Final Report Summary - DYSTRUCTURE (The Dynamical and Structural Basis of Human Mind Complexity: Segregation and Integration of Information and Processing in the Brain)
Perceptions, memories, emotions, and everything that makes us human, demand the flexible integration of information represented and computed in a distributed manner. The human brain is structured in a large number of areas in which information and computation is highly segregated. Normal brain function requires the integration of functionally specialized but widely distributed brain areas. Furthermore, human behaviour entails a flexible task-dependent interplay between different subsets of these brain areas in order to integrate them according to the corresponding goal-directed requirements. Hence, the understanding of the mechanisms underlying the integration of distributed segregated representations is a central question for the entire field of cognition and brain function.
In this project, we studied how whole-brain computational modelling based on and constrained by multimodal neuroimaging and neurophysiological data can be used to gain new insights into segregation and integration of information in the brain. We investigated the key findings of resting-state activity covering a range of neurophysiological recordings in animals (multi–unit recordings, LFP) and neuroimaging modalities (fMRI, EEG and MEG) both in animals and in humans. We described how to best define and analyze anatomical and functional brain networks. The main premise of our models comes from statistical physics where it has been shown that macroscopic physical systems obey laws that are independent of their mesoscopic constituents. One of the main difficulties of computational brain modeling is to strike the best balance between complexity and realism. Given the astronomical number of neurons in the human brain and the lack of accurate information of specific connectivity at the neural level, it is neither feasible, nor desirable, to create intricate models of, say, each individual neuron and its connections. Instead, whole-brain computational models have typically used various mesoscopic top-down approximations of the underlying complexity. The dynamics of a whole-brain computational model use the anatomical structural connectivity between brain regions in a given parcellation as a description of the synaptic connections between neurons in those areas. These interregional connections are weighted by the strength specified in the structural connectivity matrix and by control parameters of the conductivity of the fibers. These parameters can then be varied systematically to simulate and compare the dynamics of the whole brain with functional data from neuroimaging and neurophysiological experiments. This functional data contains highly structured spatiotemporal activity patterns that emerge across the brain at rest. Whole-brain computational models can thus give a mechanistic explanation of the origin of normal spontaneous resting state networks. Several of our studies have successfully done so for fMRI, MEG, EEG, multiple cell recording and LFP data in both healthy humans and animals. We further extended the results to consider task/stimuli evoked-activity (i.e. cognitive tasks).
We then investigated how unbalancing these networks may lead to problems with mental health. The explicit linkage of human neuroimaging data with whole-brain computational modeling has shown great potential not only for a deeper understanding of the computational and biophysical mechanisms underlying healthy resting state and task/stimuli-evoked activity, but also for the discovery of the causes of the breakdown in neuropsychiatric disorders. This mechanistic information is then useful as biomarkers for individual patients. In addition, this information can be used to monitor the progress for existing therapies, helping to predict the outcome at an early stage, which opens the possibility of tailoring specific treatments to specific patient groups in a stratified neuropsychiatry. Importantly, this will also help our understanding of the origin and mechanistic causes of disease and open up for novel interventions and treatments.
In sum, we believe that the research achievements deepen our understanding of a fundamental question in the entire field of cognition and brain function, namely to identify the mechanisms underlying the integration of distributed segregated representations. Furthermore, as many cognitive processes involve distributed activity across the brain, understanding the dynamics at large scale is indeed a key to understand the breakdown of those processes in neuropsychiatric diseases.
In this project, we studied how whole-brain computational modelling based on and constrained by multimodal neuroimaging and neurophysiological data can be used to gain new insights into segregation and integration of information in the brain. We investigated the key findings of resting-state activity covering a range of neurophysiological recordings in animals (multi–unit recordings, LFP) and neuroimaging modalities (fMRI, EEG and MEG) both in animals and in humans. We described how to best define and analyze anatomical and functional brain networks. The main premise of our models comes from statistical physics where it has been shown that macroscopic physical systems obey laws that are independent of their mesoscopic constituents. One of the main difficulties of computational brain modeling is to strike the best balance between complexity and realism. Given the astronomical number of neurons in the human brain and the lack of accurate information of specific connectivity at the neural level, it is neither feasible, nor desirable, to create intricate models of, say, each individual neuron and its connections. Instead, whole-brain computational models have typically used various mesoscopic top-down approximations of the underlying complexity. The dynamics of a whole-brain computational model use the anatomical structural connectivity between brain regions in a given parcellation as a description of the synaptic connections between neurons in those areas. These interregional connections are weighted by the strength specified in the structural connectivity matrix and by control parameters of the conductivity of the fibers. These parameters can then be varied systematically to simulate and compare the dynamics of the whole brain with functional data from neuroimaging and neurophysiological experiments. This functional data contains highly structured spatiotemporal activity patterns that emerge across the brain at rest. Whole-brain computational models can thus give a mechanistic explanation of the origin of normal spontaneous resting state networks. Several of our studies have successfully done so for fMRI, MEG, EEG, multiple cell recording and LFP data in both healthy humans and animals. We further extended the results to consider task/stimuli evoked-activity (i.e. cognitive tasks).
We then investigated how unbalancing these networks may lead to problems with mental health. The explicit linkage of human neuroimaging data with whole-brain computational modeling has shown great potential not only for a deeper understanding of the computational and biophysical mechanisms underlying healthy resting state and task/stimuli-evoked activity, but also for the discovery of the causes of the breakdown in neuropsychiatric disorders. This mechanistic information is then useful as biomarkers for individual patients. In addition, this information can be used to monitor the progress for existing therapies, helping to predict the outcome at an early stage, which opens the possibility of tailoring specific treatments to specific patient groups in a stratified neuropsychiatry. Importantly, this will also help our understanding of the origin and mechanistic causes of disease and open up for novel interventions and treatments.
In sum, we believe that the research achievements deepen our understanding of a fundamental question in the entire field of cognition and brain function, namely to identify the mechanisms underlying the integration of distributed segregated representations. Furthermore, as many cognitive processes involve distributed activity across the brain, understanding the dynamics at large scale is indeed a key to understand the breakdown of those processes in neuropsychiatric diseases.