0_a_introduction into eeg processing

Upload: wendy-houston

Post on 06-Jul-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/16/2019 0_a_Introduction Into EEG Processing

    1/71

    BIOLOGICALBIOLOGICAL SIGNALSSIGNALS

    PROCESSINGPROCESSINGIntroductionIntroduction

     Assoc. Prof. Assoc. Prof. Vladim Vladim í írr KrajKrajččaa, MSc., PhD., MSc., PhD.

    Department of neurology Department of neurology Faculty Hospital Na Bulovce,Faculty Hospital Na Bulovce,

    Prague, Czech RepublicPrague, Czech Republic

    phonephone +420+420--22--66086608 23072307

    ee--mailmail krajcakrajca v  v @@fnbfnb.cz.cz

    mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]

  • 8/16/2019 0_a_Introduction Into EEG Processing

    2/71

    Source materials on webSource materials on web  www.skolicka.fbmi.cvut.cz www.skolicka.fbmi.cvut.cz

     password: signaly password: signalyContents of web pages:

    e-learning courses (2009, 10 lectures, in Czech)

    selected publications - international journals and conferences

     video - EEG, EMG, EP signals recording in EEG lab FN Bulovka andFN Ostrava

    requirements and terms of exams (enrolment in KOS)

    e-learning course (in Czech - pictures, formulas)

    Mohylová J., Krajča V. Zpracování signál ů v lékař ství , Ostrava 2007 animated examples and simulations (some in DOS),

     programs for simulation of filtering, brain mapping, cluster analysis…

    homeworks

    http://www.skolicka.fbmi.cvut.cz/http://www.skolicka.fbmi.cvut.cz/http://www.skolicka.fbmi.cvut.cz/

  • 8/16/2019 0_a_Introduction Into EEG Processing

    3/71

    MetMethods of computerhods of computer--assistedassisted

    electroencephalogramelectroencephalogram analysis realizedanalysis realizedinin EEGEEG lablab FN Na BulovceFN Na Bulovce

    Overview Overview 

    Examples of real world EEG biosignal processing.Case studies from clinical practice and research.

  • 8/16/2019 0_a_Introduction Into EEG Processing

    4/71

    Outline of lectureOutline of lecture

    1. Introduction – what it is electroencephalogram, properties ofthe signal

    2. Problems of visual evaluation

    3. Selection of proper methods of analysis

    4. Practical assets of methods5. Publications overview, courses , source materials

  • 8/16/2019 0_a_Introduction Into EEG Processing

    5/71

    1.1. What is electroencephalogramWhat is electroencephalogram (EEG) ?(EEG) ?

    • complex elektrical biosignal, measurement on the scull by

    electrodes (µV)• reflects the electrical activity of brain functions

    • apart from structural methods (CT - computer tomography), itshows dynamic functional manifestation of living brain

    • basic diagnostic tool for the treatment and diagnosis of epilepsy

    and sleep analysis• manifests behavioural states (psychology), consciousnes,

    disturbances of brain functions, sleep stages, coma, lesions

  • 8/16/2019 0_a_Introduction Into EEG Processing

    6/71

    1.a.1.a. Basic spectral activityBasic spectral activity -- waves, waves, graphoelementsgraphoelements

    • Principal frequency content for diagnosis (backgroundactivity)

    • Delta 0.5 - 4 Hz - children, deep sleep, tumours• Theta 4 - 8 Hz - normal•  Alpha 8 - 13 Hz - awake, eyes closed, blocked by eyes opened• Beta 13 - 22 Hz - normal, attention, pain, unrest

    • Graphoelements• epileptic abnormalities - transients - spikes, spike and wave complexes(SWC)

    • spike - transient clearly distinctive from the background activity,

     pointed peak, 70 msec wide (sharp wave 200msec)

    • Artefacts• Defects in recordings technical (50/60 Hz noise, drop off of electrode)

    and physiologic (sweating, movement) caused by external influence.Must be displaceed from recording (detection, ICA, PCA)

  • 8/16/2019 0_a_Introduction Into EEG Processing

    7/71

    1.b.1.b. ExamplesExamples

    Amplitude artefactEpileptic spikes

  • 8/16/2019 0_a_Introduction Into EEG Processing

    8/71

    2.2. Problems of Problems of  visual evaluation visual evaluation -- why why useusecomputercomputerss

    1. Visual evaluation / interpretation

    • subjective

    • „more art than science“ - evaluation depends on the teacher and hisexperience

    • education of top specialists is difficult ( 5 years of maturation)

    • tedious, requires permanent attention

    2. Length of recording

    • ambulatory recording - 20 min -> epileptic activity doesn't have to manifestitself 

    • long term monitoring is necessary (24 hours and more)

    • sleep analysis - 8 hours = 864 m of paper recording3. Archivation - (paper machines obsolete); DVD is compact, databases provide

    fast information retrieval

    4. Data processing (montage changing, filtering, not possible on papertraces )

    5. Numeric EEG quantification and analysis (qEEG, digital signal processing)

  • 8/16/2019 0_a_Introduction Into EEG Processing

    9/71

    2.a.2.a. Aims of computer Aims of computer--assisted analysisassisted analysis

    Support of physician's decision Extension of his capabilities by objective data Graphic data presentation Normal/abnormal activity distinction Signal classification Trends evaluation Data reduction and archiving Quantification - assesment of exact value of

    dominant frequency Hidden information extraction (not visible by a

    naked eye) Concentration on interesting parts of long-term

    recordings an skipping the uninterested parts (byartificial intelligence)

  • 8/16/2019 0_a_Introduction Into EEG Processing

    10/71

    3.3. LongLong--termterm xx detaildetailed analysised analysis

    • Long-term EEG• significant events

    detection• significant

    graphoelements (spikes)

    searching• trends evaluation

    • sleep stages classifiation

    • epileptic seizuresdetection

    • Detailed analysis• topographic mapping of

    frequenct bands power• local and interhemispheric

    coherence mapping

    • frequency shifts in spectrum• source derivation

    • dipole analysis

    •  phase specrum- epilepticfocus localization

  • 8/16/2019 0_a_Introduction Into EEG Processing

    11/71

    4.4. Methods overview Methods overview 

    PRE -PROCESSING

    DIGITAL FILTERING

    TRENDS REMOVING

    LONG- TERM

    PROCESSINGDETAILEDANALYSIS

    SPECTRUM

    ANALYSIS

    LOCALCOHERENCE

    WAVEFINDER 

    SPIKE

    DETECTIONCSA

    BRAIN

    MAPPING

    CORDANCE

    PRINCIPALCOMPONENT

    ANALYSIS

    INDEPENDENTCOMPONENT

    ANALYSIS

    NEURALNETWORKS

    PHASEMAPPING

    PCASEGMENT

    ATION

    ARTEFACTELIMINATION

    POWER

    SPECTRUMPhoticdriving

    3D PROJECTION3D PROJECTION

    LORETALORETA

    NEONATALSLEEP

    ANALYSIS

    NEONATALSLEEP

    ANALYSIS

    EEG SIGNAL

  • 8/16/2019 0_a_Introduction Into EEG Processing

    12/71

    CooperationCooperation

    withwith SZSZÚÚ

    4.1.4.1. CooperationCooperationCooperationCooperation withwith PsychiatricPsychiatric

    CenterCenter

    Cooperation withCooperation with

    AcademyAcademy of of ScienceScience

    Cooperation withCooperation withTechnicalTechnical

    University OstravaUniversity Ostrava

    PRE -PROCESSING

    DIGITAL FILTERING

    TRENDS REMOVING

    LONG- TERM

    PROCESSINGDETAILEDANALYSIS

    SPECTRUM

    ANALYSIS

    LOCALCOHERENCE

    WAVEFINDER 

    SPIKE

    DETECTIONCSA

    BRAIN

    MAPPING

    CORDANCE

    PRINCIPALCOMPONENT

    ANALYSIS

    INDEPENDENTCOMPONENT

    ANALYSIS

    NEURALNETWORKS

    PHASEMAPPING

    PCASEGMENT

    ATION

    ARTEFACTELIMINATION

    POWER

    SPECTRUMPhoticdriving

    CooperationCooperation

    withwith ČČ

    VUTVUTaandnd ÚÚPMDPMD

    3D PROJECTION3D PROJECTION

    LORETALORETA

    NEONATALSLEEP

    ANALYSIS

    NEONATALSLEEP

    ANALYSIS

    EEG SIGNAL

  • 8/16/2019 0_a_Introduction Into EEG Processing

    13/71

    Signal ProcessingSignal Processing -- visualisation visualisation

    A li d hi iA lit d t hi i

  • 8/16/2019 0_a_Introduction Into EEG Processing

    14/71

     

    13.5

    -98.3

    5.6

    85.0

    -100

    +100  

    13.5 -98.3

    85.0

    5.6

     1. iteration  2. iteration  Topographic map 

    Conversion of numbers

    into color scale

    Average from four

    neigbours 

    The new points are

    included into computing… . . . till the whole area is

    covered

     

    V  

    V  

    a    b  

    c  

     Amplitude topographic mapping Amplitude topographic mapping

  • 8/16/2019 0_a_Introduction Into EEG Processing

    15/71

    13.5 -98.3

    85.0

    5.6

    13.5

    -

    98.3

    5.6

    85.0

    mV

    a)

    2. iteration

    … new computed valuesenter next iteration

    Numbers are replaced by colors-100

    100

    µ Vb)

    Topographic map

    . . . final map

    1. iteration

    average of 4 neighbours

    c)

    Mapping of amplitudeMapping of amplitude1. In multichannel signal choose a time

    instant – Fig. a)

    2. Numerical values are replaced bycolors with from a color scale – Fig. b)

    3. In several iterations do interpolationfrom neigbouring points – Fig. c)

    4. Repeat (refine) interpolation byincluding the new values to cover all

    plane5. The values of EEG amplitude are

    color coded into topographic map.

    Example in DOS

  • 8/16/2019 0_a_Introduction Into EEG Processing

    16/71

    amplitude brain mapping - between cursors

  • 8/16/2019 0_a_Introduction Into EEG Processing

    17/71

    4.3.1.4.3.1. Signal processingSignal processing -- unprocessed signalunprocessed signal

  • 8/16/2019 0_a_Introduction Into EEG Processing

    18/71

    4.3.1.4.3.1. FIR filteringFIR filtering

  • 8/16/2019 0_a_Introduction Into EEG Processing

    19/71

    4.3.3.4.3.3. Spectral componentsSpectral components

    Signal + noise 50 Hz

    Filtering 0.5-25 Hz

  • 8/16/2019 0_a_Introduction Into EEG Processing

    20/71

    5.5. Spectrum anlysisSpectrum anlysis

     Assesment of frequency components in a single / all Assesment of frequency components in a single / all

    channelschannels

    Four main EEG frequency bands / single spectralFour main EEG frequency bands / single spectralline (power spectrum density)line (power spectrum density)

    S l i EEG 1 h l

  • 8/16/2019 0_a_Introduction Into EEG Processing

    21/71

    Spectrum analysisSpectrum analysis –  – EEG 1EEG 1 channelchannel

  • 8/16/2019 0_a_Introduction Into EEG Processing

    22/71

    S l iS l i ff

  • 8/16/2019 0_a_Introduction Into EEG Processing

    23/71

    Spectrum analysisSpectrum analysis -- frequency curvesfrequency curves

    Spectrum analysisSpectrum analysis -- detailed maps fordetailed maps for

  • 8/16/2019 0_a_Introduction Into EEG Processing

    24/71

    Spectrum analysisSpectrum analysis -- detailed maps fordetailed maps forevery frequency lineevery frequency line

    S l iS t l i f b df b d

  • 8/16/2019 0_a_Introduction Into EEG Processing

    25/71

    Spectrum analysisSpectrum analysis -- frequency bandsfrequency bands

    S t l iS t l i 3D3D d ld l h i l lih i l li

  • 8/16/2019 0_a_Introduction Into EEG Processing

    26/71

    Spectrum analysis,Spectrum analysis, 3D3D modelmodel -- spherical splinesspherical splines

  • 8/16/2019 0_a_Introduction Into EEG Processing

    27/71

  • 8/16/2019 0_a_Introduction Into EEG Processing

    28/71

  • 8/16/2019 0_a_Introduction Into EEG Processing

    29/71

    6.6. CaseCase reportreport :: Local and interhemispheric coherenceLocal and interhemispheric coherence

    Local coherence – indicator of cross correlation (cooperation) ofneighbouring areas of the brain cortex (prof. Rappelsberger)

    Interhemispheric coherence – analysis of interhemispheric synchrony

    Quantitative evaluation of focal lesion, not visible in native EEG

  • 8/16/2019 0_a_Introduction Into EEG Processing

    30/71

    Used also for topographic spectrum andUsed also for topographic spectrum and

    coherence mappingcoherence mapping

  • 8/16/2019 0_a_Introduction Into EEG Processing

    31/71

    Local coherenceLocal coherence -- normalized cross spectrumnormalized cross spectrum

    COHG

    G Gxyxy

    x y

    =

    2

    magnitude squared coherence

    [ ]G f E X f X f   x ( ) ( ) ( )= ⋅

    [ ]G f E X f Y f   x y ( ) ( ) ( )= ⋅

    auto-spectrum

    Cross-spectrum

    COHG

    G Gxy

    xy

    x y

    =

    amplitude coherence

    Examples (expert evaluation - without conclusive focal changes):

  • 8/16/2019 0_a_Introduction Into EEG Processing

    32/71

     

    Ischémie l.dx – vpravo (EEG závěr: …bez průkazných ložiskových změn) 

    Local coherence mapping - focus

    temporally right in agreement with

    CT

    Classic spectrum

    mapping - symmetrical

  • 8/16/2019 0_a_Introduction Into EEG Processing

    33/71

    Ictus (CMP)Ictus (CMP) -- right side hemiparesisright side hemiparesis

  • 8/16/2019 0_a_Introduction Into EEG Processing

    34/71

    7.7. LongLong--term analysisterm analysis CSACSA -- compressedcompressed

    spectral arraysspectral arrays CSACSA –  – compressedcompressed spectralspectral arraysarrays

    Spectral curves fromSpectral curves from 2 sec2 sec segments in pseudosegments in pseudo

    3D projection, like in theater3D projection, like in theater

    Intensive Care UnitsIntensive Care Units -- spectral shifting, brainspectral shifting, brain

    nmonitoring nmonitoring 

    77 CSACSA-- epileptic seizureepileptic seizure

  • 8/16/2019 0_a_Introduction Into EEG Processing

    35/71

    7.7. CSA CSA - epileptic seizureepileptic seizure

    CSACSA h f ich n of i

  • 8/16/2019 0_a_Introduction Into EEG Processing

    36/71

    CSACSA -- change of view change of view 

    CaseCase study:study: PhoticPhotic drivingdriving in workers exposed to mercuryin workers exposed to mercuryVV i i h fi i h f U b SZU b SZÚÚ

  • 8/16/2019 0_a_Introduction Into EEG Processing

    37/71

     Vapers Vapers -- cooperation with prof.cooperation with prof. Urbanem, SZUrbanem, SZÚÚ

    Photic driving – reaction of EEG to photostimulation fordifference freq of stroboscope (respons of harmonics in EEGcorresponding to stimulation).

     Aim – asses CSA for indication of changes in PD for early

    neurotoxical contamination of workers exposed to Mercuryvapors.

    Significant changes in PD in comparison with the control groupearly before clinical manifestations

    CSAsCSAs in ain a personperson withwith wellwell expressedexpressed photicphotic dridrivingving

  • 8/16/2019 0_a_Introduction Into EEG Processing

    38/71

    CSAsCSAs in ain a person person with with well well expressedexpressed photic photic dridri ving ving

    Backgroundα activity

    Driving on thefundamental

    frequency

    Driving on the

    1st harmonic

    frequency

    Driving on the

    2nd harmonic

    frequency

    CSAsCSAs in ain a personperson withwith absentabsent photicphotic drivingdriving

  • 8/16/2019 0_a_Introduction Into EEG Processing

    39/71

    CSAsCSAs in ain a person person with with absentabsent photic photic drivingdriving

    Background

    α activity

    Automatically generated protocolAutomatically generated protocol

  • 8/16/2019 0_a_Introduction Into EEG Processing

    40/71

     Automatically generated protocol Automatically generated protocol

    Quantitativeparams

    8.8. LongLong--term signal analysisterm signal analysis -- significantsignificant

  • 8/16/2019 0_a_Introduction Into EEG Processing

    41/71

    graphoelements identification by WFgraphoelements identification by WF

    WaveFinder: Hierarchic system of automatedEEG processing1. adaptive segmentation2. features extraction3. automatic classification4. visualization and quantification

    System propertiesSystem properties

  • 8/16/2019 0_a_Introduction Into EEG Processing

    42/71

    y p py p p

    LongLong--termterm (24(24 hours) signal processinghours) signal processing

    Warning of the physician about some eventWarning of the physician about some event -- physician canphysician canmake his own conclusion bazsed on the original EEGmake his own conclusion bazsed on the original EEG

    Mimetic techniqueMimetic technique -- physicianphysician´́s work immitations work immitation duringduring EEGEEGevaluationevaluation

    Precision is not a goalPrecision is not a goal -- suspected eventssuspected events 1010 %% ,, 55%%artefacts/false warningsartefacts/false warnings -- time saving for 24 hours of EEGtime saving for 24 hours of EEG

    User (physician) friendly, no parameter setting necessaryUser (physician) friendly, no parameter setting necessary

    Graphic presentation of resultsGraphic presentation of results

    TransparentTransparent -- switching to and from real original EEG signalswitching to and from real original EEG signal..

     Adaptive segmentation based on two connected windows Adaptive segmentation based on two connected windows

  • 8/16/2019 0_a_Introduction Into EEG Processing

    43/71

    EEG – nonstationary signal, but

    quasistationary (piece-wisestationary). Fixed segmentsdoes not respect signalproperties. Better way - divide itinto variable length segmentsdepending on the occurence of

    EEG graphoelementsAdaptive segmentation(Bodenstein and Praetorius,Michel and Houchin, Krajča,Varri) :

    1. Two connected windows aresliding along the signal

    2. The deviation of stationarityis estimated from thedifference of parameters of

    the two indows3. The segment boundary is

    located at the local maximaof the difference measure

      1 dvě spojená okna hranice segmentu

    2 míra rozdílu oken  3 lokální maximum

      mez

    Two connectedwindows

    Localmaximum

    Differencemeasure

    Segmentboundary

    Threshold for smallfluctuations

    A i i f iAd i i f i l h l

  • 8/16/2019 0_a_Introduction Into EEG Processing

    44/71

    Adaptive segmentation of a single channelAdaptive segmentation of a single channel

    EEG signal and

    segments boundary

    Amplitude differencemeasure

    Frequency differencemeasure

    Total differencemeasureand the threshold

    Multichannel adaptive segmentationMultichannel adaptive segmentation

  • 8/16/2019 0_a_Introduction Into EEG Processing

    45/71

    Color identification of graphoelements by cluster analysisColor identification of graphoelements by cluster analysis

  • 8/16/2019 0_a_Introduction Into EEG Processing

    46/71

    Features extractionFeatures extraction

  • 8/16/2019 0_a_Introduction Into EEG Processing

    47/71

    Classification should be relevant to the visualevaluation of the physician

    Set of 10 features describing the time and frequency propertiesof the signal graphoelements:

    variance of amplitude of the segment difference of maximal positive and minimal negative amplitude

    power spectrum in delta, theta, alpha, beta1 a beta2 frequency bands

    maximal value of the first derivative of the signal (proportional to the slope)

    maximal value of the second derivative of the signal (proportional to the sharpness)

    average value of the frequency of the signal

    now - research of the relevant features (selection, evaluation)

     Automated classification: supervised and unsupervised, Automated classification: supervised and unsupervised,l i l nd f zzclassical and fuzzy

  • 8/16/2019 0_a_Introduction Into EEG Processing

    48/71

    classical and fuzzyclassical and fuzzy MotivationMotivation -- to detect the hidden data structure, EEG graphoelementsto detect the hidden data structure, EEG graphoelements

    identificationidentification

    Methods of classification:Methods of classification: CLUSTER ANALYSISCLUSTER ANALYSIS

    searching the natural data structure (if exists)searching the natural data structure (if exists) learning without a teacher (unsupervised)learning without a teacher (unsupervised) -- NO apriori informationNO apriori information

    about the dataabout the data datadata in a cluster is more similar (close) than data in different clasin a cluster is more similar (close) than data in different classesses feature oriented classification methodfeature oriented classification method (statistical pattern recognition)(statistical pattern recognition) onon--lineline classification not possible (all segments enter the computing)classification not possible (all segments enter the computing)

    LEARNING CLASSIFIERLEARNING CLASSIFIER

    the new incomer is compared to etalon, prototype (identified bythe new incomer is compared to etalon, prototype (identified by aateacher during the learning phaseteacher during the learning phase

    problems with object unknown in the etalons selection during leaproblems with object unknown in the etalons selection during learningrningphasephase

    cluster analysis can be used in learning phase for etalonscluster analysis can be used in learning phase for etalonsidentificationidentification

    Classical methods andClassical methods and Neural Networks Classic andClassic and Fuzzy sets theorysets theory

    k k --meansmeans algorithm (MacQueen)algorithm (MacQueen)

  • 8/16/2019 0_a_Introduction Into EEG Processing

    49/71

    1. Cluster centers, prototypes are selectedrandomly (can be the data)

    2. The (Euclidean) distances forom thecenters to all the data are computed

    3. The object is classified to the clusterwith minimal distance to the center 

    4. The new centers ( of gravity) are re-

    computed

    5. If the optimum is not reached (classmembership does not change), go to thestep 2

    Learning kLearning k NN classifierNN classifier

  • 8/16/2019 0_a_Introduction Into EEG Processing

    50/71

    Learning k Learning k --NN classifierNN classifier

    LEARNING PHASE

    Etalons, prototypes are determined with the help of the teacher

    - which etalons belongs to the specific class

    CLASS1 CLASS 2 CLASS3

    CLASSIFICATION PHASE

    1. The newcomer is compared with all prototypes

    2. It is put into the closest class (most similar)

    3. Next objects are evaluated

    Hard and fuzzy sets (Lofti Zadeh 1970)Hard and fuzzy sets (Lofti Zadeh 1970)

  • 8/16/2019 0_a_Introduction Into EEG Processing

    51/71

    Hard and fuzzy sets (Lofti Zadeh, 1970)Hard and fuzzy sets (Lofti Zadeh, 1970)

    Classic sets no class sharing in different classes object belongs/does not belong to the set (class) black and white view

    Fuzzy sets multiple class membership with a different grade

    X={ x1, x2,....xN }, set of objects.

    Classic set | Fuzzy set _____________________________________|____________________________ 

    characteristic function u A: X - > {0,1} | generalization u A: X -> |

    1 , xk   ∈  Ai |

    u Ai(xk) = { | u Ai(xk)=uik fuzzy membership0 , xk   ∉  Ai |

    FuzzyFuzzy sets and classificationsets and classification

  • 8/16/2019 0_a_Introduction Into EEG Processing

    52/71

    FuzzyFu y sets and classificationsets and c ass cat on

      Tř ída 1 Tř ída 2 Tř ída 3

      0.70 0.30 0

     1.0 0 0

     0.20 0.25 0.55

     0 1.0 0

     0 0.80 0.20 smíšený objekt může patř it do r ůzných tř íd s r ůzným stupněmčlenství

      JABLKA VIŠNĚ  ŠVESTKY

    proměnný stupeňčlenství v intervalu 0 -

     FUZZY TŘÍDĚNÍ

    70 %  JABLKA

    30 %  VIŠNĚ

    70 %  JABLKA

     ?30 %  VIŠNĚ

      Tř ída 1 Tř ída 2 Tř ída 3

      0 1 0

      0 0 1

    1 0 0

    1 0 0

      1 0 0  smíšený objekt musí 

    být zař azen do jedné z tř íd

      JABLKA VIŠNĚ  ŠVESTKY

      0 ... patř í do tř ídy

      1 ... nepatř í

      KLASICKÉ (HARD) TŘÍDĚNÍ

    OVOCNÉMOŠTY

     Automated classification of EEG graphoelements Automated classification of EEG graphoelements

  • 8/16/2019 0_a_Introduction Into EEG Processing

    53/71

    MotivationMotivation -- disclose the hidden structure of data, EEGdisclose the hidden structure of data, EEGgraphoelements identification.graphoelements identification.

    MethodMethod –  – Cluster analysisCluster analysis

    Example of classic clusteringExample of classic clustering

  • 8/16/2019 0_a_Introduction Into EEG Processing

    54/71

    Class 6 with mixed objects of class 5Class 6 with mixed objects of class 5Classic class number 5.

    Fuzzy elimination of hybrid segments (grey color)Fuzzy elimination of hybrid segments (grey color)

  • 8/16/2019 0_a_Introduction Into EEG Processing

    55/71

    Fuzzy elimination of hybrid segments (grey color)y y d g (g y )

    alpha cut 0.5

    Summary sheetsSummary sheetsTypical, representative segments

  • 8/16/2019 0_a_Introduction Into EEG Processing

    56/71

    yp , p g

    (closest to the center of gravity),

    percentual occurence - multichannel

    summary information

    Temporal profile shows dynamic

    class membership in the course of

    time

    Classes are ranged according to the

    decreasing amplitudeColor - indicates the segment class

    membership in original EEG recording

    Color graphic EEG segments identificationColor graphic EEG segments identification

  • 8/16/2019 0_a_Introduction Into EEG Processing

    57/71

    Significant graphoelements are identified by a color in the original EEG record

    Temporal profilesTemporal profiles -- dynamic changes in EEG, significant events detectiondynamic changes in EEG, significant events detection

  • 8/16/2019 0_a_Introduction Into EEG Processing

    58/71

    Lékař  má k dispozici efektivní nástroj,

    umožňující pomocí kursoru vybrat př íslušnou

    část originálního záznamu a pokračovat v

     prohlížení buďto v časové oblasti EEG, nebo ve

    schematickém diagramu. K tomu může vyvolat

    sumární multikanálovou informaci o typu EEGaktivity

    EEG with a seizureEEG with a seizure

  • 8/16/2019 0_a_Introduction Into EEG Processing

    59/71

     Časový profil dlouhodobých EEG. Př íklad zobrazení 20 minut. Možnost zobrazit až 3 hodiny

    záznamu na edné stránce. Dva e ile tické záchvat o trvání 2 minut .

     Z ačátek epi lept ického záchva tu v 11 m inutě  ,vybraný k ursorem z

    Sleep EEG analysisSleep EEG analysis (1.5(1.5 hour/page,hour/page, 11 channel)channel)

  • 8/16/2019 0_a_Introduction Into EEG Processing

    60/71

    Comparison with another methodsComparison with another methods -- CSACSA -- spectral informationspectral information only,only,shape of graphoelements is lostshape of graphoelements is lost

  • 8/16/2019 0_a_Introduction Into EEG Processing

    61/71

    CaseCase study:study: Detection and quantification of sleep stages inDetection and quantification of sleep stages innewbornsnewborns (Dr.(Dr. Paul,Paul, ÚÚPMD PodolPMD Podolíí))

  • 8/16/2019 0_a_Introduction Into EEG Processing

    62/71

    Temporal profiles processingTemporal profiles processing -- sleep stages detection in neonatal EEGsleep stages detection in neonatal EEG (120(120minmin of recordings)of recordings)

  • 8/16/2019 0_a_Introduction Into EEG Processing

    63/71

    VISUAL EVALUATION

    AWAKE QUIET SLEEP ACTIVE SLEEP QUIET SLEEP ACTIVE SLEEP QSLEEP AWAKE

    Selected publications with impact factor - www.skolicka.fbmi.cvut.cz

  • 8/16/2019 0_a_Introduction Into EEG Processing

    64/71

    Book chapterBook chapter

  • 8/16/2019 0_a_Introduction Into EEG Processing

    65/71

    pp

    Book chapterBook chapter

  • 8/16/2019 0_a_Introduction Into EEG Processing

    66/71

    pp

    CoursesCourses

  • 8/16/2019 0_a_Introduction Into EEG Processing

    67/71

    Cooperation with Psychiatric CenterCooperation with Psychiatric Center

  • 8/16/2019 0_a_Introduction Into EEG Processing

    68/71

    Prague (Dr. Brunovsky, PhD)Prague (Dr. Brunovsky, PhD)

  • 8/16/2019 0_a_Introduction Into EEG Processing

    69/71

  • 8/16/2019 0_a_Introduction Into EEG Processing

    70/71

  • 8/16/2019 0_a_Introduction Into EEG Processing

    71/71