P323 Modularity and inhibition: the transition from burst-suppression to healthy EEG signals in a microscale model
Guido Wiersma1,*, Michel van Putten1,2, Nina Doorn1 ● Department of Clinical Neurophysiology, University of Twente, 7500 AE Enschede, The Netherlands ● Department of Neurology and Clinical Neurophysiology, Medisch Spectrum Twente, 7500 KA Enschede, The Netherlands
* email: wiersmaguido@gmail.com Introduction
Burst-suppression (BS) is an electroencephalogram (EEG) pattern consisting of high voltage patterns (bursts, >20 µV) alternated with low voltage or even isoelectric periods (suppression)[1]. It can be categorized into BS with identical bursts, observed in comatose patients after brain ischemia indicating poor prognosis, and BS with heterogeneous bursts[2]. Where past research did not identify the neural origin of BS, recent work showed that the shift of heterogeneous to identical BS is caused by the loss of either inhibition, or modularity in the connectivity between neurons[3]. Here, we hypothesize that when both inhibition and modularity are included in a network, the transition of BS to a healthy network state can be modelled.
Methods To simulate the pathological and healthy states, a network of 2000 adaptive integrate and fire (IF) neurons is constructed. Such networks are known to generate both BS and a wide variety of healthy characteristics as observed in EEG (e.g. alpha or gamma activity)[4]. The adaptation mechanism of the IF neurons is conductance-based, preventing unrealistically negative membrane voltages during suppression periods as described in e.g.[5]. Inspired by Gao et al., simulation based inference is used to explore a wide variety of dynamics resulting from a broad range of free parameters[4].
Results The results show the influence of inhibition and modularity on the simulation of BS and healthy network states. Furthermore, by using one channel EEG data as target observations for the parameter inference, combined with the broad parameter range, we show to what extent the proposed microscale model can simulate these target EEG signals.
Discussion The roles of inhibition and modularity provide new insights into the mechanisms behind the transition of healthy brain states to BS. This opens potential pathways for treatments in comatose patients after ischemia. Although the model consists of only 2000 neurons, the striking similarity between BS patterns generated in-vitro and those observed in EEG recordings highlights the potential of microscopic models to capture features of large-scale brain activity[3,6,7]. This study demonstrates the potential of these biophysically detailed models to uncover cellular level insights from EEG signals.
Acknowledgements We thank Maurice van Putten, PhD, for his invaluable support, expertise, and generous provision of the code to implement synaptic parallel computing for dynamic load balancing. References [1] 1. https://doi.org/10.1097/01.nrl.0000178756.44055.f6 [2] 2. https://doi.org/10.1016/j.clinph.2013.10.017 [3] 3.https://doi.org/10.12751/nncn.bc2024.146 [4] 4. https://doi.org/10.1101/2024.08.21.608969 [5] 5. https://doi.org/10.1162/neco_a_01342. [6] 6. https://doi.org/10.1152/jn.00316.2002.
