Project

Back to overview

A Bayesian Inference Approach to Intracranial EEG Seizure Dynamics

English title A Bayesian Inference Approach to Intracranial EEG Seizure Dynamics
Applicant Schindler Kaspar
Number 155950
Funding scheme Project funding (Div. I-III)
Research institution Universitätsklinik für Neurologie Inselspital Bern
Institution of higher education University of Berne - BE
Main discipline Neurology, Psychiatry
Start/End 01.11.2014 - 31.10.2017
Approved amount 334'500.00
Show all

All Disciplines (2)

Discipline
Neurology, Psychiatry
Other disciplines of Physics

Keywords (4)

Intracranial EEG; Pharmacoresistance; Bayesian Approach; Human Epilepsy

Lay Summary (German)

Lead
Um Patienten/innen mit Epilepsie wirkungsvoller und gezielter behandeln zu können, ist es notwendig, die Auswirkungen therapeutischer Interventionen auf Entstehung und Dynamik epileptischer Anfälle besser zu verstehen. Ein moderner Weg zu einem tieferen Verständnis dieser Auswirkungen ist die Computer-unterstützte Entwicklung von Modellen der elektrischen Aktivität von Teilen des menschlichen epileptischen Gehirns. Mit Hilfe dieser Modelle können dann die Effekte präzise umschriebener ("minimal invasiver") therapeutischer Interventionen simuliert werden. Ziel dieses Projektes ist die Entwicklung solcher Modelle, um einen Beitrag zur weiteren Verbesserung der Therapie von Epilepsiepatienten/innen zu leisten.
Lay summary

Inhalt und Ziel des Forschungsprojektes

Patienten/innen, die an epileptischen Anfällen leiden, welche mit Medikamenten nicht kontrolliert werden können, kann unter Umständen mit einem epilepsiechirurgischen Eingriff geholfen werden. Typische Voraussetzungen dafür sind, dass 1. alle Anfälle in einem einzigen Hirnareal entstehen, 2. dieses Hirnareal genau lokalisiert werden kann und 3. die chirurgische Entfernung dieses Hirnareales nicht zu für den Patienten/in unakzeptablen neurologischen Folgeschäden führt. Um das Hirnareal, in welchem die Anfälle entstehen, genau zu lokalisieren, ist es häufig notwendig, die elektrische Aktivität mit speziellen Sensoren (Elektroden) zu messen, die direkt in das Hirn eingelegt werden. Diese Signale werden als "intrakranielles Elektroencephalogramm (iEEG)" bezeichnet. Das Ziel des vorliegenden Projektes ist es, aus diesen iEEG Signalen mit Hilfe leistungsfähiger mathematischer Methoden ("Bayessche Statistik") Modelle zu entwickeln - sogenannte "Probabilistic Graphical Models" (PGMs) - die dann die Simulation gezielter therapeutischer Interventionen, wie zum Beispiel die chirurgische Entfernung oder elektrische Modulation bestimmter Hirnareale, erlauben. 

Wissenschaftlicher und gesellschaftlicher Kontext des Forschungsprojektes

Dieses Projekt beinhaltet die mathematische Analyse und Modellierung der elektrischen Aktivität von Teilen des menschlichen epileptischen Gehirns und soll zur Verbesserung der therapeutischen Möglichkeiten bei  einer der häufigsten - und deshalb nicht nur für den individuellen Patienten/In, sondern auch für die Gesellschaft als Ganzes relevanten - neurologischen Krankheit beitragen. Die dabei zu entwickelnden und angewandten Analysemethoden sind zudem nicht nur für die Analyse hirnelektrischer Signale hoch relevant, sondern können auf zahlreiche weitere komplexe Systeme angewendet werden, wie sie in unserer heutigen zunehmend vernetzten Welt allgegenwärtig geworden sind.

Direct link to Lay Summary Last update: 25.09.2014

Responsible applicant and co-applicants

Employees

Publications

Publication
Personalized structural image analysis in patients with temporal lobe epilepsy.
Rummel Christian, Slavova Nedelina, Seiler Andrea, Abela Eugenio, Hauf Martinus, Burren Yuliya, Weisstanner Christian, Vulliemoz Serge, Seeck Margitta, Schindler Kaspar, Wiest Roland (2017), Personalized structural image analysis in patients with temporal lobe epilepsy., in Scientific reports, 7(1), 10883-10883.
An optimal strategy for epilepsy surgery: Disruption of the rich-club?
Lopes Marinho A, Richardson Mark P, Abela Eugenio, Rummel Christian, Schindler Kaspar, Goodfellow Marc, Terry John R (2017), An optimal strategy for epilepsy surgery: Disruption of the rich-club?, in PLoS computational biology, 13(8), 1005637-1005637.
Predictive modeling of EEG time series for evaluating surgery targets in epilepsy patients.
Steimer Andreas, Müller Michael, Schindler Kaspar (2017), Predictive modeling of EEG time series for evaluating surgery targets in epilepsy patients., in Human brain mapping, 38(5), 2509-2531.
All together now: Analogies between chimera state collapses and epileptic seizures.
Andrzejak Ralph G, Rummel Christian, Mormann Florian, Schindler Kaspar (2016), All together now: Analogies between chimera state collapses and epileptic seizures., in Scientific reports, 6, 23000-23000.
Estimation of brain network ictogenicity predicts outcome from epilepsy surgery
Goodfellow M., Rummel C., Abela E., Richardson M. P., Schindler K., Terry J. R. (2016), Estimation of brain network ictogenicity predicts outcome from epilepsy surgery, in Scientific Reports, 6, 29215-29215.
Ictal time-irreversible intracranial EEG signals as markers of the epileptogenic zone
Schindler Kaspar, Rummel Christian, Andrzejak Ralph G., Goodfellow Marc, Zubler Frédéric, Abela Eugenio, Wiest Roland, Pollo Claudio, Steimer Andreas, Gast Heidemarie (2016), Ictal time-irreversible intracranial EEG signals as markers of the epileptogenic zone, in Clinical Neurophysiology, 127(9), 3051-3058.
Chow-Liu trees are sufficient predictive models for reproducing key features of functional networks of periictal EEG time-series
Steimer Andreas, Zubler Frédéric, Schindler Kaspar (2015), Chow-Liu trees are sufficient predictive models for reproducing key features of functional networks of periictal EEG time-series, in NeuroImage, 118, 520-537.
Random Sampling with Interspike-Intervals of the Exponential Integrate and Fire Neuron: A Computational Interpretation of UP-States.
Steimer Andreas, Schindler Kaspar (2015), Random Sampling with Interspike-Intervals of the Exponential Integrate and Fire Neuron: A Computational Interpretation of UP-States., in PloS one, 10(7), 0132906-0132906.
Resected Brain Tissue, Seizure Onset Zone and Quantitative EEG Measures: Towards Prediction of Post-Surgical Seizure Control.
Rummel Christian, Abela Eugenio, Andrzejak Ralph G, Hauf Martinus, Pollo Claudio, Müller Markus, Weisstanner Christian, Wiest Roland, Schindler Kaspar (2015), Resected Brain Tissue, Seizure Onset Zone and Quantitative EEG Measures: Towards Prediction of Post-Surgical Seizure Control., in PloS one, 10(10), 0141023-0141023.

Collaboration

Group / person Country
Types of collaboration
University Department for Neurosurgery, Inselspital, Bern Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
University Institute for Diagnostic and Interventional Neuroradiology, Inselspital, Bern Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Exchange of personnel
University of Exeter Great Britain and Northern Ireland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Universitat Pompeu Fabra, Department of Information and Communication Technologies, Barcelona Spain (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Fröhlich Lab, Department of Psychiatry, University of North Carolina at Chapel Hill United States of America (North America)
- in-depth/constructive exchanges on approaches, methods or results
Centre for Epilepsy Surgery, Niguarda Hospital, Milano Italy (Europe)
- in-depth/constructive exchanges on approaches, methods or results

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
First International Summer Institute on Network Physiology (ISINP) Poster Evaluating Resective Surgery Targets in Epilepsy Patients: A Comparison of Quantitative EEG Methods 24.07.2017 Como, Italy Müller Michael;
5th BENESCO Winter Research Meeting Talk given at a conference Predicting Critical Brain Targets For Resective Surgery in Pharmacoresistant Epilepsy Patients 09.03.2017 Wengen, Switzerland Steimer Andreas; Schindler Kaspar;
3rd SFCNS Congress Talk given at a conference Assessing epileptogenic brain regions/networks by intracranial EEG 05.11.2016 Basel, Switzerland Schindler Kaspar;
12th European Congress on Epileptology Talk given at a conference Synchronization in human focal seizures 11.09.2016 Prag, Czech Republic Schindler Kaspar;
Dynamics Days 2016 Talk given at a conference Distributional Clustering of Multivariate Time Series 06.06.2016 Korfu, Greece Steimer Andreas;
4th BENESCO Winter Research Meeting Poster Predictive modeling of post-surgical outcome in epilepsy treatment 10.03.2016 Wengen, Switzerland Steimer Andreas; Müller Michael;
3rd BENESCO Winter Research Meeting Talk given at a conference ’Predictive modeling of the spatio-temporal dynamics of periictal iEEG time-series 26.03.2015 Wengen, Switzerland Steimer Andreas;
10th Annual Meeting Clinical Neuroscience Bern Poster Chow-Liu Trees Are Sufficient Predictive Models For Reproducing Key Features of Functional Networks of Periictal EEG Time-Series 22.01.2015 Bern, Switzerland Steimer Andreas; Gast Heidemarie; Schindler Kaspar;


Associated projects

Number Title Start Funding scheme
122010 Assessing functional networks in human epileptic brains 01.01.2009 Project funding (Div. I-III)
140332 Imaging large scale neuronal networks in epilepsy 01.05.2012 SPUM

Abstract

Background: The primary treatment goal in patients suffering from epilepsy is complete seizure freedom. However, seizure freedom is still not achieved in 20-30% of the patients. It is therefore mandatory to find biomarkers, i.e. measurable characteristics of seizure generation that have the potential to improve current therapies and to guide the design of novel approaches like electric brain stimulation. Here we set out to invoke the powerful tools offered by Bayesian inference methods to assess seizure dynamics, i.e. the spatiotemporal characteristics of intracranial EEG (iEEG) that are critical for seizure onset, seizure propagation and seizure termination. A crucial aspect of our approach is the implementation of versatile, realistic models of multi-channel iEEG time series, which should allow to predict how specific iEEG signals, generated in different parts of the epileptic brain, influence overall iEEG seizure dynamics. We expect that the Bayesian inference approach will clearly surpass the limitations of state-of-the-art methods of iEEG analysis, such as functional networks, which are purely descriptive only and which do not allow to model the effect of localized changes onto the overall iEEG patterns. Working hypothesis:Using multi-channel iEEG time series, the Bayesian inference approach allows for realistic predictions regarding the probability of seizure occurrence. In particular, this approach allows to test, how specific iEEG signals in different parts of the epileptic brain influence overall iEEG seizure dynamics and seizure probability.Patients and methods:In patients with pharmaco-resistant epilepsies who undergo invasive monitoring, iEEG will be prospectively recorded with high spatio-temporal resolution using macroelectrodes and microwires. Bayesian inference -in particular Probabilistic Graphical Models (PGMs) that accurately represent seizure and non-seizure brain states- will be learned from iEEG signals. The PGMs then allow to extensively test, how changes of one or more iEEG signals influence seizure dynamics and thereby help to identify the brain regions and electric brain activities that are key for seizure control. To further validate the PGMs, we will retrospectively analyze an existing database consisting of >200 iEEG seizures from patients with known post-surgical outcomes. Expected value of the project:By invoking the powerful framework of Bayesian inference applied to iEEG signals this project will deepen the understanding of how localized changes of electric activity influence the overall dynamics of human epileptic brains during seizures. We expect the results of this study to potentially be key to better predict the outcome of current resective epilepsy surgeries. In addition, they may also provide guidance for future neuromodulatory therapies, by helping to improve target selection and identifying those iEEG signals that should be induced by therapeutic interventions.
-