LOC, neural activity was recorded for more 25-30 minutes. Heart rate and breathing were monitored throughout the entire experiment.
Iv Signal Processing and Granger Causality in the Frequency Domain
A reject-band IIR-notch filter was used to attenuate components of the signal at 50Hz.
The signal was down-sampled from 1KHz to 200Hz.
The signal was divided into windows of 1000 points (equivalent to a five-second recording of neural activity).
For each of the 128 time series, the trend was removed and the average was subtracted.
For the computation of association values using Granger causality in the frequency domain, were used with some adaptations the following libraries: MVGC GRANGER TOOLBOX, developed by PhD. Anil Seth (Sussex University, UK), described in (Seth, 2010), available at www.anilseth.com, and the library BSMART toolbox (Brain-System for Multivariate AutoRegressive Timeseries toolbox) described in (Cui et al., 2008) and available at www.brain-smart.org.
Computation of Causal Interactions
To find the model order (number of observations to be used in the regression model), it was used the criteria of selection of models from Akaike (AIC) and Bayes/Schwartz (BIC). Both methods returned an order of the model equal to seven.
At each window of 1000 points, Granger causality in the frequency domain interactions were pair-wise computed among the 128 time series, by the use of the function cca_pwcausal() (MVGC GRANGER TOOLBOX).
Granger causality interactions were calculated in five physiological frequency bands: Delta (0-4Hz), Theta (4-8Hz), Alpha (8-12Hz), Beta (13-30Hz) and Gamma (25-100Hz).
The interaction values obtained were saved in adjacency matrices.
Graphs and Networks
For each sequence of graphs respective to a frequency band, a threshold was chosen, and only the interactions with magnitude values higher this threshold were considered as edges of the graphs.
Delta (0-4Hz), threshold =
Theta (4-8Hz), threshold =
Alpha (8-12Hz), threshold =
Beta (13-30Hz), threshold =
Gamma (25-100Hz), threshold =
As discussed in (Bullmore and Sporns, 2009; Sporns, 2011), it is possible for scientists to use different criteria in order to determine this parameter. In the present study, due to experimental conditions, each sequence of networks contained graphs with distinct connectivity. Thresholds were chosen in such a way to prevent graphs with lower connectivity in each sequence from presenting many disconnected parts or vertices, which might introduce distortions in the analysis.
After obtaining non-weighted graphs, the directions of the edges were removed, resulting in undirected and non-weighed networks.
Analysis of Topology
Network measures were used in order to characterize the topology and properties of the graphs.
Topological alterations on the functional brain networks were observed along the anesthetic induction process, being verified the occurrence of alterations on the distinct measures used to characterize the networks. Those results reveal that there are differences in the way how cortical areas functionally interact regarding awake conditions and during the induced state of anesthesia.
Within about one and a half minutes after the administration of the anesthetic cocktail, abrupt changes were observed on several networks properties, revealing the existence of a rapid transition between the two states (awake/anesthetized). Alterations were also observed after the macaque was blindfolded, demonstrating that different stimuli presented to the animal were also able to alter the structure of its functional neural networks.
I Average Degree
Considerable alterations on the networks average degree were observed on the five frequency bands analyzed during the experiment (see Fig.1 ; Table 1).
Considering the transition between the periods in which the macaque was with eyes open and blindfolded, alterations on the average connectivity of the networks were verified in some frequency bands (see Fig.1). After the placement of a patch over the eyes, an increase and a higher variation on the values of the average degree of the graphs occurred.
Considering the transition between the awake and anesthetized states, prominent changes in the average degree of the graphs were also observed (see Fig.1), being verified a considerable reduction and a smaller variation of the average degree, the graphs assumed a tendency of possessing the same connectivity over time.
|Average Degree||Eyes Open||Eyes Closed||Anesthesia|
Low Frequencies (0-4Hz)
After the placement of the blindfold, the average degree respective to Delta frequency band, assumed a higher variation and approximately doubled in mean values (see Fig.1 a, Table 1). After the administration of the anesthetics, the degree of the graphs had diminished, keeping approximately constant for about four minutes, later, an increase and a higher variation that lasted around 10 minutes were observed. After that time, the average degree had fallen again keeping approximately constant until the end of the experiment. The networks respective to the awake (blindfolded) and anesthetized states had approximately the same mean connectivity (see Table 1).
Medium Frequencies 4-30Hz
The average degree on Theta and Alpha bands had increased significantly after the placement of the blindfold, having assumed a higher variation, and approximately tripled in mean values (see Table 1). In Beta band, an alike dynamic behavior was also observed, in this frequency band the average connectivity of the graphs increased in the order of 20% due to the placement of the blindfold (see Table 1).
After the administration of the anesthetics, the Theta, Alpha and Beta bands presented a considerable reduction in the average degree, compared to the time when the macaque was awake (blindfolded) (see Fig.1 b, c, d) the average degree was reduced about three times in Theta, eight times in Alpha and six times in Beta band (see Table 1). Under those experimental conditions, the variation of the graphs connectivity had diminished, the networks assumed a quite constant degree along the anesthetized state.
High Frequencies 25-100Hz
Between 25 and 100Hz, after the eyes of the monkey were covered, the average degree increased in the order of 35%, and the connectivity of the networks assumed a higher variation, reflecting in a variance four times higher compared to the time when the macaque eyes were open (see Table 1). After the administration of the anesthetics, the variation of the values decreased, and the average connectivity was reduced by the order of 30% compared to the time when the macaque was awake (eyes closed) (see Table 1).
Ii Correspondence Between Vertices Degree and Anatomical Areas
Through the use of a color gradient, it was possible to verify the relation between the degree of the vertices of the functional networks and their corresponding cortical anatomical areas. It was noted that the majority of the nodes that were physically closer to each other had a tendency of presenting a similar degree, reflecting in an alike color. It was also noted that not all nodes had the same degree, the differences observed seemed to have relation to anatomical areas and divisions (see Fig. 2).
Analysing the networks estimated serially over time, it was possible to observe changes among the patterns from one network to another, revealing that the connectivity of the vertices is dynamic. Besides the existence of some changes in consecutive (time) networks, two evident and distinct patterns were observed along the experiment. The first one respective to the period when the monkey was awake, and the second pattern was observed when the macaque was anesthetized.
ii.1 Pattern of the Awake State
The presence of nodes of high degree over the frontal and parietal lobes characterized the pattern observed in the awake state (see Fig.2 A - E). Highly connected nodes were also observed with a considerable likelihood on the frontal medial wall and anterior parts of the temporal lobe. It was also noticed the existence of areas in which the vertices degree was relatively smaller, those areas comprehended mainly the occipital lobe, occipital medial walls and medial-posterior areas of the temporal lobe.
ii.2 Pattern of the Anesthetized State
The absence of high degree nodes over the entire network characterized the pattern observed during the induced state of anesthesia. Despite the overall decrease in the functional connectivity, at certain moments, some regions presented vertices with higher degree. Those events of connectivity raise were mainly located at the occipital lobe and some times on the frontal and parietal areas (see Fig.2 P - T).
Within about one and a half minutes after the Ketamine-Medetomidine cocktail was injected, an abrupt change on the degree of the vertices was observed, revealing a new distinct pattern that remained present while the macaque was anesthetized. The transition between the two patterns occurred in a rapid way, from the observation of the (Figure 2), it is possible to note that the transition took approximately 20-25 seconds (4-5 networks) (see Fig.2 I - O).
Iii Average Degree - Cortical Lobes
Aimed at observing and comparing the alterations that occurred due to the anesthetic induction in each one of the four cortical lobes, networks respective to each lobe were assembled (corresponding to subgraphs of the total network).
|Average Degree||Eyes Open||Eyes Closed||Anesthesia|
|Subgraph Frontal Lobe||9.80||44.3||6.65||18.0||112||10.6||1.00||0.24||0.49|
|Subgraph Parietal Lobe||3.46||4.41||2.10||6.82||10.1||3.17||0.65||0.32||0.56|
|Subgraph Temporal Lobe||3.75||3.57||1.89||7.52||7.00||2.65||1.72||0.51||0.71|
|Subgraph Occipital Lobe||2.17||1.96||1.40||4.80||12.4||3.52||6.34||10.7||3.27|
Significant alterations in the average degree at each one of the four brain lobes analysed were verified over the experiment (see Fig.3 , Table 2).
The frontal, parietal, and temporal regions presented a quite alike dynamic behavior. On those regions, after the placement of the blindfold over the eyes, an increase and a higher variation in the mean connectivity degree of the networks were observed. A minute and a half after the injection of the anesthetic cocktail, an expressive decrease in the connectivity was noted (see Fig.3 a, b, c), the average degree was reduced 18 times in the the subgraphs of frontal region, 10 times in the parietal lobe and four times on temporal areas; in comparison to the awake state (blindfolded) (see Table 2). A substantial decrease on the variation of the average degree on those cortical lobes was also observed, as the connectivity remained quite constant during general anesthesia (see Fig.3 a, b, c).
A distinct dynamic behavior in response to the experimental conditions was observed on the sub-graphs respective to the occipital lobe. After the eyes of the monkey were covered, the average degree of the sub-graphs almost doubled and also presented a higher variation (see Fig.3 d). After the administration of the anesthetics, a decrease on the mean connectivity that lasted for about seven minutes was verified. Then, the mean connectivity started to increase and also presented a higher variation (see Fig.3 d).
Iv Average Path Length
Significant changes on values of average path length were observed during the anesthetic induction process (see Fig.4 , Table 2).
Considering the effects of blindfolding the monkey, a subtle decrease on the average path length occurred. It was also possible to note that under this experimental condition, there was a tendency for the graphs to present quite the same values, keeping the average path length almost constant, except by short periods when higher values were observed.
Considering the transition between the awake (blindfolded) and the anesthetized state, the anesthetic induction led to an expressive increase in the average path length of the networks, and also a higher variation of this property was verified during general anesthesia (see Fig.4 b - e).
|Average Path||Eyes Open||Eyes Closed||Anesthesia|
Low Frequencies 0-4Hz
In the Delta band, no substantial changes in the average path length were observed along the experiment (see Fig.4 a). After the placement of the blindfold, a decrease in the order of 15% on the average length was registered, compared to the time when the monkey had its eyes open. After the administration of the anesthetics, a decrease in the order of 5% was observed, compared to the awake state (blindfolded) (see Table 3).
Medium Frequencies 4-30Hz
On medium frequencies, the act of blindfolding the macaque led to a slight decrease on the average path length of the networks. A similar behavior was observed in Theta, Alpha and Beta bands during the anesthetic induction process. A minute and a half after the administration of the Ketamine and Medetomidine cocktail, a substantial increase in the average path length in the order of 45% in Theta and 50% in Alpha and Beta occurred (see Table 3; Fig.4 b - d). A higher variation on the average path length was also observed when the monkey was anesthetized.
Higher Frequencies 25-100Hz
On the Gamma frequency band, putting a patch over the eyes of the macaque has led to a higher variation on the average path length of the networks. After the anesthetic injection, the average path length increased, four minutes later some reduction was observed on the values. A decrease in the variation over time of the average path length was also verified during general anesthesia (see Fig.4 e).
Changes in the diameter of the networks were observed along the experiment (see Fig.5; Table 4).
The diameter of the graphs respective to the Delta band presented a distinct dynamic behavior compared to the other frequency bands. From 0 to 4Hz, the anesthetic induction led to a reduction and a decrease on the variation of the diameter (see Fig.5 a).
|Diameter||Eyes Open||Eyes Closed||Anesthesia|
A similar behavior was observed on the Theta, Alpha, Beta and Gamma bands. After blindfolding the macaque, the diameter length subtly decreased, and a slight increase in the variation amplitude of the values was noted. One minute after the administration of the anesthetics, the diameters of the graphs increased substantially, staying this way until the end of the experiment (see Fig.5 b - e).
Vi Average Betweenness Centrality Degree
On all frequency bands analyzed no significant changes on the average betweenness centrality degree of the vertices were observed after blindfolding the monkey (see Fig.6), however, the administration of the anesthetics, promoted changes on this property of the networks.
In the Delta band, the administration of the anesthetics led to a decrease on the vertices average betweenness centrality degree that lasted for about 15 minutes. Then, the values started to increase up to the end of the recording section (see Fig.6 a).
|Centrality||Eyes Open||Eyes Closed||Anesthesia|
In both Theta and Gamma bands, after the administration of the anesthetics, an increase on the average betweenness centrality degree and a reduction in the variation of the values was observed, compared to the awake state (eyes open and blindfolded) (see Fig.6 b, e).
In the Alpha band, after the administration of the anesthetics, the vertices average centrality degree was reduced and an increase in the variation of the values was observed (see Fig.6 c).
On the Beta band one and a half minutes after the anesthetic injection, the average betweenness centrality degree of the nodes was reduced. Within about 15 minutes later, the values started to increase (see Fig.6 d).
Vii Vertices Betweenness Centrality Degree and Anatomical Areas
Through the use of a color gradient, it was possible to verify the correspondence between the intermediation degree of the vertices of the networks and their corresponding anatomical areas.
It was observed that nodes physically closer from each other presented a quite similar intermediation degree. Not all vertices of the networks had the same betweenness centrality degree. It was possible to note that the betweenness centrality degree of the vertices was related to anatomical areas and divisions (see Fig.7).
It was verified that the intermediation degree of the vertices is dynamic, once variations from consecutive (time) networks were noted. Besides the existence of those variations, during the experiment, two prominent patterns were observed, the first respective to the time when the monkey was awake, and the second to when it was anesthetized.
Awake State Pattern
In awake conditions, a high participation of all nodes on geodesic paths was verified. It was possible to notice that some vertices had a higher intermediation degree, and they were frequently located closer from each other. Those vertices of higher degree extended and involved continuous large areas of the cortex (see Fig.7 A-E).
Anesthetized State Pattern
While the macaque was anesthetized, a small number of nodes (cortical areas) had high betweenness centrality degree, while the other nodes of the network had a reduced degree of centrality. It was possible to observe that the intermediation had been monopolized by vertices located in some specific regions. Another remarkable change observed was the discontinuity of the coverage of those areas of high betweenness centrality, that were located far from each other being separated by regions characterized by vertices with reduced intermediation degree (see Fig.7 P - T).
Within about one minute and a half after the administration of the anesthetics, a transition in the patterns of the vertices intermediation degree was observed. The pattern found in the awake state was not observed anymore, giving place to a different pattern that prevailed while the monkey was anesthetized. The transition between the two distinct patterns happened in approximately 40 to 50 seconds (see Fig.7 J - Q).
The placement of a blindfold over the eyes of the monkey did not change the assortative coefficient of the graphs significantly (see Fig.8). Small changes were observed on each frequency band, but no remarkable alteration in the dynamic behavior occurred due to the placement of the blindfold (eyes open and closed).
|Assortativity||Eyes Open||Eyes Closed||Anesthesia|
Significant changes in the assortative character of the networks occurred after the administration of the anesthetics (see Fig.8 ; Table 6). A quite similar behaviour was observed in Delta, Theta and Alpha bands. On those frequency bands while the macaque was awake, the assortativity varied assuming positive and negative values. In average there was a prevalence of a non assortative character. A few seconds after the administration of the anesthetics, an accentuated change occurred, the graphs assumed an assortative character, having positive assortativity (see Fig.8 a, b, c).
In the Beta band, an assortative character prevailed all over the experiment. After the anesthetics were administrated, the assortativity gradually lowered for about 10 minutes. After that time, the assortativity started to increase, and assumed one same dynamic behaviour until the end of the experiment (see Fig.8 d).
While the monkey was awake, the networks respective to the Gamma band were disassortative. Almost right after the administration of the drugs, the functional brain networks started to assume a high variation in the assortativity, being sometimes assortative, and other times disassortative, revealing in this frequency band an expressive dynamics of networks structural alterations during general anesthesia (see Fig.8 e).
Blindfolding the monkey, has led to an increase on both the variation and the mean values of the transitivity coefficient; in comparison to the time when the monkey had the eyes open (see Fig.9; Table 7).
After the anesthetic injection, clear alterations in the transitivity coefficient (Newman, 2001) could be noted on some frequency bands (see Fig.9).
|Transitivity||Eyes Open||Eyes Closed||Anesthesia|
After the administration of the anesthetics, the Delta band presented an increase in the transitivity coefficient, having the variation of the values been kept almost constant (see Fig.9 a).
In the Theta and Alpha bands, during general anesthesia, a decrease in transitivity and also a reduction in the variation of the values occurred (see Fig.9 b, c).
In Beta, one minute and a half after the anesthetic injection, an expressive decline was observed, the transitivity coefficient was relatively smaller compared to the period when the macaque was awake (eyes open and closed), remaining this way while the monkey was anesthetized (see Fig.9 d).
In Gamma a reduction in the values was observed one minute and a half after the administration of the anesthetics, later the transitivity coefficient started to increase. Another remarkable alteration that occurred in this frequency band, was a smaller variation of the values, compared to the period when the monkey was awake (see Fig.9 e).
In the experiment, the administration of the Ketamine-Medetomidine cocktail led to alterations on various networks properties on the five physiological frequency bands analyzed. The most prominent alterations were observed in Theta, Alpha and Beta bands. Those experimental observations indicate that the Ketamine-Medetomidine cocktail affected in a more expressive manner neural activities given between 4 and 30Hz. Once the anesthetic agents promoted physiological changes in the animal model, reducing drastically its cognitive capacity and level of consciousness, those results suggest that neural processes and activities related to cognition and consciousness may be mainly given in frequencies between 4 and 30 Hz.
Changes on many topological properties of the networks occurred within about one and a half minutes after the administration of the anesthetics. This observation of prominent alterations in such a fast and simultaneous way on distinct properties of the graphs highly indicates a phase transition on the networks architecture. The fact of being registered quite after the injection of the anesthetic cocktail, brings high confidence that the changes observed during the experiment are directly related to the physiological effects of the anesthetics on the animal (rapid and expressive reduction in the level of consciousness).
Those results are very interesting according to the perspective of the modern network science, once they bring empirical evidence involving relations between a change in the behavior of a real system and alterations into its respective networks. Being in agreement with one of the greatest premises of complex systems science, which states that the behavior exhibited by a system is intimately related to its structure, that is, with the specific way how the elements of the system establish interactions.
This is one of the first papers to report the existence, and to bring an estimative of the structure and dynamics of large scale functional brain networks respective to the induced state of anesthesia.
I Alterations on the Networks Properties
The average degree reflects the average number of connections of the vertices of the network (Rubinov and Sporns, 2010), bringing information related to the global connectivity of the graph, indicating how interactive the elements of the system are.
The results obtained in this research revealed that after the administration of the anesthetics, an expressive reduction in the average degree respective to Theta, Alpha, and Beta bands occurred (see Fig.1 ; Table 1). From that experimental evidence, it is verified that globally the anesthetic induction leads to a considerable reduction of the cortical functional connectivity, demonstrating that functional interactions established among several areas of the cortex are drastically reduced during general anesthesia.
Vertices Degree and Anatomical Areas
Considering the functional brain networks, the degree of each individual vertex brings information regarding the capacity of the area that it represents in influencing or receiving influence from other distinct areas (Sporns, 2011).
The results obtained in this experiment, demonstrate that in the awake state, high degree vertices are located covering all frontal and parietal lobes, and also in temporal areas anterior to the superior temporal sulcus, indicating that in awake conditions those cortical regions present high connectivity and functional integration. After the administration of the anesthetics, highly connected vertices were not observed on those cortical areas anymore, evidencing the occurrence of a drastic reduction in the functional connectivity. The results reveal that the Ketamine-Medetomidine cocktail affected mainly the great connectivity that existed in areas of the secondary associative cortex of the frontal and parietal lobes, strongly suggesting that events of consciousness may be highly dependent on the high functional integration involving these anatomical regions. Such experimental evidence is in accordance with several scientific reports relating the same regions to neural correlates of consciousness. There are experimental evidences that the administration of the general anesthetics Propofol induces a large decrease in the blood flow on areas of the pre-cuneous, cuneous and posterior cingulate cortex (Fiset et al., 1999), being the inactivation of the posterior medial part associated with loss of consciousness (Kaisti et al., 2002). Damasio and Dolan reported that injuries affecting the same cortical areas are related to serious disturbances of cognition and consciousness (Damasio and Dolan, 1999). Laureys highlighted that fronto-parietal regions are preferentially deactivated in humans patients in vegetative state (Laureys et al., 2004), being the loss of consciousness of those patients associated with the functional disconnectivity between frontal and parietal regions (Laureys et al., 1999).
Average Degree - Cortical Lobes
The analysis of the average degree of the sub-graphs respective to the frontal, parietal, temporal and occipital lobes confirm some features observed in the color gradient representing the degree of the vertices over the position of the electrodes. Those results show that on Theta, Alpha and Beta111Theta and Beta frequency bands data not shown. frequency bands, the anesthetic induction led to an expressive decrease in the functional connectivity of the frontal, parietal and temporal lobes (see Fig.3 ; Table 2).
Unlike what happened in the frontal, parietal and temporal lobes, the average connectivity of the subgraphs respective to the occipital lobe was three times higher222When compared to the awake state (eyes open). during the induced state of anesthesia (see Table 2). Such event was also observed in the gradient of colors representing the degree of the vertices. These results strongly suggest the existence of coherent neural activities in the occipital lobe during Ketamine-Medetomidine induced state of anesthesia.
Average Path Length
According to Latora and Marchiori, the capacity and global efficiency of a network in transmitting information is directly related to the average of its minimum paths, being the most efficient networks, those which have the shortest paths (Latora and Marchiori, 2001).
The experimental results demonstrate that the administration of the anesthetics led to a substantial increase in the average path length (Costa et al., 2007) on the Theta, Alpha and Beta frequency bands (see Fig.4 ; Table 3). Such increase on the average path length strongly indicates that during general anesthesia, the global capacity of transmission of information between several cortical areas is substantially reduced.
The diameter of the network (Costa et al., 2007) as being related only to the larger geodesic path of the graph, is a measure less informative than the average path length, once the latter takes into account all the minimum paths of the network. The diameter reflects the length of the largest minimum path, representing the largest distance existent on the network333Assuming that the graph is connected..
The experimental results reveal a considerable increase in the length of the diameter of the networks on Theta, Alpha and Beta bands, which occurred one and a half minutes after the administration of the anesthetics (see Fig.5). Such result supports the conclusions obtained analysing the average path length, that the global transmission of information is reduced during the induced state of anesthesia.
According to Latora and Marchiori, the local efficiency of transmission of information of a network is directly related to its transitivity coefficient, the larger the coefficient, the greater the local efficiency of the network (Latora and Marchiori, 2003).
The results obtained in this study demonstrate that the administration of the anesthetics led to a reduction of the transitivity coefficient444Mainly when compared to the awake state (eyes closed), in Theta,Alpha and Beta, being the most prominent reduction observed in the Beta band. (see Table 7 ; Fig.9). Such decrease observed, indicates that during the induced state of anesthesia, the efficiency of transmission of information is reduced at local levels.
The assortativity coefficient (Boccaletti et al., 2006) is related to the existence of preferential attachment between vertices with respect to their connectivity degree (Boccaletti et al., 2006). Alterations in this coefficient reveal the existence of structural changes on the graphs, in the way in which are established the connections among the vertices.
The results of this experiment revealed that the administration of the anesthetics led to significant alterations in the assortativity character of the networks. Being the most accentuated changes observed in Delta, Theta and Gamma bands (see Fig.8). These results demonstrate the existence of structural changes in the character of organization of neural activities on those frequency bands. The graphs respective to Delta and Gamma bands, showed no prominent changes in other network measures after the anesthetic induction process. However, alterations in the assortativity coefficient, demonstrate that the anesthetics led to structural rearrangements on the large scale functional brain networks on those frequency bands.
As reported by Costa, the assortativity character of the networks may have a great influence in the dynamic processes supported by the system (Costa et al., 2007). The fact that the networks respective to the induced state of anesthesia assumed a predominantly assortative character, may imply in a higher instability of those networks (Brede and Sinha, 2005), and a reduction on their synchronization capacity (di Bernardo et al., 2005).
Intermediation Degree and Anatomical Areas
The betweenness centrality degree of a vertex (Freeman, 1979) is related to the importance of this vertex given its participation in the minimum paths of the network. The larger the number of minimum paths passing through a node, the larger the impact on the integration of the network performed by this vertex. The analysis of the relation between the intermediation degree of the vertices and anatomical areas, brings an estimative about how the transmission of information on the network is performed, highlighting the areas that monopolize the flux of information and perform an important role in the functional integration.
The results obtained experimentally demonstrate that during awake conditions (eyes open and closed), the vast majority of the vertices of the graph had a high intermediation degree (see Fig.7), indicating that the flow of information on the network is in a certain way distributed between several cortical areas. Those results suggest that during awake conditions the vast majority of the cortex gives support and possibly is actively involved in the transmission of information. Another remarkable characteristic is that vertices of high intermediation degree showed up extending continuously covering large cortical areas, without being separated by areas characterized vertices possessing a low intermediation degree (see Fig.7 a - e).
After the administration of the anesthetics, a considerable change regarding the intermediation degree of the vertices and their respective cortical areas occurred (see Fig.7). It was possible to observe that some areas monopolized the integration of the network, and that the vast majority of the cortex presented a reduced degree of intermediation, revealing that the structure of the network did not provide the distribution of the flow of information among “the entire” cortex anymore. Another remarkable characteristic observed was the discontinuity of the occupancy of the areas having a high degree of intermediation, being those physically segregated by extensive regions characterized by vertices possessing a low intermediation degree (see Fig.7 P - T). From the analysis of the results it is possible to intuit the hypothesis that probably activities and neural processes related to conscious experiences may require the participation of the vast majority of the cortex on the integration of the network as a whole.
Average Betweenness Centrality Degree
A direct association between the average intermediation degree of the vertices and functional or structural properties of the network is not trivial, once different alterations in the topology of the graph may lead to similar changes in the mean intermediation degree of the vertices. Besides not being possible to comprehend in a straightforward manner the meanings and the implications of the alterations observed on this property, the existence of changes confirm the occurrence of structural alterations on the graphs.
In this experiment, after the administration of the anesthetics considerable changes were observed on the average betweenness centrality degree in Theta, Alpha, Beta and Gamma bands (see Table 5 ; Fig.6), confirming the occurrence of structural rearrangements on the large scale functional brain networks during the induced state of anesthesia.
Ii Alterations due to Anesthesia on Small World Architecture
Regarding the five frequency bands analyzed, the most prominent alterations observed during the anesthetic induction, were verified on the networks respective to the Theta, Alpha and Beta bands. In general lines, those frequency bands presented mainly the same kind of alterations:
An expressive reduction on the average degree of the vertices.
A considerable increase in the average path length.
A decrease in the transitivity coefficient.
The considerable reduction in the average degree and transitivity, and the great increase in the average path length, have direct consequence, impacting the small world architecture (Watts and Strogatz, 1998) observed during the awake state. According to Olaf Sporns, the small world architecture, that is characterized by high values of the transitivity coefficient and a small average path length, provides an structural substrate of great relevance, being important in many aspects of the brain functional organization. This architecture supports the processing of information segregated locally and integrated globally (Sporns, 2011). The small world architecture is also considered to promote efficiency of transmission and processing of information, wiring economy and to give support to a diverse and complex dynamic of the network (Bullmore and Sporns, 2012).
Several authors consider the segregation and integration as two of main principles of organization of the activities that occur in the cerebral cortex (Zeki, 1978; Zeki and Shipp, 1988; Tononi et al., 1994; Tononi and Edelman, 1998; Friston, 2002, 2005, 2009). Sporns goes even beyond and highlights that the balance between those two properties constitute a key mechanism necessary for the brain in order to perform its activities (Sporns, 2011).
The results of this study demonstrate that on the frequency bands Theta, Alpha and Beta, two of the properties of the network that were most impacted after the anesthetic induction were the integration capacity at global levels (increase in the average path length) and integration capacity at local levels (decrease in the transitivity coefficient). Those factors have direct consequence on the breakdown of the small world architecture that was observed in the awake state. Those experimental results are in accordance and support what has been pointed out by many authors, relating the loss of key structural properties of functional networks to alterations and suppression of brain capacities.
Alterations Observed After the Placement of the Blindfold
Several alterations on the topology and dynamic behavior of the properties of the large scale functional brain networks were observed after the placement of a blindfold over the eyes of the macaque, being possible from the figures and graphs, to note clear distinctions regarding the time when the monkey had its eyes open and when it had the eyes closed.
The researchers that realized the experiments and provided the database of the neural activity records, did not provide enough information regarding the placement of the blindfold on the state of the animal. The effects may depend on several factors, such as, the conditioning of the macaque to the experimental conditions, the relationship of the animal with the researchers, the handling used during the experiment, and others. Placing a blindfold over the eyes of the monkey, being the animal restrained in a chair, might have led to a calm and relaxed state or might have led to fear and apprehension. Thus, conclusions regarding the relation of the network measures and the specific state in which the monkey was, can only be drawn under a complete detailed description of the experiment.
The fact that alterations were observed on the networks measures right after blindfolding the monkey, demonstrate that the functional networks are dependent on the conditions presented to the animal. Those phenomena observed are in accordance to some suppositions of Olaf Sporns that stated that functional connectivity may vary considerably over time, being modulated by demands due to different types of activities or due to different sensory stimuli presented (Sporns, 2011).
Point of Loss of Consciousness (LOC)
The scientists that realized the experimental procedures and recorded the data, reported the occurrence of the loss of consciousness point (LOC), at the moment when the animal no longer responded to stimuli such as touching the nostrils and the hands. However, no prominent alterations on network properties were observed at the (LOC), which was registered by the researchers within about ten minutes after the administration of the anesthetics. Probably the end of the awake state took place when several changes on the properties of the networks were observed, which occurred in less than two minutes after the anesthetic injection. However, after that time, the monkey was still able to present involuntary reactions to stimuli, being in an intermediary condition between deep sedation and general anesthesia, probably in a similar condition as the pharmacological effects of the administration o the Ketamine in Humans patients are described (Bergman, 1999).
The supposed view of general anesthesia being given by a "whole brain shutdown" is not supported by the experimental results of this study. It was possible to observe that during the Ketamine-Medetomidine induced anesthesia, the brain entered into a highly specific, complex and dynamic state.
From the modeling of the interactions established among several cortical areas through the use of graphs and complex networks, it was possible to verify that the networks respective to induced state of anesthesia had shown structurally distinct from the networks respective to the awake state. Those results reveal that anesthetics are able to impact several properties of the large scale functional brain networks, resulting in graphs of different architectures. Those changes observed in the experiment bring indications that the behavior exhibited and the processes supported by the brain may have a direct relation to structural properties of large scale functional networks.
It was possible to observe that functional neural activities are dynamic, considerable changes were verified on the network measures in short time intervals. The observation of the degree of the vertices through the use of a color gradient also revealed the existence of changes in the patterns in short intervals of time. Such evidence demonstrate that functional neural activities are not static, they reveal that the brain functional organization is constantly changing in short intervals of time555In both awake and anesthetized conditions..
The Ketamine-Medetomidine cocktail did not alter the functional activities in only some specific and restrict areas of the cortex, a general change in the state of the brain in which all the cortex presented alterations in functional connectivity was observed.
The greatest changes observed on the large scale functional brain networks due to the anesthetic induction were an accentuated decrease in functional connectivity, and a reduction in the capacities of integration of the networks in both global and local levels.
From the characterization of the functional networks during the anesthetic induction process, it was possible to verify a transition between the awake and anestetized states. The transition occurred in a quite fast manner, it took about 20 to 30 seconds.
This manuscript contains results respective to a single experiment involving one macaque. The results may be validated, in a similar experiment involving a second macaque of the same species.
Statistical analyses will be performed in order to provide intervals of confidence regarding the results.
Those procedures will be performed and included on future versions of the manuscript.
All the experimental procedures involving the animal model and data record (Methods Sections II and III), were idealized and performed by researchers from the laboratory of adaptive intelligence at the
RIKEN BRAIN SCIENCE INSTITUTE, laboratory under supervision of the PhD. Naotaka Fujii.
Eduardo C. Padovani worked with the neural records database (Methods Section IV) . Idealized and performed the procedures, analyzed the results and wrote the manuscript.
This research was partially financed by CAPES (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior).
During part of the development of the study, the author used the structure of the Laboratory Vision-eScience at Universidade de São Paulo, laboratory supported by FAPESP grant 2011/50761-2, CNPq, CAPES, NAP-eScience, PRP and USP.
- Alkire (1998) Michael T Alkire. Quantitative eeg correlations with brain glucose metabolic rate during anesthesia in volunteers. Anesthesiology, 89(2):323–333, 1998.
- Alkire et al. (1995) Michael T Alkire, Richard J Haier, Steven J Barker, Nitin K Shah, Joseph C Wu, and Y James Kao. Cerebral metabolism during propofol anesthesia in humans studied with positron emission tomography. Anesthesiology, 82(2):393–403, 1995.
- Alkire et al. (2000) MT Alkire, RJ Haier, and JH Fallon. Toward a unified theory of narcosis: brain imaging evidence for a thalamocortical switch as the neurophysiologic basis of anesthetic-induced unconsciousness. Consciousness and cognition, 9(3):370–386, 2000.
- Bergman (1999) Stewart A Bergman. Ketamine: review of its pharmacology and its use in pediatric anesthesia. Anesthesia progress, 46(1):10, 1999.
- Boccaletti et al. (2006) Stefano Boccaletti, Vito Latora, Yamir Moreno, Martin Chavez, and D-U Hwang. Complex networks: Structure and dynamics. Physics reports, 424(4):175–308, 2006.
- Brede and Sinha (2005) Markus Brede and Sitabhra Sinha. Assortative mixing by degree makes a network more unstable. arXiv preprint cond-mat/0507710, 2005.
- Bullmore and Sporns (2009) Ed Bullmore and Olaf Sporns. Complex brain networks: graph theoretical analysis of structural and functional systems. Nature Reviews Neuroscience, 10(3):186–198, 2009.
- Bullmore and Sporns (2012) Ed Bullmore and Olaf Sporns. The economy of brain network organization. Nature Reviews Neuroscience, 13(5):336–349, 2012.
- Costa et al. (2007) L da F Costa, Francisco A Rodrigues, Gonzalo Travieso, and Paulino Ribeiro Villas Boas. Characterization of complex networks: A survey of measurements. Advances in Physics, 56(1):167–242, 2007.
- Cui et al. (2008) Jie Cui, Lei Xu, Steven L Bressler, Mingzhou Ding, and Hualou Liang. Bsmart: a matlab/c toolbox for analysis of multichannel neural time series. Neural Networks, 21(8):1094–1104, 2008.
- Damasio and Dolan (1999) Antonio Damasio and Raymond J Dolan. The feeling of what happens. Nature, 401(6756):847–847, 1999.
- di Bernardo et al. (2005) Mario di Bernardo, Franco Garofalo, and Francesco Sorrentino. Synchronization of degree correlated physical networks. arXiv preprint cond-mat/0506236, 2005.
- Fiset et al. (1999) Pierre Fiset, Tomás Paus, Thierry Daloze, Gilles Plourde, Pascal Meuret, Vincent Bonhomme, Nadine Hajj-Ali, Steven B Backman, and Alan C Evans. Brain mechanisms of propofol-induced loss of consciousness in humans: a positron emission tomographic study. The Journal of neuroscience, 19(13):5506–5513, 1999.
- Flohr (1995) Hans Flohr. An information processing theory of anaesthesia. Neuropsychologia, 33(9):1169–1180, 1995.
- Freeman (1979) Linton C Freeman. Centrality in social networks conceptual clarification. Social networks, 1(3):215–239, 1979.
- Friston (2002) Karl Friston. Beyond phrenology: what can neuroimaging tell us about distributed circuitry? Annual review of neuroscience, 25(1):221–250, 2002.
- Friston (1994) Karl J Friston. Functional and effective connectivity in neuroimaging: a synthesis. Human brain mapping, 2(1-2):56–78, 1994.
- Friston (2005) Karl J Friston. Models of brain function in neuroimaging. Annu. Rev. Psychol., 56:57–87, 2005.
- Friston (2009) Karl J Friston. Modalities, modes, and models in functional neuroimaging. Science, 326(5951):399–403, 2009.
- Friston et al. (1993) KJ Friston, CD Frith, PF Liddle, and RSJ Frackowiak. Functional connectivity: the principal-component analysis of large (pet) data sets. Journal of Cerebral Blood Flow & Metabolism, 13(1):5–14, 1993.
- Fukushima et al. (2014) Makoto Fukushima, Richard C Saunders, Matthew Mullarkey, Alexandra M Doyle, Mortimer Mishkin, and Naotaka Fujii. An electrocorticographic electrode array for simultaneous recording from medial, lateral, and intrasulcal surface of the cortex in macaque monkeys. Journal of neuroscience methods, 233:155–165, 2014.
- Granger (1969) Clive WJ Granger. Investigating causal relations by econometric models and cross-spectral methods. Econometrica: Journal of the Econometric Society, pages 424–438, 1969.
- Green et al. (2011) Steven M Green, Mark G Roback, Robert M Kennedy, and Baruch Krauss. Clinical practice guideline for emergency department ketamine dissociative sedation: 2011 update. Annals of emergency medicine, 57(5):449–461, 2011.
- Hameroff (2006) Stuart Hameroff. The entwined mysteries of anesthesia and consciousness. Anesthesiology, 105(2):400–12, 2006.
- Hameroff et al. (1998) Stuart R Hameroff, Alfred W Kaszniak, and Alwyn Scott. Toward a science of consciousness II: The second Tucson discussions and debates, volume 2. Mit Press, 1998.
- Hamilton (1989) James D Hamilton. A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica: Journal of the Econometric Society, pages 357–384, 1989.
- Imas et al. (2005) Olga A Imas, Kristina M Ropella, B Douglas Ward, James D Wood, and Anthony G Hudetz. Volatile anesthetics disrupt frontal-posterior recurrent information transfer at gamma frequencies in rat. Neuroscience letters, 387(3):145–150, 2005.
- Iriki and Sakura (2008) Atsushi Iriki and Osamu Sakura. The neuroscience of primate intellectual evolution: natural selection and passive and intentional niche construction. Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1500):2229–2241, 2008.
- Isa et al. (2009) Tadashi Isa, Itaru Yamane, Miya Hamai, and Haruhisa Inagaki. Japanese macaques as laboratory animals. Experimental Animals, 58(5):451–457, 2009.
- John and Prichep (2005) E Roy John and Leslie S Prichep. The anesthetic cascade: a theory of how anesthesia suppresses consciousness. Anesthesiology, 102(2):447, 2005.
- Kaisti et al. (2002) Kaike K Kaisti, Liisa Metsähonkala, Mika Teräs, Vesa Oikonen, Sargo Aalto, Satu Jääskeläinen, Susanna Hinkka, and Harry Scheinin. Effects of surgical levels of propofol and sevoflurane anesthesia on cerebral blood flow in healthy subjects studied with positron emission tomography. Anesthesiology, 96(6):1358–1370, 2002.
- Kennedy and Norman (2005) Donald Kennedy and Colin Norman. What don’t we know? Science, 309(5731):75–75, 2005.
- Kwiatkowski et al. (1992) Denis Kwiatkowski, Peter CB Phillips, Peter Schmidt, and Yongcheol Shin. Testing the null hypothesis of stationarity against the alternative of a unit root: How sure are we that economic time series have a unit root? Journal of econometrics, 54(1):159–178, 1992.
- Latora and Marchiori (2001) Vito Latora and Massimo Marchiori. Efficient behavior of small-world networks. Physical review letters, 87(19):198701, 2001.
- Latora and Marchiori (2003) Vito Latora and Massimo Marchiori. Economic small-world behavior in weighted networks. The European Physical Journal B-Condensed Matter and Complex Systems, 32(2):249–263, 2003.
- Laureys et al. (1999) Steven Laureys, Serge Goldman, Christophe Phillips, Patrick Van Bogaert, Joël Aerts, André Luxen, Georges Franck, and Pierre Maquet. Impaired effective cortical connectivity in vegetative state: preliminary investigation using pet. Neuroimage, 9(4):377–382, 1999.
- Laureys et al. (2004) Steven Laureys, Adrian M Owen, and Nicholas D Schiff. Brain function in coma, vegetative state, and related disorders. The Lancet Neurology, 3(9):537–546, 2004.
- Lewis et al. (2012) Laura D Lewis, Veronica S Weiner, Eran A Mukamel, Jacob A Donoghue, Emad N Eskandar, Joseph R Madsen, William S Anderson, Leigh R Hochberg, Sydney S Cash, Emery N Brown, et al. Rapid fragmentation of neuronal networks at the onset of propofol-induced unconsciousness. Proceedings of the National Academy of Sciences, 109(49):E3377–E3386, 2012.
- Mashour (2004) George A Mashour. Consciousness unbound: toward a paradigm of general anesthesia. 2004.
- Mashour (2006) George A Mashour. Integrating the science of consciousness and anesthesia. Anesthesia & Analgesia, 103(4):975–982, 2006.
- Mitchell (2009) Melanie Mitchell. Complexity: A guided tour. Oxford University Press, 2009.
- Nagasaka et al. (2011) Yasuo Nagasaka, Kentaro Shimoda, and Naotaka Fujii. Multidimensional recording (mdr) and data sharing: an ecological open research and educational platform for neuroscience. PloS one, 6(7):e22561, 2011.
- Newman (2001) Mark EJ Newman. The structure of scientific collaboration networks. Proceedings of the National Academy of Sciences, 98(2):404–409, 2001.
- Newman (2003) Mark EJ Newman. The structure and function of complex networks. SIAM review, 45(2):167–256, 2003.
- Rubinov and Sporns (2010) Mikail Rubinov and Olaf Sporns. Complex network measures of brain connectivity: uses and interpretations. Neuroimage, 52(3):1059–1069, 2010.
- Schwartz et al. (2010) Robert S Schwartz, Emery N Brown, Ralph Lydic, and Nicholas D Schiff. General anesthesia, sleep, and coma. New England Journal of Medicine, 363(27):2638–2650, 2010.
- Seth (2010) Anil K Seth. A matlab toolbox for granger causal connectivity analysis. Journal of neuroscience methods, 186(2):262–273, 2010.
- Seth and Edelman (2007) Anil K Seth and Gerald M Edelman. Distinguishing causal interactions in neural populations. Neural computation, 19(4):910–933, 2007.
- Sporns (2011) Olaf Sporns. Networks of the Brain. MIT press, 2011.
- Stam and Van Straaten (2012) CJ Stam and ECW Van Straaten. The organization of physiological brain networks. Clinical Neurophysiology, 123(6):1067–1087, 2012.
- Stam and Reijneveld (2007) Cornelis J Stam and Jaap C Reijneveld. Graph theoretical analysis of complex networks in the brain. Nonlinear biomedical physics, 1(1):3, 2007.
- Strogatz (2001) Steven H Strogatz. Exploring complex networks. Nature, 410(6825):268–276, 2001.
- Tononi and Edelman (1998) Giulio Tononi and Gerald M Edelman. Consciousness and complexity. Science, 282(5395):1846–1851, 1998.
- Tononi et al. (1994) Giulio Tononi, Olaf Sporns, and Gerald M Edelman. A measure for brain complexity: relating functional segregation and integration in the nervous system. Proceedings of the National Academy of Sciences, 91(11):5033–5037, 1994.
- Uhrig et al. (2014) L Uhrig, S Dehaene, and B Jarraya. Cerebral mechanisms of general anesthesia. In Annales francaises d’anesthesie et de reanimation, volume 33, pages 72–82. Elsevier, 2014.
- Watts and Strogatz (1998) Duncan J Watts and Steven H Strogatz. Collective dynamics of ”small-world” networks. Nature, 393(6684):440–442, 1998.
- White and Alkire (2003) Nathan S White and Michael T Alkire. Impaired thalamocortical connectivity in humans during general-anesthetic-induced unconsciousness. Neuroimage, 19(2):402–411, 2003.
- Yanagawa et al. (2013) Toru Yanagawa, Zenas C Chao, Naomi Hasegawa, and Naotaka Fujii. Large-scale information flow in conscious and unconscious states: an ecog study in monkeys. PloS one, 8(11):e80845, 2013.
- Young et al. (1999) SS Young, AM Schilling, S Skeans, and G Ritacco. Short duration anaesthesia with medetomidine and ketamine in cynomolgus monkeys. Laboratory animals, 33(2):162–168, 1999.
- Zeki and Shipp (1988) Semir Zeki and Stewart Shipp. The functional logic of cortical connections. Nature, 1988.
- Zeki (1978) Semir M Zeki. Functional specialisation in the visual cortex of the rhesus monkey. Nature, 274(5670):423–428, 1978.