Skip to main content

"Investigation of the interaction between external stimulation and ongoing brain activity in cortical networks: analysis, modeling and empirical corroboration"

Final Report Summary - INTERACTIONS (Investigation of the interaction between external stimulation and ongoing brain activity in cortical networks: analysis, modeling and empirical corroboration)

For the most part of the 20th century a notion of brain function has prevailed based in the idea that each part of the brain specialises in processing particular functions. This perspective has been well founded by physiological, surgical and lesion studies. A prominent problem of this perspective, however, is that it is very difficult to imagine how a collection of specialised functions alone could give rise to a coherent perception of the reality and consciousness. For that to happen it is also necessary that the different parts of the brain communicate with each other to share and to combine their information. During the last decades a richer perspective of the brain has been introduced in which networks of segregated but interacting processes govern neural dynamics on top of the processing of the specialised regions. This capacity of the brain to process information of different modalities separately (e.g. visual, auditory or olfactory) and to simultaneously combine that information is known as the balancing between segregation and integration.

The last two decades have seen many advances to observe the brain “at work”. Classical electroencephalography (EEG) allowed to register the electrical activity of the brain with very good temporal resolution but we had very poor precision of where that activity was coming from. Neuroimaging techniques such as functional magnetic resonance imaging (fMRI) allow today to look inside the brain and detect where activity happens, at the cost of low temporal resolution, approximately one snapshot every two or three seconds. Another contribution of imaging techniques is the posibility of measuring the diffusion of water molecules in the brain. Using that information neuroscientists use computer models to guess the direction and starting points of the communication pathways between distant brain regions. Investigation of the properties of those networks of communication have helped to understand the properties of the brain regarded as a network. It has been found that the regions of the brain are compartamentalised into modules, that is, the regions form densely interconnected groups which are less connected with the regions in other groups. This division is a structural characteristic that supports segregation of information into parts with specialised function. It does not explain, however, where the combination and the integration of the multisensory information could happen. The identification of highly connected brain regions, usually refered as the hubs, which span their connections along diverse areas of the cortex have been identified. Their capacity to access information from different specialised modules makes them very good candidates to perform the integration of multisensory information.

These network properties of the human brain are also shared by the brains of mice, cats and macaque monkeys. Even a worm, the C. elegans, whose whole nervous system is composed of only 302 neurones, shares those features. In the Marie-Curie project INTERACTIONS, funded by the European Comission through the 7th People’s Framework, we have further investigated the plausibility of those implications. For that we have targeted at particular questions related to the problem, e.g. what kind of dynamical system is the brain and how does the underlying network of communications support those dynamics? what kind of network topologies are optimal to host the coexistence of segregated/specialised information processes and their integration?

To answer those questions we have investigated the structural connectivity of the brains of cats, macaque monkeys and humans, we have devolped dynamical models to simulate their activity and we have compared the results with empirical observations of the human brain’s dynamics using functional magnetic resonance data. We have found that the class of modular and hierarchical network structures to which the brain network belong are optimal structures to host complex dynamical processes in comparison with other hierarchical structures. For that we have developed new measures to quantify the interaction complexity of networked dynamical systems. These can be applied both to empirical data such as fMRI and to the results of models.

Within a modular and hierarchical network every node plays a different function. A challenge we faced was how to identify the role every node is playing in the network and how to classify them accordingly. In this scenario the importance of a node is not only characterised by its number of connections, its functional category is more determined by how its connections are spanned along the many modules. Hence, we observe that brain regions that are specialised in processing one type of sensory information, e.g. visual, auditory or sensory-motor, rarely connect with regions of other modalities, while the hubs are connected to regions of all sorts of functional specialisations. We have developed network analysis tools to determine and classify the roles of the nodes in a network according to those characteristics.

So far, we have discussed the properties and the organisation of the communication pathways between brain regions. However, they exist to support the dynamical processes happening in the brain which we can measure via fMRI or EEG. So, the following question is to identify which kind of dynamical system the brain is and how the underlying physical network of communications support its dynamical state. To answer these questions we have investigated fMRI data of healthy subjects at rest, that is, when they are comfortably lying in a scanner with their eyes closed and are asked to perform no particular cognitive task, simply to relax without falling asleep during five to ten minutes. We have analysed the resting-state dynamics by applying for the first time in the field recurrence analysis, a set of time-series analysis tools well-known for nonlinear dynamics and chaos. The purpose of recurrence analysis is, as its name states, to investigate whether a dynamical system randomly wanders exploring the vast amount of states it can access, or if in the contrary, the system bounds itself toaccess only a finite set of states. In the latter case, the system returns from time to time to states which it already visited before. Our findings suggest that the brain dynamics are partially recurrent favouring a rich but finite set of states despite the almost infinite number of possible states it could set itself into. This finite but rich dynamical behaviour is the signature, as quantified by the complexity explored before, of the underlying hierarchical network topology organised into modules of specialised brain regions interconnected via hub regions who can “listen” to the information from different sensory modalities by widely spaning their connections and who can share that information with the other hubs.

In the course of our investigations we have developed computational tools to analyse complex networks. We have written a library in the Python programming language, GAlib (Graph Analysis library) and we have made those tools freely available for anyone to use and/or to further develop:

The results obtained during the INTERACTIONS project strongly deepen in our understanding about the organisation of the brain, its dynamical behaviour and the manner we create models to reproduce its activity. The shift in paradigm that has posed the interpretation of the brain as a set of interacting and mutually “speaking” regions contrasts but completes the prevalent perspective of the brain as if every function where perfectly localised. In the future this understanding will have extensive medical applications by providing practitioners and pharmacologists with a more hollistic view of the brain. It is well-known that several mental disorders are not caused by the degradation of a particular region of the brain, but by a disturbance in the communication capabilities between them. Pharmacological treatments of the future can benefit of this knowledge and lead to the design of treatments that better target the source of the neural diseases.

We have found that evolution has shaped neural and brain networks into a form that optimises their complexity to balance between their capabilities to perform several tasks simultaneously by segregating specialised agents and to integrate the different sources of information towards coordination and consciousness. Such network configuration into modules that are interconnected through a set of interconnected hubs with access and understanding of the information from the many sources, could become a proxy for the structure of human organisations. Finding the ideal organigram of a bussiness or of a political/administrative organisation is the source of constant debate in administration schools. Maybe, there are a few lessons to learn from the shape that nature has found optimal for the brain to achieve its goals: to acquire information through several channels (sensory systems), to extract as much information from each channel as possible, and to put it together (integrate it) towards a holistic perspective and a coherent behaviour.