Abstracts

Rationale: Epilepsy is a network disease with distributed nodes over various brain regions with network dynamics that change over time.If multiple high density DBS leads are placed near regions of interest, then electrical source imaging (ESI) techniques using the recordings of such electrodes can locate nodes of epileptic activity.We hypothesize that using an array of intracranial electrodes such as high density DBS leads, we can obtain precise location of epileptic network nodes and consequently offer enhanced targeted stimulation at critical nodes in a subject and event specific manner.For epilepsy, this translates to detecting the initiation of epileptic events and localizing the functional nodes involved in the epileptic network and to determine the underlying dynamics using Granger causality analysis.Exploiting this data we can determine when to turn the stimulation on (employing the same DBS array), to target the functional nodes that are critical in the generation and progression of seizures, and tune stimulation parameters optimally.Methods: We tested this idea in a computer simulation study.DBS leads with 80 electrodes were placed bilaterally in the brain.3 nodes located at thalamus (Th), hippocampus (Hp) and the cortex (in temporal lobe) were selected as nodes within an epileptic network.A causal relation between thalamus and the rest of the nodes, and hippocampus and cortex was assigned.Applying ESI techniques and causal modeling, we identified the causal nodes (Th & Hp in this example) and the underlying network dynamics.Given the target area, we determined how to activate the DBS electrodes to optimally stimulate them.The Poisson equation was solved to determine the forward relation between the stimulation current distribution of DBS leads and the resulting current density distribution within the brain volume.Consequently, the desired current density (maximal at target nodes and minimal elsewhere) was achieved by solving the inverse problem and determining stimulation current patterns.Results: The location of the 3 nodes within the simulated epileptic network were estimated accurately when DBS recordings were used (2 mm error), whereas a 14 mm localization error was achieved when EEG was used.The activity of the underlying sources, i.e. estimated time series of each node, were input into the Granger causality analysis to find the central nodes of the epileptic network.Analysis of these time series when DBS recording was used, revealed the true underlying dynamics of the simulated network (the direction and relative strength of the links between different nodes) whereas EEG-only analysis did poorly and missed some of the causal links and their directions, hence establishing misleading causality.Conclusions: Our results suggest good specificity and resolution in sensing epileptic networks and targeting them using high density DBS leads.This provides a novel framework in studying epilepsy, in a high spatio-temporal resolution and also provides a means to intervene and modulate in a subject/seizure-specific manner.This work was supported in part by NIH RO1 EB006433 and NSF CBET-1450956