Epilepsy supervision based on brain and heart activity
analysis
Anton Popov, PhD
Associate Professor, Physical and Biomedical Electronics Dept., National Technical University of Ukraine
“Igor Sikorsky Kyiv Polytechnic Institute”,
Kusatsu, Japan, 2017
Contents
– What is EEG
– EEG methodology
– EEG spectral analysis
– EEG in epilepsy
– Epileptiform patterns localization
– Epileptic seizures prediction
– Heart activity in epilepsy
– Brain-Computer Interface
2
3
EEG history-1
In August 28, 1875 "Electrical currents of the brain". British Medical Journal, 2 (765): 278.
Richard Caton records brain
potentials from cortex of dogs
and apes
4
EEG history-2
5
In 1924, Hans Berger
• Recorded brain activity from the closed skull
• Reported brain activity changes according to thefunctional state of the brain (1929)
– Alpha and Beta rythms
– Sleep, Hypnosis, mental tasks, telepathic transmission
– Pathological states (epilepsy)
– Cerebral injury
– Drugs: phenobarbital,
morphine, cocaine
6Archive für Psychiatre und Nervenkrankheiten
Electroencephalography
• Is the measurement of the difference in voltage between two different cerebral locations, which then is studied over time
7
EEG definition
An electroencephalorgam (EEG) is a record of multichannel signal of potential differences, that can be registered on the intact head surface.
• EEG is the only measurement modality describing the functioning of the brain, since all processes in the brain related to the information processing are of electrical nature
8
EEG of healthy subject
9
Applications of EEG• Monitoring alertness, coma, and brain death;
• locating areas of damage following head injury, stroke, and tumour;
• testing afferent pathways (by evoked potentials);
• monitoring cognitive engagement (alpha rhythm);
• producing biofeedback situations;
• controlling anaesthesia depth (servo anaesthesia);
• investigating epilepsy and locating seizure origin;
• testing epilepsy drug effects;
• assisting in experimental cortical excision of epileptic focus;
• monitoring the brain development;
• testing drugs for convulsive effects;
• investigating sleep disorders and physiology;
• investigating mental disorders;
• providing a hybrid data recording system together with other imaging modalities;
• deciphering thoughts and intentions in BCI.10
Origin of EEG signal - 1
11
Origin of EEG signal - 1
12
Electrodes location –10/20 system
13
Nasion: point between the forehead and the skull
Inion: bump at the back of the skull
Location: Frontal, Temporal, Parietal, Occipital, Centralz for the central line
Numbers: Even numbers (2,4,6) - right hemisphere, odd (1,3,5) – left hemispere
EEG electrode placement
14
EEG recording
15
Channel: Recording from a pair of electrodes (here with a common reference: A1 –left ear)
Multichannel EEG recording: up to 256 channels recorded in parallelReference electrode: Mastoid, nose, ear lobe…
Unipolar (monopolar) measurement
Bipolar measurement
EEG reference systems
16
EEG recording
17
EEG recorders – 2
18
OpenEEG project
19
http://openeeg.sourceforge.net/doc/modeeg/modeeg.html
EEG of healthy subject
20
EEG rhythms – 1
21
22
Rhythm Frequency
(Hz)
Amplitude
(uV)
Location State of mind,
Source of activity
Alpha (α) 8 – 13 50 – 100 Occipital, Parietalregions
Adults, rest, eyes closed. Alpha blockadeoccurs when new stimulus is processed.Oscillating thalamic pacemaker neurons.
Beta (β) 14 - 40 2 – 20 Frontal region
Adult, mental activity.
Reflects specific information processing betweencortex and thalamus
Gamma (γ) 40-70 (100) 5 – 7 Variable location
Is believed to be reflecting mental activity, consciousness, meditation states etc. Associated with sensory processing.
Delta (δ) 0.5 – 4 Above 50 Variable location
Deep sleep, children in sleep.Oscillations in Thalamus and deep corticallayers. Usually inibited by ARAS (AscendingReticular Activation System)
Theta (θ) 5 – 7 5 – 100 Occipital Children, drowsy adult, sleepiness, emotional distress.Nucleus reticularis slows oscillating thalamicneurons , therefore sensory throughput to cortexdiminished.
Mu (μ) 8 – 13 up to 50 Motor cortex
Associated with sensorimotor transformation. Suppressed when person performs real of imaginary motor action.
Changes in EEG patterns
• Ageing: reduction of the alpha waves, bilateral slowing in the theta and delta waves, increase in beta wave activity(intellectual loss)
• Dementia: alpha rhythm slows, delta and theta wave activities increase, beta wave activity may decrease, epileptiform discharges and triphasic waves can appear
• Epileptic Seizure and Nonepileptic Attacks: sudden change of frequency, sudden desynchronization of activity, gradual change from chaotic to ordered waveforms
• Psychiatric Disorders: associated EEG features have been rarely analytically investigated
• External Effects: pharmacological and drug, music, cognitive
23
EEG spectral analysis
• In classical Fourier analysis the signal is represented as a sum of sinusoidal oscillations of harmonically increasing frequencies
24
EEG spectrum, ageing
25
EEG spectrum, sleep
26
EEG spectrum, drugs
27
EEG spectrum, dementia
28
Brain functional regions
29
Quantitative EEG (qEEG)
30
Topographical maps plot EEG data on a map of thebrain surface.
Usual data plotted:
• ERP maps– potential changes
• Spectral maps– frequency changes
• Statistical maps– comparison of
measurements
Time-frequency analysis
• TFA allows to track time changes in spectral contents of EEG signals
• Short-Time Fourier transform is used to analyze EEG
31
32
Real-time Brain activity mapping
33
Cross-spectra and Coherence
34
Cross-correlation function of two signals
Cross- spectra of two signals
Spectral coherence of two signals
Coherence during relaxation
35
Coherence during meditation
36
Epilepsy
• Epilepsy is a group of diseases characterised in most cases by repeated sudden unpredictable seizures caused by excessive abnormal synchronous activity of neuronal groups in the brain.
• Clinical manifestations of epilepsy are unforeseen and abrupt motor phenomena like tremor and convulsions, lost of consciousness, breathing pauses and respiratory changes, hallucinations, screaming and other psychic and sensory symptoms.
– Causes: insults, trauma, vascular diseases, infections, neurodegeneration, inflammations, cysts, tumors, genetic inclinations, developmental defects.
37
Epileptic seizures
38
• Partial (focal)• Simple partial
• Motor
• Somatosensory
• Autonomic
• Psychological
• Complex partial
• Simple partial with impaired consciousness
• Partial seizures with secondary generalization
• Generalized• Absence
• Tonic
• Clonic
• Tonic-clonic
• Atonic
• Myoclonic
EEG during epileptic seizure
39
EEG activity in Epilepsy
Epilepsy
Normal activity
Abnormal activity
Pre-ictal
Inter-ictal Ictal
Focal activity
Generalized activity
Multifocal
40
Targets of EEG analysis in epilepsy
Automated analysis of
EEG
Epileptic activity
detection
Epileptic spike
detection
Epileptic seizure analysis
Seizure detection
Seizure prediction
Seizure classification
Seizure focus localization
41
Epileptic seizure analysisFeatures of
seizure activity
Spectral and wavelet
characteristics
Amplitude distribution
Spatial distribution
Chaotic characteristics
42
Classificationof seizure
activity
Nearest neighbor
Artificial Neural Nets
Support Vector
Machines
Decision Trees
Fuzzy Classifiers
Epileptic Pattern Localization• Epileptic oscillation complexes in human electroencephalogram -
complexes of a sharp wave and a slow wave, indicating the presence of neuronal epileptic discharges in the brain.
43
44
45
Epileptiform Complexes – 1• At the early stages of disease, epileptiform complexes
are rare, distorted, non-prominent and low-amplitude.
46
Objective – to detect and properly localize the epileptiform complexes in EEG.
Early detection –earlier start of treatment –
more favorable prognosis of recovery
47
«generic» epileptiform complex
Parameters of possible acceptable distortions
Template for the Class of Epileptiform Complexes
1 1, 1,, min maxA K
i Q n zd f n f n
Pseudometric:
Template for the class: 1 1 1 1 11
,K K K K KKn C S A L V
Magnitude Alternation
t
U
Offset relative to isoelectric line
Magnitude Ratio AlternationDuration alternation
Presence of slow trend
Template matching – 1
48
1
11 12 11
21 22 22
1 2
N
N
K
M M MNM
x x xX
x x xXX
x x xX
1
1 M
ij ki kjk
c x xM
1
det 0KC I
1
11 12 1
21 22 2
1 2
N
N
K
N N NN
c c c
c c cC
c c c
1K k k kC v v
1 1,
maxK kk N
1 1 1 1K K K KC
1 1 2[ , , , ]K N Main eigenvector keeps the most information about member’s prevailing and dominating properties.
Template matching – 2
49
Duration:
Magnitude:
Offset:
Trend:
Magnitude Ratio:
Template matching – 3
50
TP – 85%
FP (while all diagnostically meaningful complexes are detected) – 15%
Sensitivity – 82%
Selectivity – 72%
Percentage P of TP (○) and FP (□) from threshold dА
Template matching – 4
51
Template matching – 5
52
Template matching – 6
53
Template matching – 7
54
Epileptic seizures control
55
Epileptic seizure prediction – 1
Epilepsy is a group of long-term neurological disorders characterized by epileptic seizures.
• About 1-2% of people worldwide (65 million)have epilepsy;
• 36% patients have seizures, that cannot becontrolled with medication
• Control of seizures:
– Detection
• Prediction
– Prevention
56
Seizure prediction and prevention
57
Neurostimualator
Simulator control
Classification and prediction
Multiparametric data analysis system
Feature extraction from EEG and/or ECG
Linear/nonlinear analysis and coupling estimation
Situation analysis and optimized feature set
Neurostimulator control strategy selection
Generation of signals for direct brain/nerve tissue stimulation
Neurostimulator with Feedback
58
Seizure predictionIctal
Pre-ictalInter-ictal
Seizure prediction approaches
Patient-specific
• Easier to build
• A lot of data from each patient is needed
• Possibility to use feature selection for each subject
Non-patient-specific
• Harder to build
• Comprehensive dataset is needed
• Extensive and descriptive set of features are needed due to inter-subject diversity of EEG characteristics
60
A4OP3
Слайд 60
A4 тут в тексте надо упомянуть и про пациент-специфический подход?Admin; 20.08.2016
OP3 да, спасибо. дописалOleg Panichev; 22.08.2016
61
Correlation features extraction
1
2 2
1 1
,
Nm m n ni i
imn
N Nm m n ni i
i i
x x x x
r
x x x x
Channel
Channel
Patient-specific classification
Window length for largest AUC. Mean AUC is 0.97±0.028.
Patient
Win
dow
length
, s
Classification performance in prediction of epileptic seizures
63
Seizure prediction performance (AUC values) for SVM classifier (left), and DA classifier (right) depending on epoch length and duration of pre-ictal segment for
one patient from group of patients with focal seizures
64
Derivation scheme influence
Permutation Entropy
65
PE is a descriptor of time series complexity, unpredictability, disorder, chaoticity, nonlinearity and stochasticity and can reflect complex nonlinear interconnections between anatomical and functional subsystems emerged in brain in healthy state and during various diseases.
1
j
q jp
N m l
!
1
log
,log !
m
j jj
norm
p p
PermEn m lm
66
`
Lag=1
Lag=2
Lag=3
Lag=4
67
0 1 2 3 4 5 6 7 8 9 10-50
0
50
x(t),
uV
0 1 2 3 4 5 6 7 8 9 10-50
0
50
x(t)
, uV
0 1 2 3 4 5 6 7 8 9 10-1000
0
1000
Time, sec
x(t)
, uV
0
50
100
23
45
67
80.4
0.6
0.8
1
1.2
1.4
LagOrder
PE
0.5
0.6
0.7
0.8
0.9
1
0
50
100
23
45
67
80.2
0.4
0.6
0.8
1
1.2
LagOrder
PE
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
100
101
102
103
0.4
0.5
0.6
0.7
0.8
0.9
1
Lag
PE
PE as a function of time lag and order
for tree types of EEG signals
PE saturates starting from some lag value, different for various signal types
Permutation Entropy averaged values for 5th order for different combination of lags and
downsampling factor
1 2 3 4 5 6 7 8 9 10
12
34
56
78
910
0
0.2
0.4
0.6
0.8
1
l
Order m=5
M
Me
an P
E
0.4
0.5
0.6
0.7
0.8
0.9
PE for seizure prediction
68
0 2000 4000 6000 8000 10000 12000-400
-200
0
200
400EEG Signal # 2
0 50 100 150 200 250 300 350 400 4500.7
0.8
0.9
1PermEn of order m=3
Interhemispheric Interactions- Left: mostly analytics (speech, input
signals), telic actions,
- Left hemisphere is mainly responsible for speech
- Right: interaction with environment, automatic actions, shape, distance and space perception. Signing!
- People with right hemisphere injuries have difficulties with perception, orientation, memory, face recognition and attention.
- Natural opposition, like sense-intuition, science-art, logic-mystics.
69
Synchrony and Symmetry of brain functioning
It is known that the electrical brain activity is more or less synchronous in two hemispheres.
Symmetry is not exact but is highly essential for healthy brain.
Spatial characteristics of brain activity:-Rhythm distribution between zones – areas of dominant activity;-Symmetry of bioelectric activity – similarity of oscillation parameters (e.g.
magnitude and frequency content) in different hemispheres.
• Applications of interhemispheric symmetry estimation:
- Response to functional tests, rhythm assimilation (hyperventilation, photostimulation etc.);
- Right- left-handed persons research;
- Dynamics of activity propagation, dispersion, penetration and entrapping;
- Pathological changes in symmetry (amplification or attenuation of differences);
- Epilepsy research (seizures etc.)70
Symmetry experiment setup
71
asymmetricleads
symmetric leads
Phase-Locking Value
72
Healthy EEG Epileptic EEG
Symmetric 0.22 ± 0.04 0.12 ± 0.02
Asymmetric 0.10 ± 0.01 0.03 ± 0.02
• In application to evaluation of degree of symmetry between two hemispheres in EEG of healthy people and in the presence of epileptiform activity, significant differencewas shown, and thus it could potentially be an important clinical diagnostic tool.
• EEG phase synchronization for main clinically meaningful bands was analyzed for healthy subjects and the difference between synchronization in symmetric and asymmetric channels was shown with high validity according to the Kruskal-Wallis criteria (p≤0.001) in all frequency bands.
Averaged synchronization
Synchronization for healthy brain
73
• EEG phase synchronization for main clinically meaningful bands was analyzed for healthy subjects and the difference between synchronization in symmetric and asymmetric channels was shown with enough validity according to the Kruskal-Wallis criteria.
Delta Theta Beta
Gamma 1, 2 and 3:symmetric, symmetric channels and surrogate signals1. Kolmogorov-Smirnov test (non-normal,
p=0.05)2. Kruskal-Wallis (distributions with equal means), existance of synchronization was validated at p≤0.001 (for gamma (p≤0.05 )
Alpha
Heart activity in epilepsy
• heart rate variability (HRV) is promicing for epileptic seizures’ onsets detection
• interactions between cortical and cardiac parameters might also help to contribute to a more accurate prediction of epileptic seizures.
74
HRV changes in epilepsy
• Recurrence plots – characteristics of periodicity and cycling behavior of time series.
75
Classification results
76
77
Nonlinear Methods- Phase locking- Transfer entropy- Mutual information- Synchronization likelihood- Granger causality- Cross-spectral analysis- Coherence- Multivariate autoregression- Scatter plots, recurrent plots
Quantitative analysis of interconnections
78
Synchronization between brain regions during movements
Brain-Computer Interface• Brain–computer interface (BCI) is a direct communication pathway
between the human brain and an external device.
79
BCI system
80
81
Device Allows Brain To Bypass Spinal Cord, Move Paralyzed Limbs – 2014
82http://wexnermedical.osu.edu/mediaroom/pressreleaselisting/new-device-allows-brain-to-bypass-spinal-cord-move-paralyzed-limbs
2 years later
83
OpenBCI Project
84
A customizable and fully open brain-computer interface platform that gives you access to high-quality brain wave data.
BCI for WoW
85