Final Report Summary - ROUTES TO AROUSAL (Routes to arousal: a simultaneous EEG-FMRI investigation of pharmacological sedation in humans)
Functional magnetic resonance imaging (fMRI) studies can reveal connectivity between signals from different brain regions. This permits spatial and temporal characterisation of cortical and subcortical functional circuits during wakefulness and sedation. Combining information from simultaneous electroengephalogram (EEG) and fMRI measurements affords a still finer physiological fractionation of brain networks. Our main aim is to use simultaneous EEG-fMRI to examine altered functional connectivity during pharmacologically altered arousal in healthy human volunteers.
We used functional connectivity (FC) to investigate possible alterations induced by mild propofol sedation on resting state fMRI data. We tested for changes in cortico-cortical and cortico-subcortical FC, using regions of interest that encompass major subdivisions of the cortex (frontal, temporal, parietal, occipital). In particular, prefrontally connected portions of the thalamus were shown to play an important role in maintaining high the synchronisation between cortical regions whose connectivity may be reduced by the sedation. We found that the thalamus may be part of a network (including the cingulate cortex and a portion of brainstem), which coordinates the different stages of the descent to an unconscious state. The method we used to investigate cortical functional connectivity suggests that the temporal cortex is more sensitive to local connectivity variations, than other regions reported in previous studies. These results were presented as an electronic poster at the 20th ISMRM Annual Meeting (5 - 11 May 2012, Melbourne, Australia).
The main limitation of standard functional connectivity analysis resides in the bias introduced every time a region of interest (ROI) is defined. The definition of the sise and the location of an ROI introduces a considering amount of variability across subjects and conditions. To overcome this issue we decided to use graph theory assisted functional connectivity analysis instead of simple cross correlation analysis. The use of graph theory has the great advantage to avoid any operator dependent choice of ROIs needed for cross correlation: each voxel of the brain is considered as an ROI. Obviously, removing the constraint of averaging timeseries in specific regions increases enormously the amount of data to handle (around 40 000 timeseries considered simultaneously!). Fortunately, the workstation bought using part of the budget provided by the fellowship, allowed an easy solution of this problem. Graph theory analysis returned interesting results. We considered brain voxels as nodes of a complex network and measured eigenvector centrality (EC) to characterise brain network properties. The EC mapping of fMRI data in healthy humans under propofol mild sedation, demonstrated a decrease of centrality of the thalamus versus an increase of centrality within the pons of the brainstem, highlighting the important role of these two structures in regulating consciousness. Specifically, the decrease of thalamus centrality results from its disconnection from a widespread set of cortical and subcortical regions, while the increase of brainstem centrality may be a consequence of its increased influence over a few highly central cortical regions key to the default mode network such as the posterior and anterior cingulate cortices. These results were shown as traditional poster at the OHBM 2012 Annual Meeting (10 - 14 June 2012, Beijing, China), and submitted as a paper to the Journal of Neuroscience. The paper has been accepted subject to revisions.
The last period of the fellowship was devoted to the analysis of resting state EEG-fMRI data that we collected during wakefulness and mild propofol sedation. Instead of investigating propofol induced alterations in EEG-fMRI functional connectivity and patterns of within network temporal sequence separately, we conceived a method to use phase slope index properties of EEG signals as a probe for the measurement of connectivity between fMRI brain voxels. We decided to consider four macro-regions of the brain (anterior, posterior, left and right) and to average electrodes signals in those regions. We calculated phase slope index between signals coming from every pair of regions of electrodes as a function of the repetition time used in the fMRI sequence. In such a way, we created a set of regressors for a general linear model fitting of fMRI signals. This kind of approach will allow us to avoid any deconvolution between neuronal and hemodynamic contribution to the fMRI signal. This is because we preferred to not introduce any bias forcing signals to fit a specific model (the hemodynamic response function) preserving all the degrees of freedom typical of a data-driven approach. Results of this analysis are almost ready and their presentation as a manuscript is in progress.
Impacts
The project has made a significant contribution to our understanding of pharmacological sedation in the human brain. This provides a platform for future investigations to apply the methods developed to disorders of consciousness and cognition such as dementia. In particular the development of graph theory based methods offers an efficient way of evaluating key nodes within the brain and their altered function. While the project examines basic systems neuroscience in the human brain, it lays the groundwork for improvements in anaesthesia and offers markers of brain function that in the long term could be useful for drug development in neurological and psychiatric disorders.
The project has also had a significant impact on the development of Dr Gili's research career. In the last two years, Dr Gili realised the experimental project proposed, analysed data and disseminated results through conferences and publications, more of which will be disseminated in the coming months. His skills as neuroimaging researcher were significantly improved. He learnt to run autonomously EEG-fMRI experiments and in particular the optimisation of such techniques for the investigation of pharmacological effects in the human brain. He developed and implemented several advanced methods for data analysis, whose flexibility will allow their use for analyse data obtained in different fMRI experiments and last but not least he improved considerably his spoken and written English.
We used functional connectivity (FC) to investigate possible alterations induced by mild propofol sedation on resting state fMRI data. We tested for changes in cortico-cortical and cortico-subcortical FC, using regions of interest that encompass major subdivisions of the cortex (frontal, temporal, parietal, occipital). In particular, prefrontally connected portions of the thalamus were shown to play an important role in maintaining high the synchronisation between cortical regions whose connectivity may be reduced by the sedation. We found that the thalamus may be part of a network (including the cingulate cortex and a portion of brainstem), which coordinates the different stages of the descent to an unconscious state. The method we used to investigate cortical functional connectivity suggests that the temporal cortex is more sensitive to local connectivity variations, than other regions reported in previous studies. These results were presented as an electronic poster at the 20th ISMRM Annual Meeting (5 - 11 May 2012, Melbourne, Australia).
The main limitation of standard functional connectivity analysis resides in the bias introduced every time a region of interest (ROI) is defined. The definition of the sise and the location of an ROI introduces a considering amount of variability across subjects and conditions. To overcome this issue we decided to use graph theory assisted functional connectivity analysis instead of simple cross correlation analysis. The use of graph theory has the great advantage to avoid any operator dependent choice of ROIs needed for cross correlation: each voxel of the brain is considered as an ROI. Obviously, removing the constraint of averaging timeseries in specific regions increases enormously the amount of data to handle (around 40 000 timeseries considered simultaneously!). Fortunately, the workstation bought using part of the budget provided by the fellowship, allowed an easy solution of this problem. Graph theory analysis returned interesting results. We considered brain voxels as nodes of a complex network and measured eigenvector centrality (EC) to characterise brain network properties. The EC mapping of fMRI data in healthy humans under propofol mild sedation, demonstrated a decrease of centrality of the thalamus versus an increase of centrality within the pons of the brainstem, highlighting the important role of these two structures in regulating consciousness. Specifically, the decrease of thalamus centrality results from its disconnection from a widespread set of cortical and subcortical regions, while the increase of brainstem centrality may be a consequence of its increased influence over a few highly central cortical regions key to the default mode network such as the posterior and anterior cingulate cortices. These results were shown as traditional poster at the OHBM 2012 Annual Meeting (10 - 14 June 2012, Beijing, China), and submitted as a paper to the Journal of Neuroscience. The paper has been accepted subject to revisions.
The last period of the fellowship was devoted to the analysis of resting state EEG-fMRI data that we collected during wakefulness and mild propofol sedation. Instead of investigating propofol induced alterations in EEG-fMRI functional connectivity and patterns of within network temporal sequence separately, we conceived a method to use phase slope index properties of EEG signals as a probe for the measurement of connectivity between fMRI brain voxels. We decided to consider four macro-regions of the brain (anterior, posterior, left and right) and to average electrodes signals in those regions. We calculated phase slope index between signals coming from every pair of regions of electrodes as a function of the repetition time used in the fMRI sequence. In such a way, we created a set of regressors for a general linear model fitting of fMRI signals. This kind of approach will allow us to avoid any deconvolution between neuronal and hemodynamic contribution to the fMRI signal. This is because we preferred to not introduce any bias forcing signals to fit a specific model (the hemodynamic response function) preserving all the degrees of freedom typical of a data-driven approach. Results of this analysis are almost ready and their presentation as a manuscript is in progress.
Impacts
The project has made a significant contribution to our understanding of pharmacological sedation in the human brain. This provides a platform for future investigations to apply the methods developed to disorders of consciousness and cognition such as dementia. In particular the development of graph theory based methods offers an efficient way of evaluating key nodes within the brain and their altered function. While the project examines basic systems neuroscience in the human brain, it lays the groundwork for improvements in anaesthesia and offers markers of brain function that in the long term could be useful for drug development in neurological and psychiatric disorders.
The project has also had a significant impact on the development of Dr Gili's research career. In the last two years, Dr Gili realised the experimental project proposed, analysed data and disseminated results through conferences and publications, more of which will be disseminated in the coming months. His skills as neuroimaging researcher were significantly improved. He learnt to run autonomously EEG-fMRI experiments and in particular the optimisation of such techniques for the investigation of pharmacological effects in the human brain. He developed and implemented several advanced methods for data analysis, whose flexibility will allow their use for analyse data obtained in different fMRI experiments and last but not least he improved considerably his spoken and written English.