Projects
Computational Modeling
The goal of these projects is to understand the mechanisms that cause epileptic phenomena in brain tissue, and identify ways to identify and characterize these abnormal signals.
Method 1: Mathematical models of seizure dynamics
In a collaboration with Viktor Jirsa and Christophe Bernard (Marseille, France), we developed a dynamical model of seizure activity, based upon the bifurcations and state variables necessary to produce focal seizures. We found that the most common bifurcations were a saddle node at seizure onset, and a homoclinic bifurcation at offset. These bifurcations led to several predictions about canonical seizure dynamics, several of which we were able to validate in multiple species such as rats, mice, zebrafish, and humans. Using this framework, we are now categorizing human seizure dynamics on a larger scale, with the hopes to improve our understanding of the time course of seizures.
Method 2: NEURON modeling of High Frequency Oscillations
Oscillations are very common in brain tissue. One of the most intriguing phenomena are High Frequency Oscillations (HFOs) that arise in both normal and epileptic tissue. Several theories have been proposed to explain them. Based upon experimental and modeling using NEURON software, we have developed a detailed computational model of hippocampal tissue that shows how HFOs can arise under many different conditions. We found that they can arise even in the absence of any coupling between cells. This unusual result suggests that HFOs are an an inherent property of active neurons, regardless of underlying connectivity.
Method 3: Constructed Local Field Potentials
It is still unclear what creates the local field potentials that are recorded by electrodes. Are they produced by postsynaptic potentials (PSP), action potentials (AP), or a mixture of both? This uncertainty has become very important in the evaluation of High Frequency Oscillations as a biomarker of epilepsy, because they represent very different underlying mechanisms. However, technology currently is unable to monitor with the necessary spatial and temporal resolution to answer this question experimentally. We invented a novel model of constructed LFPs, in order to demonstrate what the recorded potential would look like when comprised of just APs or PSPs. We found that APs are capable of producing a wide range of frequencies and are very robust to variability between cells. PSPs, on the other hand, are unlikely to produce any frequency over 200 Hz simply due to the signal characteristics, regardless of the network structure. This work also validated that the hypotheses of the NEURON model were robust and did not depend upon specific network structure