Â鶹ӰԺ

Decoding Epileptogenesis in a Reduced State Space

;Ìý

Date and time:Ìý

Thursday, October 8, 2015 - 2:00pm

³¢´Ç³¦²¹³Ù¾±´Ç²Ô:Ìý

ECCR 257

´¡²ú²õ³Ù°ù²¹³¦³Ù:Ìý

In many areas of science the only method to study a complex system entails making indirect time-resolved measurements of the state of the system. In the absence of a detailed mathematicalÌýmodel that can be used to explain the measurements, we have to resort to machine learning methods toÌýlearn the association between the state of the system and the measurements. An example of such aÌýproblem involves the definition of a biomarker to monitor epileptogenesis following a traumaticÌýbrain injury.

Ìý

In this talk I will describe the recent results of a multidisciplinary effort to actively and continuously decode the progressive changes in neural network organization leading to epilepsy.ÌýUsing an animal model of acquired epilepsy, we chronically recorded hippocampal auditory evokedÌýpotentials during epileptogenesis. ÌýOur approach combines in a unique manner applied harmonicÌýanalysis, spectral graph theory, and state-space models to infer the hidden distinct stages ofÌýepileptogenesis. Using our decoding algorithm, we were able to show that archetypal changes in theÌýwaveform morphology have universal predictive value for the development of spontaneous recurrentÌýseizures.

Ìý

This work was done in collaboration with Daniel Barth, Alexander M. Benison, ÌýZachariah Smith, and Lukas Ruediger Nels Goetz-Weiss