Dynamics of largDynamics of largy gbrain ne

Klaus L

y gbrain ne

Klaus LKlaus LMarie-There

Stephan

Klaus LMarie-There

Stephan

DepNeu

Interdisciplinary Center University of

p yfor Complex Systems

Supported by the Deutsche Forschungsgemeinschaft

ge-scale epileptic ge-scale epileptic g p petworks

ehnertz

g p petworks

ehnertzehnertz ese Kuhnert Bialonski

ehnertz ese Kuhnert Bialonski

t. of Epileptologyrophysics Group

Helmholtz-Institutefor Radiation- and

Bonn, Germany Nuclear Physics

Epile

Greek term for seizure; disease

~ 1 % of world population suffe

famous people suffering from eSokrates, Alexander the GreatFlaubert, Dostojevski, Carroll, Händel, van Gogh, Newton, P

epsy

e first mentioned ~ 1750 BC

ers from epilepsy

epilepsy: t, Julius Caesar, Lenin,Poe, Berlioz, Paganini, ascal, Helmholtz, Nobel

Treatment

antiepileptic drugs; primar

epilepsy surgery; option fop p y g y; prequirement: localize and dfrom functionally relevant brsuccess: ~ 60 % (15 % – 8success: ~ 60 % (15 % – 8

alternative therapies; for ~seizure prediction, seizure csuccess: ?success: ?

of Epilepsy

ry therapy; success: ~ 70 %

or ~ 5 – 10 % of patientspdelineate epileptic focus rain areas85 %)85 %)

~ 22 % of patientscontrol

Prob

where and when and why do seizu- where and when and why do seizu

- where and why do seizures spread

h d h d i d?- when and why do seizure end? co

lems

ures start? seizure precursors?ures start? seizure precursors?

d? consistency?

i t ?onsistency?

Epileptic Focus vs. Eptraditional concept: epileptic fo- circumscribed area of the brain - critical amount of neurons → ep

recent evidence: epileptic brainp p- functionally and anatomically co- activity in any one part affects a

vulnerability to seizures in any o- vulnerability to seizures in any oby activity everywhere else in th

- seizures may entrain large neur- growing evidence from imaging,and modeling studies

pileptic Brain Networkocus

pileptic seizures

n networkonnected brain structures activity in all the othersone part of the network influencedone part of the network influencedhe networkral networks from any given part, electrophysiological,

Epileptic Brastructural brain networks nodes – neurons li klinks – synapses hard (impossible?) to access

functional brain networks (largenodes – sensors (dynamics of neli k i t ti ( i ht d dlinks – interactions (weighted and

constructing and characterizingg g- time series analysis (strength an- global/local network characterist(clustering coefficient average p(clustering coefficient, average p

ain Networks

e scale: EEG, MEG, fMRI)tworks of neuron networks)

d/ di t d)d/or directed)

g functional brain networksgnd direction of interactions)tics path length centrality etc )path length, centrality, etc.)

Constructing Functiomultichannel recordings of brain dy

interaction matrix I

nsor

sen

- thres- signi- …

sensore.g. E. Bullmore & O. Sporns, Nat. Rev. Neurosci. 10, 186, 2009

onal Brain Networksynamics

adjacency matrix A

nsor

A = f(I)

sen

sholdingficance testing

sensor

Functional Brain Networkepileptic brain networks are mo

- 21 patients, 23 controls- scalp EEG recordings (29 sites)p g ( )- eyes-open (15 min) - eyes-closed (15 min) - mean phase coherence

(frequency-adaptive; -selective)( q y p )- binary networks

(fixed mean degree, thresholding) - weighted networks

(different normalizations) ( )- clustering coefficient C- average shortest path length L

M.T. Horstmann et al., Clin. Neurophysiol.. 121, 172, 2010

ks: Epilepsy vs. Controls

weighted binaryre regular than healthy ones

* *

* *

* p < 0.050.5 – 5 Hz (δ-band)

Epileptic Brain Netwofunctional topology

from more randommore random

tomore regular

b k tback tomore random

- 60 patients, 100 seizures- intracranial EEG recordings

(53 ± 21 sites)( )- cross-correlation- thresholding (A fully connected)- clustering coefficient C- average shortest path length L

K. Schindler et al., Chaos 18, 033119, 2008

g p g- comparison with random networks (prescribed de

orks during Seizures

see also: Ponten et al., Clin. Neurophysiol. 118, 918, 2007Kramer et al., Epilepsy Res. 79, 173, 2008Kramer et al., J. Neurosci. 30, 1007, 2010

egree sequence)

Regular Functional Topnetwork sync: a mechanism for

- 60 patien- intracran

(53 ± 21 - cross-co

K. Schindler et al., Brain 130, 65, 2007

- spectrum

pology during Seizuresseizure termination?

nts, 100 seizuresnial EEG recordingssites)

orrelationm of eigenvalues

Long-Term Dynamics of mainly reflects daily rhythms, ep

seizures

t

- 13 patients, 75 seizures- intracranial EEG recordings (> 2100 h)

pat

intracranial EEG recordings ( 2100 h)(56 sites, range: 24-72)

- mean phase coherence (frequency-adaptive) - thresholding (fixed mean degree)- clustering coefficient C

M.T. Kuhnert et al., Chaos 20, 043126, 2010

clustering coefficient C- average shortest path length L

Epileptic Brain Networkspileptic process only marginally

exemplary frequency distributions

power spectral density estimatest 13

Cn Lndaily rhythms

power spectral density estimates(grand average)

t.: 13

precursor dynamics ?

Pr

- reduction of complexitymultivariate dynamics →multivariate dynamics →

t k h t i ti bi- network characteristics as biom

- improved characterization of epseizure precursor ?other network characterisother network characteris

w

os

single numbersingle number

k (?)marker (?)

pileptic process

sticsdegree centrality

Pat A Pat Bstics

work in progress …epileptogenic area

p g

Challenges and Fu

- node identification spatial sampling, transitivit

- link identificationlink identificationindirect vs. direct interactiotransitivity (common sourcereliabilit of estimators forreliability of estimators for

- interpretation of findingscomparison of empirical neappropriate null models(1) /appropriate null models /

uture Requirements

yMEG recording spatial model

ons, es) interactions

(S. Bialonski et al., Chaos 20, 013134, 2010)

interactions

etworks/ surrogate networks(2)/ surrogate networks

(1) S. Bialonski et al (submitted) (2) G. Ansmann & K.L. (submitted)