PhD Scientific Days 2017

Poster Presentation: Neurosciences

P51: Analysis of EEG data of epileptic patients based on the estimation of Hawkes’ self-exciting point processes

Előadó neve

Perczel, György

Előadó munkahelye

National Institute of Clinical Neurosciences, Department of Functional Neurosurgery and Center of Neuromodulation

Előadó telefonszáma

+36207700787

Előadó e-mail címe

perczel.gyorgy.miklos@itk.ppke.hu

Az előadás címe

Analysis of EEG data of epileptic patients based on the estimation of Hawkes’ self-exciting point processes

Szerző(k) neve és munkahelye

György Perczel 1 2, Loránd Erőss 1 2, Dániel Fabó 3, László Gerencsér 4, Zsuzsanna Vágó 1
1 Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, Budapest
2 Department of Functional Neurosurgery and Center of Neuromodulation, National Institute of Clinical Neurosciences, Budapest
3 Department of Epileptology, National Institute of Clinical Neurosciences, Budapest
4 Institute for Computer Science and Control, Hungarian Academy of Sciences, Budapest

Szekció

Poster Presentation: Neurosciences

Data of the presenter

Doctoral School: Roska Tamás Doctoral School of Sciences and Technology, Pázmány Péter Catholic University
Program: Program 1: Bionics, Bio-inspired Wave Computers, Neuromorphic Models.
Supervisor: Erőss Loránd, Gerencsér László, Vágó Zsuzsanna
E-mail address: perczel.gyorgy.miklos@itk.ppke.hu

Text of the abstract

Introduction
A widely accepted hypothesis is that seemingly insignificant, but interacting events on a microscale level that may lead to the sudden appearance of macroscale phenomena are common in the dynamics of epileptic seizures, earthquakes and financial crises. Based on this assumption, the dynamics of EEG signals of epileptic patients is modeled using random point processes with feedback effect (Hawkes’ processes).
Aims
The objective of our study is to create an efficient framework to fit a Hawkes’ process to electroencephalographic (EEG) data and to develop and test new methods for detecting changes in the underlying dynamics of the EEG in the periictal state.
Method
First of all, we developed a method for simulating a Hawkes’ process via its dynamical representation. Then, the correctness and viability of the maximum likelihood estimation (MLE) method, following Ozaki et al., was tested. The potential of our MLE algorithm was explored by graphical representations of various cross-sections of the cost function. We fitted our model to the EEG by deriving point processes defined as level crossings of the signal.
Preliminary results
Simulations of Hawkes processes were successfully done. The characterization of the MLE shows that a high fidelity parameter estimation can be obtained in a wide region of the parameter space.
Conclusion
After successful simulations of Hawkes processes in real-time for a wide class of response functions, we present our experimental results for fitting a Hawkes process, using a maximum likelihood method, to simulated and real data, both prior to and during seizure.

Azonosító

P51

Kind

Szabad

Előadás fájl jóváhagyás

nem rendelkezett róla

Előadó

1217

Authors (legacy)

György Perczel 1 2, Loránd Erőss 1 2, Dániel Fabó 3, László Gerencsér 4, Zsuzsanna Vágó 1
1 Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, Budapest
2 Department of Functional Neurosurgery and Center of Neuromodulation, National Institute of Clinical Neurosciences, Budapest
3 Department of Epileptology, National Institute of Clinical Neurosciences, Budapest
4 Institute for Computer Science and Control, Hungarian Academy of Sciences, Budapest