Poster No:
1659
Submission Type:
Abstract Submission
Authors:
Dominic Dunstan1, Mark Richardson2, Eugenio Abela2, Samantha Chan3, Alexander Shaw1, Alessia Caccamo1, Marc Goodfellow1
Institutions:
1University of Exeter, Exeter, United Kingdom, 2King's College London, London, United Kingdom, 3University College London, London, United Kingdom
First Author:
Co-Author(s):
Introduction:
Neural mass models (NMMs) are a useful tool for interpreting observations of brain dynamics. By incorporating assumptions about how neuronal populations (excitatory and inhibitory) interact, they provide a means to understand data by examining model parameters as a proxy for underlying mechanisms. To interpret data using NMMs we need to quantitatively compare the output of NMMs with data, and thereby find parameter values for which the model can produce the observed dynamics. Though abstract, NMMs still comprise of many parameters that are difficult to constrain a priori.
Existing approaches to model fitting make assumptions to simplify a particular optimisation scheme. These include exploring only a subset of the parameter space by fixing the values of many parameters a priori (Wendling 2005; Nevado-Holgado 2012), and, in the case of dynamical causal modelling (Friston 2012), linearising models and assuming Gaussian distributions. Therefore, we still have little knowledge of how different regions of parameter space of NMMs can yield dynamics that approximate data, how nonlinearities in models can affect parameter mapping or how best to quantify similarities between model output and data.
Here, we take a global nonlinear approach for mapping parameters of NMMs. We use evolutionary algorithms to explore large regions of parameter space and to demonstrate how using different objective functions can affect the inferences made. We demonstrate some advantages of using multiobjective evolutionary algorithms (MOEA), combining the weighted horizontal visibility graph with spectral properties to quantify similarities between model and data. We illustrate the usefulness of this approach in two applications: (i) understanding the slowing of the alpha rhythm in epilepsy (Abela 2019); and (ii) understanding why children with epilepsy have altered EEG during slow wave sleep (Eriksson 2022).
Methods:
To explore the former application, we recorded EEG alpha activity during the eyes closed resting state from 20 healthy individuals and 20 people with focal epilepsy. For the latter application, EEG polysomnography was acquired from 15 children with drug naïve epilepsy. We then compared the slow wave sleep obtained from the frontal electrodes to 16 healthy age-matched controls. In both cases, we used the MOEA to optimise the dynamics of an NMM to the data and were able to accurately recreate properties of the data in the time and spectral domain.
Results:
We found that the mean excitatory gain parameter had the largest effect in explaining the shift in alpha power observed (Dunstan 2023). Counterintuitively, this parameter was reduced in people with epilepsy compared to control. By mapping the differences in slow-wave sleep seen in children with epilepsy to an NMM, we found enhanced neuronal firing rate (excitability) in model excitatory and inhibitory populations, which comes predominantly from enhanced excitatory synaptic currents. Furthermore, using the model to infer mechanisms underpinning a classical epileptiform spike-wave rhythm reveals that these differences in currents place patients closer to seizure dynamics than controls.
Conclusions:
These results demonstrate that the MOEA framework proposed is a potentially useful tool to map differences seen in EEG recordings to underlying mechanisms. In particular, we found that mechanisms of epilepsy are manifested in non-seizure states and that patient EEG can be mapped to specific synaptic deficits which are causative for epilepsy.
Modeling and Analysis Methods:
EEG/MEG Modeling and Analysis 1
Methods Development 2
Keywords:
Computational Neuroscience
Electroencephaolography (EEG)
Epilepsy
Modeling
PEDIATRIC
Sleep
1|2Indicates the priority used for review
Provide references using author date format
Abela, E. (2018), ‘Slower alpha rhythm associates with poorer seizure control in epilepsy’, Annals of Clinical and Translational Neurology, vol. 6, no. 2, pp. 333-343.
Dunstan, D.M. (2023), ‘Global nonlinear approach for mapping parameters of neural mass models’, PLOS Computational Biology, vol. 19, no. 3, e1010985.
Eriksson, M.H. (2023), ‘Sleep homeostasis, seizures, and cognition in children with focal epilepsy’, Developmental Medicine & Child Neurology, vol. 65, no. 5, pp. 701-711.
Friston, K.J. (2012), ‘DCM for complex-valued data: cross-spectra, coherence and phase-delays’, Neuroimage, vol. 59, no. 1, pp. 439-455.
Nevado-Holgado, A.J. (2012), ‘Characterising the dynamics of EEG waveforms as the path through parameter space of a neural mass model: Application to epilepsy seizure evolution’, NeuroImage, vol. 59, no. 3, pp. 2374-2392.
Wendling, F. (2005), ‘Interictal to ictal transition in human temporal lobe epilepsy: insights from a computational model of intracerebral EEG’, Journal of Clinical Neurophysiology, vol. 22, no. 5, pp. 343-356.