An EEG based Neural Mass Model of Traumatic Brain Injury and Recovery

5 Analysis of brain activity reveals the presence of synchronous oscillations 6 over a range of frequencies. These oscillations can be observed using 7 electro-neurological measurements such as electroencephalogram (EEG), 8 magnetoencephalogram (MEG) or electrocorticogram (ECoG) . Further, 9 these rhythms can traverse different connected parts of the brain forming a 10 “system of rhythms”. These systems are analyzed in this paper using a 11 lumped-parameter, interconnected, neural mass models. This model allows 12 the analysis of the dynamics of the neural population in the frontal cortex 13 and their synapses using a few state variables. It is assumed here that the 14 neurons share the inputs and synchronizes their activity. The present work 15 is motivated by a recent paper by Bhattacharya et al who have proposed an 16 adaptation of Ursino’s neural mass model for the study of the changes in 17 alpha rhythms during the course of Alzheimer's disease. In that work, the 18 synaptic organization and connectivity in the lumped thalmo-cortico19 thalmic model was modified using experimental data. The authors were able 20 to reproduce the slowing of alpha rhythms (8-12 Hz) and decrease in power 21 of these rhythms associated with the Alzheimer's disease. Using this 22 research as the basis, the present work employs a pathophysiologic 23 understanding of traumatic brain injury to create a computational model of 24 traumatic brain injury that recreates the multimodal 25 electroencephalographic changes observed to occur with mild, moderate, 26 and severe traumatic brain injury. The focus is on recreating the observed 27 changes in the alpha and gamma rhythms (30-100Hz) due to traumatic brain 28 injury. Eight coupled neural mass models are used to represent the frontal 29 cortex. Numerical simulations are conducted using a well-known software 30 package. It is shown that the present model accurately reproduces the power 31 spectral density of the normal frontal cortex under white-noise excitation 32 conditions. Three degrees of traumatic brain injuries are then modeled by 33 decreasing the connection strengths in the neural mass model. A comparison 34 of the power spectral densities of the outputs of the normal and injured 35 neural mass models indicates that the present model is capable to 36 reproducing clinically-observed changes due to traumatic brain injuries. 37