Epilepsy Prediction from EEG and ECOG Data

43
Epilepsy Prediction from EEG and ECOG Data Nathan Intrator Computer Science Tel-Aviv University cs.tau.ac.il/~nin Yaari & Beck, 2002; Lopes da Silva et al., 2003; Collaborators TAU Hospital: Talma Hendler, Itzhak Fried, Miri Noifeld, TAU: Eshel Ben Jacob, Ilana Podipsky, Andrey Zhdanov

description

Epilepsy Prediction from EEG and ECOG Data. Nathan Intrator Computer Science Tel-Aviv University cs.tau.ac.il/~nin. Collaborators TAU Hospital : Talma Hendler, Itzhak Fried, Miri Noifeld , TAU : Eshel Ben Jacob, Ilana Podipsky , Andrey Zhdanov. Outline. - PowerPoint PPT Presentation

Transcript of Epilepsy Prediction from EEG and ECOG Data

Page 1: Epilepsy Prediction from EEG and ECOG Data

Epilepsy Prediction fromEEG and ECOG Data

Nathan IntratorComputer ScienceTel-Aviv Universitycs.tau.ac.il/~nin

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

CollaboratorsTAU Hospital: Talma Hendler, Itzhak Fried, Miri Noifeld, TAU: Eshel Ben Jacob, Ilana Podipsky, Andrey Zhdanov

Page 2: Epilepsy Prediction from EEG and ECOG Data

Outline The Epilepsy Problem, Clinical Terms, and need

for prediction Sensing, eeg, ecog, depth electrodes Animal models Wavelets Eshel Vagus nerve Heart/EEG, HRV, HS Complex Network Theory bocaletti Da Silva / Cerotti, Correlation, My contribution – level sets

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

[email protected]

Page 3: Epilepsy Prediction from EEG and ECOG Data

Seizure prediction by non-linear time series analysis of brain electrical activity

Ilana Podlipsky

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 4: Epilepsy Prediction from EEG and ECOG Data

Epilepsy Epilepsy

› Synchronous firing of neurons which create high amplitude electrical discharges; this ‘storm’ inhibits other neural signals from getting through and disables function areas of the brain

Statistics› Everyone's brain has the ability to produce a seizure under the

right conditions› 1 in 20 will have an epileptic seizure at some time in their life

Treatment› Once diagnosed with epilepsy, people are generally given anti-

epileptic medication. With the appropriate treatment, up to 70% of people could be seizure free.

Characteristics / symptoms› Seizures (40 different types)› ‘Aura’, a sensory hallucination, often precludes a seizure

EEG› Recording of neural activity of targeted neurons / neural

regions in brain› Outputs brainwaves with associated rhythms and frequencies

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 5: Epilepsy Prediction from EEG and ECOG Data

Types of Epilepsy Partial Seizures (Most Common) Video

› Simple partialynchronous firing of neurons which create high amplitude electrical discharges; this ‘storm’ inhibits other neural Complex partial

› Statistics› Everyone's brain has the ability to produce a seizure under the

right conditions› 1 in 20 will have an epileptic seizure at some time in their life

Absence› Once diagnosed with epilepsy, people are generally given anti-

epileptic medication. With the appropriate treatment, up to 70% of people could be seizure free.

Characteristics / symptoms› Seizures (40 different types)› ‘Aura’, a sensory hallucination, often precludes a seizure

EEG› Recording of neural activity of targeted neurons / neural

regions in brain› Outputs brainwaves with associated rhythms and frequencies

Epilepsy.com

Page 6: Epilepsy Prediction from EEG and ECOG Data

Epilepsy Epilepsy

› Synchronous firing of neurons which create high amplitude electrical discharges; this ‘storm’ inhibits other neural signals from getting through and disables function areas of the brain

Statistics› Everyone's brain has the ability to produce a seizure under the

right conditions› 1 in 20 will have an epileptic seizure at some time in their life

Treatment› Once diagnosed with epilepsy, people are generally given anti-

epileptic medication. With the appropriate treatment, up to 70% of people could be seizure free.

Characteristics / symptoms› Seizures (40 different types)› ‘Aura’, a sensory hallucination, often precludes a seizure

EEG› Recording of neural activity of targeted neurons / neural

regions in brain› Outputs brainwaves with associated rhythms and frequencies

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 7: Epilepsy Prediction from EEG and ECOG Data

Examples of Seizure Morphologies

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 8: Epilepsy Prediction from EEG and ECOG Data

Complex Network Theory

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 9: Epilepsy Prediction from EEG and ECOG Data

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 10: Epilepsy Prediction from EEG and ECOG Data

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Prof Paul Gompers

Page 11: Epilepsy Prediction from EEG and ECOG Data

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Vagus nerve stimulation (VNS) A lead is attached to the left vagus nerve in the lower part of the neck.

• It delivers mild electrical stimulations on demand

Deep brain stimulation targets the thalamus (which relays pain, temperature, and touch sensations to the brain).

Page 12: Epilepsy Prediction from EEG and ECOG Data

Resu

lts o

f HR

V

Pre

dic

tion

Hum

ans

Rats

Successful forecasting Tachycardia period

success rate 86%

|∆RRI| Vs. RRI forecasting times

1.5-11 min.

Successful forecasting Bradycardia period

success rate 82%

|∆RRI| Vs. RRI forecasting times

2.5-9 min.

Page 13: Epilepsy Prediction from EEG and ECOG Data

Fyodor Dostoyevsky(1821-1881)

Most known epileptic novelist

Gave vivid accounts of apparent temporal lobe seizures in his novel “The Idiot”

Describes an aura he used to get before the onset of a seizure

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 14: Epilepsy Prediction from EEG and ECOG Data

Vagus Nerve Stimulation

Longest nerve in the body;

sweat, blood pressure, and heart activity (heart rate)

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 15: Epilepsy Prediction from EEG and ECOG Data

The Vagus Nerve

Longest nerve in the body; Originates in the Brainstem Goes all the way to the stomach,

passing through essential organs (Vocal cords, heart, lungs, intestines)

Also controls sweat, blood pressure, and heart activity (e.g., heart rate)

Yaari & Beck, 202; Lopes da Silva et al., 2003;

Page 16: Epilepsy Prediction from EEG and ECOG Data

The Vagus Nerve (cont)

Modulates the SYMPATHETIC and PARASYMPATHETIC system

Goes all the way to the stomach, passing through essential organs (Vocal cords, heart, lungs, intestines)

Also controls sweat, blood pressure, and heart activity (heart rate)

Yaari & Beck, 202; Lopes da Silva et al., 2003;

Page 17: Epilepsy Prediction from EEG and ECOG Data

The problem

~30% of epileptics left untreated and victim ofviolent seizures

Injuries resulting from epilepsy is most oftencaused by convulsive seizures

If a ‘lead-time’ could be provided by a seizuredetection system, physical injury would be greatly

reduced and quality of life increased

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 18: Epilepsy Prediction from EEG and ECOG Data

Distinctive Features of Epilepsy

The epileptogenic process is characterized by abnormal synchronous burst discharges in neuronal cell assemblies recordable during and in between seizures (Matsumoto & Ajmone‐Marsan 1964a, Matsumoto & Ajmone Marsan 1964b; Babb et al. 1987).

The transition to a seizure is caused by an increasing spatial and temporal non-linear summation of the activity of discharging neurons (Calvin 1971; Calvin et al. 1973).

Due to the typically unpredictable occurrence of seizures it remains difficult to investigate the rules governing the initiation of seizure activity in humans.

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 19: Epilepsy Prediction from EEG and ECOG Data

Brain as a Dynamic System

A dynamical system consists of › State› Dynamics

State – the information necessary at any time instant to describe the future evolution of a system

Dynamics – defines how the state evolves over time

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 20: Epilepsy Prediction from EEG and ECOG Data

Attractors and Dimensions

Attractor› Set of states towards which the system evolves –

Characterizes the long term behavior of the system

Dimension of a system› Describes the amount of information required to

specify a point on the attractor - the long term behavior of a system

More complex behavior – more information is required to describe this behavior – higher dimension of the system

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 21: Epilepsy Prediction from EEG and ECOG Data

Brain as a Dynamic System

The application of the theory of non-linear dynamics offers information about the dynamics of the neuronal networks.

Several authors have shown that EEG/ECoG signals exhibit chaotic behavior (Basar,1990; Frank et al,1990; Pijn et al,1991).

The correlation dimension D2 (Grassberger and

Procaccia1983), provides good information about EEG complexity and chaotic behavior. (Mayer-Kress and Layne (1987) )

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 22: Epilepsy Prediction from EEG and ECOG Data

Dynamics of Epileptic EEG The spatio-temporal dynamics of the

epileptogenic focus is characterized by temporary transitions from high-to low-dimensional system states (dimension reductions) (Lehnertz & Elger 1995,1997).

These dimension reductions allow the lateralization and possibly localization of the epileptogenic focus (Lehnertz & Elger 1995,1997).

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 23: Epilepsy Prediction from EEG and ECOG Data

Seizure prediction by non-linear time series analysis of brain electrical activityChristian E. Elger, Klaus Lehnertz (1998)

Do prolonged and pronounced transitions from high - to low - dimensional system states characterize a pre-seizure phase?

The identification of this phase would enable new diagnostic and therapeutic possibilities in the field of epileptology.

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 24: Epilepsy Prediction from EEG and ECOG Data

Methods Electrocorticograms (ECoG) and stereoelectroencephalograms

(SEEG) of 16 patients 68 EEG epochs were analyzed.

› Fifty‐two data sets of state 1; mean duration: 19.5 ± 6.9 min; range: 6–40 min; minimum distance to any seizure: 24 h.

› 16 data sets of state 2; mean duration before the electrographic seizure onset: 15.1 ± 5.8 min; range: 10–30 min; seizure onset was defined as earliest signs of ictal ECoG/SEEG patterns).

Seizure prediction by non‐linear time series analysis of brain electrical activity Christian E. Elger, Klaus Lehnertz (1998)

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 25: Epilepsy Prediction from EEG and ECOG Data

Methods

A moving window dimension analysis was applied:

1. Data sets were segmented into half-overlapping digitally low-pass filtered consecutive epochs of 30 s duration.

2. Calculation of the modified correlation integral - the mean probability that the states at two different times are close.

3. Estimate of the correlation dimension D2 for each epoch.

Seizure prediction by non‐linear time series analysis of brain electrical activity Christian E. Elger, Klaus Lehnertz (1998)

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 26: Epilepsy Prediction from EEG and ECOG Data

Calculation of correlation dimension

Digital low-pass filtering (cut-off frequency 40 Hz)

Construction of m-dimensional vectors Xm(i) (i = 1, N; m = 1,. . . , 30) from the initial ECoG samples v(i) (i = 1, N) of a given electrode using the method of delays (Takens, 1981):

Seizure prediction by non‐linear time series analysis of brain electrical activity Christian E. Elger, Klaus Lehnertz (1998)

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 27: Epilepsy Prediction from EEG and ECOG Data

Correlation Integral For a stepwise decreasing radius r of a hypersphere

centered at each vector Xm(i) for increasing m the correlation integral Cm(r) was calculated as (Grassberger and Procaccia, 1983):

Counts the number of pairs of points with distance less then r.

For small r: Cm(r) ≈ rD2

D2 = slope of (in a linear region)

r

rCmlog

)(log

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 28: Epilepsy Prediction from EEG and ECOG Data

Calculation of Correlation Dimension

The correlation dimension D2 is obtained by:

D2=slope of

for decreasing r in a linear region

Alternatively:

If no linear region is found D2 = 10

r

rCmlog

)(log

rd

rCdD m

r log

)(loglim

02

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 29: Epilepsy Prediction from EEG and ECOG Data

Results For each selected electrode of the ECoG sets, a time profile of the

estimated D2, values was constructed. The seizure (S) exhibits lowest dimension values.

Seizure prediction by non‐linear time series analysis of brain electrical activity Christian E. Elger, Klaus Lehnertz (1998)

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 30: Epilepsy Prediction from EEG and ECOG Data

Results For both states maximum dimension reductions were always

found inside the epileptogenic focus regardless of spike activity.

During state 2, maximum dimension reductions were always observed in time windows immediately preceding seizures.

In state 1:› Dimension reductions with a mean of 1.0; range 0.5-2.5.› Mean duration of 5.25min; range 1.00–10.75 min.

In state 2:› Dimension reduction mean 2.0; range: 1.0–3.5.› Mean duration 11.50 min; range: 4.25–25.00 min.

Highly significant differences between maximum state 1 and pre-seizure state dimension reductions (Dr: Z = – 3.41, P = 0.0006;Tr: Z = – 3.52, P = 0.0004).

Seizure prediction by non‐linear time series analysis of brain electrical activity Christian E. Elger, Klaus Lehnertz (1998)

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 31: Epilepsy Prediction from EEG and ECOG Data

Discussion A reduced dimensionality of brain activity, as soon as it is of

sufficient size and duration, precisely defines states which proceed to a seizure.

I was demonstrated that the features of the pre-seizure state differ clearly from the one found during seizure.

Pronounced dimension reductions of pre-seizure electrical brain activity are restricted to the area of the epileptogenic focus, they can reflect increasing degree of synchronicity of pathologically discharging neurons.

Seizure prediction by non‐linear time series analysis of brain electrical activity Christian E. Elger, Klaus Lehnertz (1998)

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 32: Epilepsy Prediction from EEG and ECOG Data

Discussion

Correlation Dimension measure as presented here is subjective.

Highly sensitive to noise.

Subject specific.

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 33: Epilepsy Prediction from EEG and ECOG Data

Introduction cont.

the brain-heart axisVagus Nerve

The existence of pre- ictal phase

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 34: Epilepsy Prediction from EEG and ECOG Data

Introduction cont.

This study Forecasting seizures

Partial complex – humans Generalized - rats

Novel method for HRV analysis Ph.D. D.H.Kerem Ph.D. A.B.Geva

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 35: Epilepsy Prediction from EEG and ECOG Data

Known Methods

Spectral analysis of the time series of R-R intervals non-linear dynamics shortcoming - inability to account for non-stationary states

and transients

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 36: Epilepsy Prediction from EEG and ECOG Data

Known Methods cont.

Time-varying power spectral density estimation

Attractors and correlation dimensions

Karhunen-Love transform-based signal analysis method

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 37: Epilepsy Prediction from EEG and ECOG Data

Fuzzy clustering approach

comet or torpedo-shaped

unsupervised method advantage

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 38: Epilepsy Prediction from EEG and ECOG Data

Chosen method

EEG-contained information of HRV. (GEVA and KEREM, 1998)

an unsupervised method designed to deal with merging and overlapping states

ability to spot and classify

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 39: Epilepsy Prediction from EEG and ECOG Data

Data

resou

rces

Hum

ans

Rats

Humans 21 patients records, archived records

The recording machinery simultaneous EEG and video recording ECG channel visual inspection by an EEG expert The actual database

Rats Hyperbaric-oxygen ECG and EEG filtering and recording

Rats effects Time period analyzing Control rats Vs. research rats

OUTPUT

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 40: Epilepsy Prediction from EEG and ECOG Data

Method cont.

Choice of analysis parameters|∆RRI| Vs. RRIembedding dimension NFor this experiment –

Both featuresN = 3

number of clusters

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 41: Epilepsy Prediction from EEG and ECOG Data

Method cont.

Forecasting criteria Appearance Disappearance Dominant

False negative - False positive

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 42: Epilepsy Prediction from EEG and ECOG Data

Resu

lts

Hum

ans

Rats

Successful forecasting Tachycardia period

success rate 86%

|∆RRI| Vs. RRI forecasting times

1.5-11 min.

Successful forecasting Bradycardia period

success rate 82%

|∆RRI| Vs. RRI forecasting times

2.5-9 min.

Yaari & Beck, 2002; Lopes da Silva et al., 2003;

Page 43: Epilepsy Prediction from EEG and ECOG Data

Resu

lts c

on

t.

Hum

ans

Rats

prediction failuresfalse negative

One casefalse positive

Two casesLonger records

prediction failuresfalse negative

nonefalse positive

Two cases Ignoring changes shown in control rats

Yaari & Beck, 2002; Lopes da Silva et al., 2003;