#LASSIlCATIONOF(UMAN(EART!BNORMALITYUSING4IMEFREQUENCYAND)MAGE 0ROCESSING4ECHNIQUE
&ADZLUL2AHIMI!HMAD"USTAMI-OHD(ANIF-D3AAD-OHD*AILANI-OHD.OR"ILKIS"ANU!ZIZ
-%-3!UTOMOTIVE2ESEARCH'ROUP
$EPARTMENTOF-ECHANICALAND-ATERIALS%NGINEERING 5NIVERSITI+EBANGSAAN-ALAYSIA
5+-"ANGI3ELANGOR -ALAYSIA
0EDIATRIC$EPARTMENT
&ACULTYOF-EDICINE 5NIVERSITI+EBANGSAAN-ALAYSIA
"ANDAR4UN2AZAK+UALA,UMPUR -ALAYSIA
2ECEIVED$ATETH!UGUST !CCEPTED$ATETH&EBRUARY
!"342!#4
4HIS PAPER DESCRIBES HEART ABNORMALITIES CLASSIlCATION PROCEDURES UTILISING FEATURES OBTAINED FROM TIMEFREQUENCYSPECTOGRAMOF%#'HEARTANDIMAGEPROCESSINGTECHNIQUES%NHANCEDSPATIALFEATURES OFTIMEFREQUENCYSPECTOGRAMWEREEXTRACTEDANDFEDINTOAFORWARDCHAININGEXPERTSYSTEMANDTHE CORRESPONDINGABNORMALITIESWEREIDENTIlED!CONlDENCEFACTORISCALCULATEDFOREVERYCLASSIlCATION RESULTINDICATINGTHEDEGREOFBELIEFTHATTHECLASSIlCATIONISTRUE)TWASOBSERVEDTHATTHECLASSIlCATION METHODWASABLETOGIVECORRECTCLASSIlCATIONBASEDONFEATURESTHATWASEXTRACTEDFROMDATASETS WHICHWEREINCLUDEDINTHEKNOWLEDGEBASEANDDATASETSWHICHWERENOTINCLUDEDINTHEKNOWLEDGE BASE
+EYWORDS(EARTABNORMALITIESCLASSIlCATIONEXPERTSYSTEMSIMULTANEOUSTIMEFREQUENCYANALYSIS
!"342!+
+ERTASINIMENERANGKANTENTANGPROSEDURKLASIlKASIJANTUNGYANGTIDAKNORMALYANGDIDAPATIDARIPADA SPEKTROGRAMMASAFREKUENSI%#'JANTUNGDANTEKNIKTEKNIKPEMPROSESANIMEJ#IRICIRIIMEJYANGTELAH DITINGKATKANDARIPADASPEKTROGRAMMASAFREKUENSIDIEKSTRAKDANDIMASUKKANKEDALAMRANTAIANKEHADAPAN SISTEMPAKARDANSETERUSNYATAHAPKETIDAKNORMALANJANTUNGTERSEBUTDIKENALPASTI&AKTORKEYAKINANYANG DIHITUNGPADASETIAPKEPUTUSANPENGKLASIlKASIANMENUNJUKKANDARJAHKEYAKINANPADASETIAPPROSESTERSEBUT
$ARIPADAUJIANDIDAPATIKAEDAHKLASIlKASIINIMAMPUMEMBERIKANKEPUTUSANTEPATTERHADAPSET DATAYANGCIRINYADIEKSTRAKDANDIMASUKKANKEDALAMPENGKALANPENGETAHUANSERTASETDATAYANGTIDAK DIMASUKKANKEDALAMPENGKALANPENGETAHUAN
+ATA+UNCI+LASIlKASIJANTUNGTIDAKNORMALSISTEMPAKARANALISISMASAFREKUENSISERENTAK
).42/$5#4)/.
2ESEARCH"ACKGROUND
)T IS GENERALLY KNOWN THAT BY MEASURING AND OBSERVING THE ELECTROCARDIOGRAM %#' OF A HUMAN HEART A QUALIlED MEDICAL PRACTITIONER WOULDBEABLETODETERMINETHECONDITIONOFTHE HEARTWHETHERTHEHEARTISNORMALORNOTNORMAL (OWEVERTHEPROCESSOFDEDUCINGTHECONDITION OFTHEHUMANHEARTUSINGTIMEDOMAINSIGNALS ONLYISDIFlCULTDUETOTHECHARACTERISTICSOFTHE
%#'PLOTINTIMEDOMAIN
4HE%#'OFANORMALHEARTRATECONSISTSOFTHE 0WAVE123COMPLEXAND4WAVEANDREPEATSAS SHOWNIN&IGURE
4HE OBJECTIVE OF THIS PAPER IS TO STUDY AND CLASSIFY HUMAN HEART ABNORMALITIES USING INFORMATION OBTAINED FROM TIMEFREQUENCY SPECTROGRAMANDIMAGEPROCESSINGTECHNIQUE
&)'52%01234COMPLEXOFAN%#'PULSE%NDERLEETAL P
P - R Q
Q - T QRS
S T
T T
4(%/29
4IMEAND&REQUENCY$OMAIN&EATURES
4IMEDOMAIN%#'PLOTSLACKTHESIGNALINTENSITY DISPLAYOFTHEFREQUENCYDOMAINCOMPONENTS4HIS ISAGREATLOSSASFREQUENCYDOMAINCOMPONENTS CONTRIBUTE SIGNIlCANTLY IN DETERMINING UNIQUE FEATURES OF MOST ENGINEERING AND SCIENTIFIC SIGNALS$RIPPS
#ONSEQUENTLYTHEFREQUENCYANALYSISOFSUCH SIGNALS VIA &OURIER TECHNIQUE IS FUNDAMENTALLY UNSATISFACTORY SINCE THEY ARE BASED UPON MODELINGTHESIGNALASALINEARCOMBINATIONOF SINUSOIDS EXTENDING THROUGHOUT THE DURATION OF THE SIGNAL4HE &OURIER ANALYSIS IS GOOD AT
DETERMINING FREQUENCIES PRESENTED IE IT PROVIDES GOOD FREQUENCY DISCRIMINATION BUT POOR AT PINPOINTING WHEN THESE FREQUENCIES OCCURIEITHASAPOORTIMELOCALIZATION #ROWE
&)'52%%#'0LOTSINA 4IME$OMAINB &REQUENCY$OMAINC 4IME&REQUENCY$OMAIN
A B C
(LAWATSCH "OUDREAUX"ARTELS (AGHIGHI-OOD4ORRY
0OPULARSIMULTANEOUSTIMEFREQUENCYANALYSIS TECHNIQUESINCLUDETHE7AVELET4RANSFORM)KEDA ETAL ANDTHE3HORT4IME&OURIER4RANSFORM 4HEEXAMPLEOFTIMEDOMAINFREQUENCYDOMAIN AND SIMULTANEOUS TIMEFREQUENCY DOMAIN DISPLAYOF%#'SIGNALSISSHOWNIN&IGURE 3HORT4IME&OURIER4RANSFORM34&4
4HESHORTTIME&OURIERTRANSFORM34&4 ISALINEAR TIMEFREQUENCY REPRESENTATION 4&2 USED TO PRESENTCHANGESINTHESIGNALTHATVARYWITHTIME 4HE&OURIERTRANSFORMDOESNOTEXPLICITLYSHOW THETIMELOCATIONOFTHEFREQUENCYCOMPONENTS BUTSOMEFORMOFTIMELOCATIONCANBEOBTAINED BY USING A SUITABLE PREWINDOWING (LAWATSCH
"OUDREAUX"ARTELS 4HE34&4APPROACH IS TO PERFORM A &OURIER4RANSFORM ON A SMALL SECTIONWINDOW OFDATAATSPECIlCTIMEFROMALL SIGNALDATATHUSMAPPINGTHESIGNALINTOATWO DIMENSIONAL$ FUNCTIONOFTIMEANDFREQUENCY 4HETRANSFORMISDESCRIBEDMATHEMATICALLYAS
X
W,a x t g t( ) ( a e) j tW dt
d d
°
WHEREGT MAYBEDElNEDASASIMPLEBOXOR PULSEFUNCTION)NDISCRETEFUNCTION34&4ISTIME DEPENDENT&OURIERTRANSFORMOFSEQUENCEX;N=
ANDDElNEDBY
WHEREω;N=ISASUITABLYCHOSENWINDOWSEQUENCE )T SHOULD BE NOTED THAT THE FUNCTION OF THE WINDOW IS TO EXTRACT A FINITE LENGTH PORTION OF THE SEQUENCE X;N= SUCH THAT THE SPECTRAL CHARACTERISTICS OF THE SECTION EXTRACTED ARE APPROXIMATELY STATIONARY OVER THE DURATION OF THEWINDOWFORPRACTICALPURPOSES
.OTE THAT IFω;N= THE DElNITION OF 34&4 GIVENIN%QUATION REDUCESTOTHECONVENTION DISCRETETIME &OURIER TRANSFORM $4&4 OF X;N=
(OWEVER EVEN THOUGH THE $4&4 OF X;N= EXISTS UNDER CERTAIN WELLDEFINED CONDITIONS THE WINDOWEDSEQUENCEIN%QUATIONS BEINGlNITE INLENGTHENSURESTHEEXISTENCEOFTHE34&4FORANY SEQUENCEX;N=)TSHOULDBENOTEDALSOTHATUNLIKE THECONVENTIONAL$4&4THE34&4ISAFUNCTIONOF TWOVARIABLESTHEINTEGERVARIABLETIMEINDEXN ANDTHECONTINUOUSFREQUENCYVARIABLEω)TALSO FOLLOWSFROMTHEDElNITIONOF%QUATION THAT 834&4EJωN ISAPERIODICFUNCTIONOFωWITHAPERIOD π
)N THIS STUDY THE "LACKMAN WINDOW WITH A WINDOWSIZEOFOVERLAPPING DISCRETE DATAOUTOFATOTALOFDISCRETEDATAWASUSED 4HE "LACKMANN METHOD OF WINDOWING WAS CHOSENBECAUSEITGIVESMINIMUMVARIATIONON SPECTRUMSHAPEANDCOLOURCOMPAREDTOOTHER TECHNIQUES "Y OBSERVING THE VIEW OF SPECTRAL PEAK "LACKMAN HAS THE WIDEST MAIN LOBE BUT THE LOWEST AMPLITUDE TAILS 3MITH 4HE DISCRETEVERSIONOFTHE"LACKMANWINDOWCANBE DESCRIBEDBYTHEEQUATIONBELOW
w k k
n k n
[ ] . . cos
. cos
¥
§¦
´
¶µ
¥
§¦
1 0 42 0 5 2 1 0 08 4
1 P P ´´
¶µ
k 0,...,n 1
,INEAR3PATIAL$OMAIN&ILTER
/NE OF THE SIMPLEST LINEAR SPATIAL IMAGE PROCESSING TECHNIQUES USED IN MACHINE VISION ISTHECONVOLUTIONOPERATION4HEOPERATIONCAN BEDESCRIBEDASFOLLOWS
&OR ANY GIVEN PLANAR IMAGE 0 AND A X ELEMENT MASK ' CAN BE DESCRIBED BY BOTH EQUATIONS AND BELOW
P
g g g g g g g g
g g g g g g g g
g g Px y Px y Px y Px
n n n n n n
1 1 1 1 1 nn n
n n n n n n n n
n
y
x y x y x y x y
x
g g
g g P P P P g g
g g P
2 1
1 1 2
1yy x y x y x y
x y x y
n n n n n n n
n n n
P P P g g
g g P P
1 1 1 1 2 1
1 2 nn n n n n
n n n n n
P P g g
g g P P P
x y x y
x y x y x
2 1 2 2 2
1 3 3 1yyn Pxn yn g g
g g g g g g g g
¨
ª
©©
©©
©©
©©
©©
©
·
¹
¸¸
¸¸
¸¸
¸¸
3 2 3
¸¸¸
¸
G
g g g
g g g
g g g
¨
ª
©©
©©
·
¹
¸¸
¸¸
11 12 13
21 22 23
31 32 33
THECONVOLUTEDIMAGE0ISGIVENAS
P x y G i j
P x i y j
j n
i m
* ( , ) ( , )
( , )
¤
¤
0 1
0 1
1 1
WHERE
0 0LANARIMAGE ' THEMASK
X XAXISCOORDINATEFOR0 Y YAXISCOORDINATEFOR0 I XAXISCOORDINATEFOR' J YAXISCOORDINATEFOR'
$IFFERENTCOEFlCIENTVALUEWILLGIVEDIFFERENT lLTERBEHAVIOUR'ONZALES7OODS 'AUSSIAN&ILTER
4HE 'AUSSIAN lLTER SMOOTHES A GIVEN IMAGE 0 'ONZALES7OODS )TISMADEFROMTHE SAMESTRUCTUREIN AND WITHSPECIALIZED VALUE OF FILTER COEFFICIENT 4HE VALUES VARIES ACCORDINGTOTHEREQUIREMENTANDSPECIlCATION
!SETOFVALUESISSHOWNINEQUATION
G 0 011 0 084 0 011 0 084 0 619 0 084 0 011 0 084 0 011
. . .
. . .
. . .
¨¨
ª
©©
©
·
¹
¸¸
¸
3OBEL%DGE$ETECTION
)N 3OBEL %DGE DETECTION TWO OPERATORS WERE USEDGx AND GyTOCALCULATEAPPROXIMATIONSOF THEDERIVATIVESONEFORHORIZONTALCHANGESAND ONEFORVERTICAL4HE3OBELOPERATORSCALCULATES THE GRADIENT OF THE IMAGE INTENSITY AT EACH POINTGIVINGTHEDIRECTIONOFTHELARGESTPOSSIBLE INCREASEFROMLIGHTTODARKANDTHERATEOFCHANGE INTHATDIRECTION4HERESULTTHEREFORESHOWSHOW hABRUPTLYv ORhSMOOTHLYv THE IMAGE CHANGES AT THATPOINTANDTHEREFOREHOWLIKELYITISTHATPART OFTHEIMAGEREPRESENTSANEDGEASWELLASHOW THAT EDGE IS LIKELY TO BE ORIENTED 'ONZALES 7OODS
4HEOPERATORSFOLLOWTHESAMESTRUCTUREAS )F0ISDElNEDASTHESOURCEIMAGEAND'XAND'Y ARETHETWOOPERATORSDESCRIBEDABOVETHELATTER ARECOMPUTEDAS
$Px Gx*P and $Py Gy *P WHERE
G and G
x
y
¨ ª
©©
©
·
¹
¸¸
¸
1 0 1
2 0 2
1 0 1
1 2 1
0 0 0
1 2
¨ ª
©©
©
·
¹
¸¸ 1 ¸
!T EACH POINT IN THE IMAGE THE RESULTING GRADIENT APPROXIMATIONS CAN BE COMBINED TO GIVETHEGRADIENTMAGNITUDEUSING
$ $P Px2 $Py2
5SINGTHISINFORMATIONTHEGRADIENTSDIRECTIONIS CALCULATEDASBELOW
1 $
$
¥
§¦
´
¶µ arctan P
P
y
x
WHERE FOR EXAMPLEΘ IS FOR A VERTICAL EDGE WHICHISDARKERONTHELEFTSIDE
.ON,INEAR 3PATIAL $OMAIN &ILTER -EDIAN
&ILTER
!MEDIANlLTERISANONLINEARlLTER)NAMEDIAN FILTER THE MEDIAN VALUE FOR THE PIXELS IN THE PROCESSEDWINDOWISUSEDTOREPLACETHECURRENT PIXELUNDERPROCESSED'ONZALES7OODS
&OREXAMPLEAPIXELWITHTHEABSOLUTEVALUEOF THATISSURROUNDEDBYOTHERPIXELSSHOWN INEQUATION BELOWISCONSIDERED%ACHPIXELS VALUE IS THEN ARRANGED INCREMENTALLY AND THE ORIGINALPIXELVALUEINTHEMIDDLEISTHENREPLACED WITHANEWVALUEWHICHISTHEVALUEMEDIANOFTHE ARRANGEDPIXELS'ONZALES7OODS &ISHER ETAL %QUATION DESCRIBESHOWMEDIAN lLTERISIMPLEMENTED
P
g g g g g g g g g g g g g g g g
g g g g g
g g g g g
g
124 126 127 120 150 125 g
g g g g
g g g g g g g g g g g g g g g g g g g g g g g g
115 119 123
¨
ª
©©
©©
©©
©©©
©©
·
¹
¸¸
¸¸
¸¸
¸¸
¸¸
/RIGINAL6ALUE
.EIGHBORHOODVALUES
-EDIAN6ALUE
0ULSE#OUNTING&ROM"INARY)MAGE
4HE SPECTROGRAM OBTAINED FROM THE 34&4 OPERATION IS CONVERTED INTO GRAYSCALE AND BINARY THRESHOLDED TO PRODUCE A BLACK AND WHITE ONLY REPRESENTATION OF THE ORIGINAL 34&4 SPECTROGRAM
!TASELECTEDFREQUENCYFSELECTED(Z THE NUMBEROFTRANSITIONSMADEBYPIXELVALUEARE CALCULATED&OREXAMPLEIFALINEOFPIXELATANY HEIGHTISEXTRACTEDTHEFOLLOWINGREPRESENTATION MAYBEOBTAINED
&ROMTHEEXAMPLEABOVETRANSITIONS.4 CAN BEOBTAINED4HEPIXELSLOCATEDINAPULSEAREA
AREINDICATEDBY3INCETHETRANSITIONINCLUDED FROMTOANDTOTHENUMBEROFTOTALPULSE IS ACTUALLY .0 .4 )N THE ABOVE EXAMPLE THEREAREACTUALLYPULSESDETECTED4HIS METHODISBYFARNOTTHEMOSTROBUSTANDPRECISE PULSE COUNTING METHOD BUT IT WAS OBSERVED THROUGHOUTTHESTUDYTHATTHEPRODUCEDRESULTS WEREACCEPTABLE
%ULER.UMBER
%ULER .UMBER IS DEFINED AS THE NUMBER OF CONNECTED COMPONENTS )T IS A TOPOLOGICAL PROPERTY FOR REGION DESCRIPTION4HE NUMBER OFHOLES(ANDCONNECTEDCOMPONENTS#INAN IMAGECANBEUSEDTODElNETHE%ULER.UMBER
%
%#n(
4HE REGIONS ARE SHOWN IN &IGURE FOR EXAMPLE IF THE %ULER NUMBERS EQUAL TO AND
&)'52%2EGIONSWITH%ULERNUMBEREQUALTOAND
RESPECTIVELYBECAUSETHELEFTlGUREHASONE CONNECTED COMPONENT AND ONE HOLE AND THE RIGHTCOMPONENTHASONECONNECTEDCOMPONENT BUTTWOHOLES'ONZALES7OODS
%80%2)-%.4!,-%4(/$
4YPES /F )NVESTIGATED (EART !BNORMALITIES )NVESTIGATED
4HE TYPES OF HEART ABNORMALITIES WHICH WERE STUDIEDINTHISRESEARCHAREDESCRIBEDIN4ABLE 3ELECTIONOF&REQUENCY"AND
4HE FREQUENCY BAND FOR THE %#' SIGNAL WAS DECIDEDTOBEBETWEEN(ERTZ(Z TO(Z
$ETAILEDSIGNALANALYSISBYPREVIOUSRESEARCHERS INDICATED THAT THE 0WAVE AND4WAVE MAINLY CONSIST OF FREQUENCY COMPONENTS BELOW (Z4HE 2WAVE MAINLY CONSIST OF FREQUENCY COMPONENTS BELOW (Z BUT IT ALSO CONSISTS
OF SOME FREQUENCY COMPONENTS ABOVE (Z 5SINGABANDPASSlLTERTHESIGNALBANDWIDTHWAS SELECTEDBETWEENTO(Z3UCHAlLTERWILL EFFECTIVELYREDUCE(ZNOISENORMALLYACQUIRED THROUGH POWERLINE AND HAVE LITTLE EFFECTS ON 0WAVEAND4WAVEWHICHlNALLYPRODUCESOME ACCEPTABLE DISTORTION ON THE 2WAVE #UTOFF FREQUENCY ABOVE (Z WAS CHOSEN TO AVOID THELOWFREQUENCYNOISEDUETORESPIRATIONAND ELECTRODEMOVEMENTBELOW(Z
&EATURES/BSERVED
4HE FEATURES FROM THE PROCESSED SPECTROGRAM IMAGE WHICH WAS EXTRACTED AND USED FOR CLASSIlCATIONISDESCRIBEDIN4ABLE
#LASSIlCATION-ETHOD
4HEABNORMALITIESWERECLASSIlEDUSINGAFORWARD CHAINING %XPERT 3YSTEMS AND A CORRESPONDING
OVERALL CERTAINTY FACTOR CF IS CALCULATED FOR EVERYCONSEQUENCES4HEREWERERULESINTHE +NOWLEDGE"ASEOFTHE%XPERT3YSTEM)NDIVIDUAL CERTAINTY FACTORS FOR EVERY CONSEQUENCE ARE EMPIRICALLY CALCULATED FROM AVAILABLE DATA 4HEYREPRESENTTHEPROBABILITYPjki OFTHEPATIENT WHICHDIAGNOSEDWITHABNORMALITYIIFFEATUREJ IS EQUIVALENT TO VALUEK 3INCE THE SAME SET OF CONSEQUENCES IS OBTAINED AS A RESULT OF THE EXECUTION OF TWO OR MORE RULES THE INDIVIDUAL CERTAINTYFACTORSOFTHESERULESISMERGEDTOGIVE ACOMBINEDCERTAINTYFACTORFORAHYPOTHESIS4HE KNOWLEDGEBASECONSISTSOFTHEFOLLOWINGRULES
2ULE )& 0ULSE7IDTHIS07
4(%. !BNORMALITYIS.ORMAL[CFCF] !BNORMALITY IS !TRIAL&IBRILLATION
[CFCF]
!BNORMALITY IS 3UPRAVENTRICULAR
!RRHYTHMIA[CFCF]
!BNORMALIT Y I S 6ENTRICULAR 4ACHY!RRHYTHMIA[CFCF]
!BNORMALITYIS
-YOCARDIAL)NFARCTION [CFCF]
2ULE )& 0ULSE7IDTHIS07
4(%. !BNORMALITYIS.ORMAL[CFCF] 4!",%&EATURES/BSERVED
&EATURES $ESCRIPTION
0ULSE7IDTH 4HEWIDTHOFINDIVIDUALPULSESINTHESPECTROGRAM (EIGHT 4HEHEIGHTOFINDIVIDUALPULSESINTHESPECTROGRAM
.UMBEROF0ULSE 4HENUMBEROFPULSESRECORDEDINSECONDINTHESPECTROGRAM )NTENSITY 4HEINTENSITYOFEACHPULSESINTHESPECTROGRAM
%ULER.UMBERFOR INDIVIDUALPULSE
4HE%ULER.UMBERCALCULATEDONINDIVIDUALPULSEINTHESPECTROGRAM %ULER.UMBERFORTHE
WHOLEIMAGE
4HE%ULER.UMBERCALCULATEDONTHEOVERALLIMAGEOFTHE SPECTROGRAM
%ULER.UMBERFOR INDIVIDUALPULSEAFTER 3OBELEDGEDETECTION OPERATION
4HE%ULER.UMBERCALCULATEDONTHEIMAGEOFCONTAININGEDGESOFA SINGLEPULSE
%ULER.UMBERFORTHE WHOLEIMAGEAFTER 3OBELEDGEDETECTION OPERATION
4HE%ULER.UMBERCALCULATEDFORTHEEDGESOFTHEOVERALLIMAGEOF THESPECTROGRAM
4!",%4YPEOF(EART!BNORMALITIESAND4HEIR$ESCRIPTION
!BNORMALITY $ESCRIPTION
.ORMAL .ORMALWORKINGHEART !TRIAL&IBRILLATION
!TRIAL&IBRILLATIONISANABNORMALITYOFTHEHEARTINWHICHTHEATRIALOFHEART ISNOTDEPOLARIZEDCOMPLETELYACTIVATED 4HE0PEAKOFTHETIMEDOMAIN
%#'ISNOTVERYCLEARBUTINTIMEFREQUENCYDOMAINTHEPULSESARENARROWER COMPAREDTOTHEPULSESFROMTHENORMALHEART
3UPRAVENTRICULAR
!RRHYTHMIA
(EARTS WITH 3UPRAVENTRICULAR !RRHYTHMIA PRODUCES THE MAXIMUM PEAK FREQUENCYBELOW(ERTZ
6ENTRICULAR 4ACHY!RRHYTHMIA
)N THIS KIND OF ABNORMALITY THE NUMBER OF PULSES IN EVERY PERIOD IS CONSIDERABLY BIGGER AS COMPARED TO OTHER KIND OF ABNORMALITIES AND NORMALHEART
-YOCARDIAL )NFARCTION
4HISKINDOFABNORMALITYGENERATESPULSESWHICHARENOTlRMCOMPARED TOTHENORMALPULSEDUETOTHEWEAKNESSOFTHEHEARTTOEXECUTENORMAL HEARTACTIVITY
!BNORMALITY IS !TRIAL&IBRILLATION [CFCF]
!BNORMALITY IS 3UPRAVENTRICULAR
!RRHYTHMIA[CFCF]
!BNORMALIT Y I S 6ENTRICULAR 4ACHY!RRHYTHMIA[CFCF]
!BNORMALITYIS
-YOCARDIAL)NFARCTION [CFCF]
2ULEJK)& ;FEATURESJ= IS ; VALUE WITHIN RANGEOFVALUEOFK=
4(%.!BNORMALITY IS .ORMAL [CF CFJK]
!BNORMALITY IS !TRIAL&IBRILLATION [CFCFJK]
!BNORMALITY IS 3UPRAVENTRICULAR
!RRHYTHMIA[CFCFJK]
!BNORMALIT Y I S 6ENTRICULAR 4ACHY!RRHYTHMIA[CFCFJK]
!BNORMALITYIS
-YOCARDIAL)NFARCTION [CFCFJK]
cfji1
T H E C O N F I D E N C E I N CONSEQUENCEI ESTABLISHED BY2ULEJ K
cfim and cfim1 ABSOLUTE MAGNITUDES OFcfim ANDcfim1RESPECTIVELY
Q NUMBEROFPOSSIBLECONDITION FORFEATUREJ
Pjki PROBABILITYOFABNORMALITYI WHENFEATUREJHASAVALUEOF K
4O CALCULATE THE COMBINED CERTAINTY FACTOR FOR CONSEQUENCESITOWHICH IS THE CLASSIlED ABNORMALITIESAMODIlEDVERSIONTHEFOLLOWINGEQUATIONWASUSED.EGNEVITSKY
$ATA
3OURCE !BNORMALITY (EIGHT J
0ULSE
J 7IDTH )NT %ULER /RI
%ULER 3OBEL
%ULER
!LL
%ULER 3OBEL
!LL
#& √ /
×
$ATATO
$EVELOP THE 3YSTEM
.ORMAL √
!TRIAL&IBRILLATION √
-YOCARDIAL
)NFARCTION √
3UPRAVENTRICULAR
!RRHYTHMIA √
6ENTRICULAR
4ACHYARRHYTHMIA √
4ESTED
$ATA
.ORMAL √
!TRIAL&IBRILLATION √
-YOCARDIAL
)NFARCTION √
3UPRAVENTRICULAR
!RRHYTHMIA √
6ENTRICULAR
4ACHYARRHYTHMIA √
4!",%#LASSIlCATION4EST2ESULTS
2%&%2%.#%3
#ROWE *! 4HE WAVELET TRANSFORM AND ITS APPLICATIONTOBIOMEDICALSIGNALS4IME&REQUENCY
!NALYSISOF"IOMEDICAL3IGNALS)%%#OLLOQUIUMON 9EAR$IGEST.O *AN
$RIPPS*(!NINTRODUCTIONTOTIMEFREQUENCY METHODS APPLIED TO BIOMEDICAL SIGNALS 4IME
&REQUENCY !NALYSIS OF "IOMEDICAL 3IGNALS )%%
#OLLOQUIUMON9EAR$IGEST.O *AN
%NDERLE*3LANCHARD3"RONZIO*)NTRODUCTION TO"IOMEDICAL%NGINEERING3AN$IEGO!CADEMIC 0RESS
&ISHER"0ERKINS37ALKER!7OLFART%3PATIAL lLTERS MEDIAN lLTER ONLINE HTTPWWWCEE HWACUKHIPRHTMLMEDIANHTML!PRIL 'ONZALES 2# 7OODS 2% $IGITAL IMAGE
PROCESSINGND%D.EW*ERSEY0RENTICE(ALL (AGHIGHI-OOD!4ORRY*.4IME&REQUENCY
!NALYSIS OF 3YSTOLIC -URMURS 4IME&REQUENCY
!NALYSISOF"IOMEDICAL3IGNALS)%%#OLLOQUIUMON 9EAR$IGEST.O *AN (LAWATSCH&"OUDREAUX"ARTELS',INEARAND
QUADRATICTIMEFREQUENCYSIGNALREPRESENTATIONS )%%%3IGNAL0ROCESS-AG
)KEDA + 6AUGHN " 1UINT 3 7AVELET
DECOMPOSITIONOFHEARTPERIODDATA0ROCEEDING OF4HE&IRST*OINT"-%3%-"3#ONFERENCE3ERVING (UMANITY !DVANCING4ECHNOLOGY /CT
-ITRA3+$IGITALSIGNALPROCESSINGACOMPUTER BASEDAPPROACH3INGAPORE-C'RAW(ILL .EGNEVITSKY - !RTIlCIAL INTELLIGENCE A GUIDE
TOINTELLIGENTSYSTEMS%SSEX0EARSON%DUCATION ,TD
3MITH374HE3CIENTISTAND%NGINEERS'UIDE TO$IGITAL3IGNAL0ROCESSINGND%D3AN$IEGO
#ALIFORNIA4ECHNICAL0UBLISHING
3PRENGER3TEPHEN-0ITCH3CALINGUSINGTHE&OURIER 4RANSFORM 4HE$30$IMENSION ONLINE HTTP WWWDSPDIMENSIONCOM!PRIL
6ERHELST7 2OELANDS - !N OVERLAPADD TECHNIQUE BASED ON WAVEFORM SIMILARITY 73/,! FORHIGHQUALITYTIMESCALEMODIlCATION OFSPEECH0ROC)%%%)#!330!PRPP
7ONG + 4HE 2OLE OF THE &OURIER4RANSFORM IN 4IME3CALE -ODIFICATION *OURNAL OF 5NDERGRADUATE2ESEARCH ONLINE HTTP WWWCLASUmEDUJURPAPERSPAPER?
WONGHTML!PRIL
cf cf
P
n k q
i i
jk i j j n
s
¤
1 , 1 2, ,.., WHERE
cfji
T H E C O N F I D E N C E I N CONSEQUENCEI ESTABLISHED BY2ULEJK
2%35,43!.$$)3#533)/.
4HERESULTSFORTHETESTEDCLASSIlCATIONARESHOWN IN4ABLE
4HElRSTlVEROWSSHOWTHEPARAMETERSANDCF OFTHECLASSIlEDABNORMALITYFORTHEDATAWHICH WASUSEDTODEVELOPTHEKNOWLEDGEBASEWHERE AS THE LAST FIVE ROWS REPRESENT THE CLASSIFIED ABNORMALITYFORTHEDATAWHICHWASNOTUSEDTO DEVELOPTHEKNOWLEDGEBASEIEEXTERNALDATA SET)NBOTHSETSOFDATATHESYSTEMWASABLETO CORRECTLYCLASSIFYTHEINVESTIGATEDABNORMALITIES WITH ACCURACY4HE CF FOR THE CLASSIFIED
ABNORMALITY VARIES BUT THE VALUES WERE EITHER CLOSETOOREQUALTOWHICHINDICATESSTRONG BELIEFINTHERESULT
#/.#,53)/.
4HE DETECTION RESULTS SHOWED THAT THE SYSTEM FUNCTIONSVERYWELLANDGIVESVERYGOODDETECTION RESULTACCURACY )TISTHEREFORECONCLUDED THAT THAT THE OBJECTIVE OF THIS RESEARCH HAS BEEN ACHIEVED &UTURE ENHANCEMENT IN THIS RESEARCHINCLUDESTHEINCLUSIONOFMOREDATAIN THE KNOWLEDGE BASE AND CF EXTRACTION &URTHER TESTINGONLARGERTESTSETISALSOPLANNEDINTHE NEARFUTURETOTESTTHEROBUSTNESSANDACCURACY OFTHISSYSTEM
!#+./7,%$'%-%.4
4HIS WORK WAS SUPPORTED BY THE -INISTRY OF 3CIENCE4ECHNOLOGYAND%NVIRONMENT-ALAYSIA UNDERTHETH-ALAYSIA0LANS)20!
32'RANT