1.INTRODUCTION
Severalstudieshavefoundthat,forultrasoundthere arenumerousfactorsthatinfluencebirthweight,including measures.thatandthereporttheperformed.absencehealthpersonbeforemodelandmacrosomia,diseases:alsoelement.elementimportantnetworkUsinggroups:manyThenoftraining.separatetraditionaltoweightrelationshipsformuladiabetinfantgender,headcircumference,maternalage,height,andes.ItisdifficultwithasimpletraditionalwithdrawaltoaddressmultidimensionalandnonlinearbetweenallofthesevariablesandfetalRecently,aneuralimplantnetwork(ANN)wasusedpredictfetalweighttoovercometheproblemsofretinaldetachment.Here,wehaveproposedaandpredictablebirthcontrolmodelbasedonwecollectedanestimateddatafromthedefaultsizetheembryo'sheadusing2Dultrasoundimages|Zenodo.weuseCNNandMulticlassSVMasalgorithmswherealgorithmsareusedtodividetheembryointotwoBW<4000g(LBW,NBW)andWW>4000g(HBW)theGLCMfeatureextractandtheneuralconvolutionisusedasacomputerview.,CNNplaysanroleandisoftenused.ItpullsoutanimageandconvertsittoalowresolutionwithoutlosingitsAtthesametime,theclassificationofdiseasesisbeingdone.inparticularweareconsideringfourmainrespiratorydisease(RDS),fainting,fetusheartfailuresoimageimagingwillbetrainedclassifiedforaspecificimagingdisorder.Asaresultourisusefulwhichhelpsgynecologisttoidentifytheriskitistoobadandthiscanbeusedwithanonmedical(patients)sopregnantmotherscanknowabouttheirandthehealthofthebabyandthishelpsthemintheoftheirgynecologist.onlyultrasoundtestswereSometimesaftertheultrasoundisdonebutforaveragepersontoknowthedetailsintheultrasoundiscomplicated,doctorsmaynotbeabletodiagnoseemergencyinthatcasepatientscanuseourmodelfindoutthehealthstatusofthebaby.Wethereforehopethishasasignificantimpactonwomen'ssafety
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 08 Issue: 12 | Dec 2021 www.irjet.net p ISSN: 2395 0072 © 2021, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page308
Abstract - Women's safety is important in society. So here's a model that will assist them during their pregnancy. Uneducated women have a lower level of pregnancy health awareness. Knowing the fetus weight at correct time of pregnancy is much important.According to the World Health Organization(WHO), therangeof Low Birth Weight (LBW)is less than 2500g, the range of High Birth Weight (HBW) is greater than 4000g, and the range of Normal Weight is between2500gand4000g.Accordingtoweight of fetusitmay undergo the disease of which some may be life taking ones. Hence to find out these traumas as soon as possible, we are building a Machine Learning model where in we use two algorithms as methods one is Convolution Neural Network(CNN)andMulticlass SVMalgorithm, whichpredicts the estimated weight and disease if suffering according. This helps clinicians to identify risks in fetus whose effectis usually to carring mother. As a result we found that multiclass SVM outperforms the best compared to CNN with approximate accuracy with 95%.
An Approach for Fetal Weight Estimation using Machine Learning for Women Safety Dr.Suvarna Nandyal1 , Sahana Khened2
1Professor at Department of Computer Science, PDA College of Engineering, Karnataka, Gulbarga, India
KeyWords: Women Safety, Machine Learning, Multiclass SVM, CNN, Fetal weight, Disease Classification.
2Student at Department of Computer Science, PDA College of Engineering, Karnataka, Gulbarga, India ***
Themostcommonwaytomeasureababy'sweighttogetan ultrasoundisdone,becauseitissafe,notdangerous.Earlier introduced various relay formulas based on different ultrasound parameters. Every setback formula has a problem,itgoestherewithnormalbirthweight,butitmay notbeasaccurateaswhenthebaby'sweightvaries.
Knowingababy'sweightiscriticalforpredictingbothshort andlong termhealthoutcomes.WHOstatesthatweightsare inthreedifferentcategories:LBWdistances(<2500),NBW distances(2500gto4000g)andHBWdistances(>4000g). Tokeepthisinmind,lowbirthweightislinkedtoshort term andchronicillnessessuchasRispiratoryDistressSyndrome (RDS) and mental retardation, learning disabilities can be considered as long term disabilities, and much more.The termHBWfetusmacrosomiareferstoanewbornbabywho is overweight. As described in the HBW range> 4000g, approximately9%ofchildrenworldwideweighmorethan 4000g and the risk factor associated with macrosomia increaseswhentheweightexceeds4500g.Sincethebabyis at risk of injury such as heart failure, there are also long termsideeffectssuchaslowbloodsugarlevels,childhood obesity.Manyriskfactorsthatmayincreasetheriskoffetal macrosomia some can be controlled and some may not. Quoting maternal diabetes mellitus, History of fetus macrosomia, maternal obesity, obesity during pregnancy, havingababy,andmuchmore.Otherrisksformothersare laborproblems,genitalpain,heavybleedingafterchildbirth, ruptureoftheuterus.
AshleyI.Naimiet.al[1](2018),Inthisstudyweobservethey have used quantile regression, random forests, Bayesian additiveregression,whicharesimpleregressionformulas thatmighthavelessaccuracycomparedtoourmodel.
Seizures:Therearenumerouscausesofneonatalseizures, including the conditions listed above. Infant seizures are morecommoninbabieswhowerebornprematurelyorwith alowbirthweight.oxygendeficiencyjustbeforeorduring birth,causedby:Difficultlabour,Umbilicalcordproblems, Placentainjury,isanothercauseofseizuresthatmayormay notindicateabirthinjury.
Macrosomia: A newborn with "foetal macrosomia" is one who is significantly larger than average. Regardless of gestationalage,ababyisdiagnosedwithfoetalmacrosomia ifheorsheweighsmorethan8pounds,13ounces(4,000 grammes).Around9%ofbabiesworldwideweighmorethan
al[3](2014),Inthispaper,theyhaveproposed amethodbasedonthesupportvectorregressionwhichwas onlyproposedtoimprovetheweightaccuracybelow2500g theLBWfetus. So far we discussed the literature, from which we can conclude saying, In previous work they have most of the timesusedthesimpleregressionformula suchassupport vector regression, random forest, XG Boost logistic thoughjustoneregressionanalysis,bayesianadditiveregressionwhichhaveortheotherdrawbacktoquotesometheyproposedtoimprovetheLBWfetus,somehaveonlymadeanalysis,theANNalgorithmisusedtoovercometheseaspects itsapproach
Maznah Dahlui et. al[4] (2016), In this study, we observe how LWB continues to be the primary cause of infant morbidityandmortality.Inthisresearchalsotheyhaveused the logistic regression analysis which is usually used for mortalityit'siseffectANN.considered,fromisimprovementnotwhichuseofregressionsurveyingthefactors.whenresearchedwithmultiplelogisticthesignificantoddratioswereshownformothersthenorthwestregion.YuLuet.al[7](2019),hereinthisstudy,againwefoundtheofrandomforest,XGBoost,andlightGBMalgorithmsarethesimpleregressionformulaswhichmightbethataccurateasperthereportthereisa12%anddecreasederrorby3%.YuehChinChenget.al[8](2012),Inthisresearch,themodelbuiltontheparametersofultrasoundwhichmayvaryfetustofetusastheyarenotthatreliabletobetheyalsotrainingthedatarandomlyusingtheSareerBadshahet.al[6](2008),ThisstudyrevealedtheofLBWinafetus.Acrosstheworld,neonatalmortality20timesmorelikelyforLBW,andasaresultofresearch,shownthatbabiesareatincreasedriskofperinatalandmorbidity. Yuet. didn’tmeettheexpectations.Sotoovercomeall theseaspects,wearebuildinganovelmodelwhichworkson ConvolutionNetworks(CNN)andMulticlassSVM’sasthey areverygoodwhenwehavenoideaonthedata.
Herewehavemadesomeresearchesrelatedtoourstudy, Miao Feng et. al[5] (2019), In this paper the authors have used the traditional techniques of machine learning, the SMOTE and DBN technique were used which are almost limitedonlyforthenormalweights.ThisSMOTEalgorithm lacks the variance and generalization which doesn’t make themodelaccurateone'sastheyaregeneralizingtheLow birthweightwithNormalweightfetus.
Jinhua
1.2 Disease of the fetus Hereweareconsideringfourmajordiseasesorriskwhich are to be diagnosed at the earliest of which the fetus may suffer from are CardiacAnomolies, RDS, Seizures, CardiacMacrosomia.Anomalies:Commonsymptomsinclude: Newborns may have problems with feeding, growth, and birth weight. Heart palpitations,Cyanosis is a condition in which there is a buildup of skin cells (bluish skin, lips, or nails)Ontheinternetclubbing(changesinnails)Difficulties breathingSwellingofatissueororganExertioncausesyou totyretoo Respiratoryquickly.DistressSyndrome:RDSstandsforrespiratory distresssyndrome. TheRDSoccurswhenthelungshavenotyetfullydeveloped and are unable to provide enough oxygen, resulting in breathingdifficulties.Diagnosisenablesdoctorstoprovide necessary treatment prior to delivery by administering certainsteroidstothemother, whicheventuallyhelpsthe foetusdeveloptheirlungsandgainsomeenergysothatthe foetusthisisusuallydonebefore2to3days.
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 08 Issue: 12 | Dec 2021 www.irjet.net p ISSN: 2395 0072 © 2021, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page309 1.1 RELATED WORK
JiaZhenget.al[2](2017),Inthisresearch,wemayknowthe risk of macrosomia also known as HBW fetus. As per research,theimportanceofmacrosomiarangedfrom5%to 20% has increased y 15 20% by last three decades. The maternal complication of the fetus with macrosomia risk may involve complexities such as operative vaginal delivery,c section emergency, perineal laceration are increasedby4.5times,birthinjury,cardiacanomalies,and many more. Infants which are weighing more than 4500g carrytheriskofshoulderdystocia,Toquotesomelong term disorders, they likely to develop obesity, type 2 diabetes. Thusthemacrosomiaisidentifiedasaworldwideproblem.
The proposedsystemhere employs machinelearningand imageprocessingtechniquestoestimatefoetalweightand disease, if any, if it is overweight or underweight in conditions.Identifyingthefoetalriskleveliscritical,asthey can become chronic for both mothers and foetuses. The primarygoalofthisstudyistoproposeamachinelearning solutiontoimprovetheaccuracyoffoetalweightestimation and to help clinicians identify potential risks before delivery.Ourmodel
isuser friendlybecauseitcanbeusedby both clinicians and non clinicians (patients/non medical people). They can use this after getting their ultrasound done,andbecausethisisprimaryinformation,theycantake precautions accordingly in the absence of their specific gynaecologist. Fig 1 ProposedMethodology outputclassifiedusinisisextractionofinformativemayabouttheweightofthatofvisiblearesomewithitrainfeatureimageconvertedcontainsubfolderscontainalgorithms.theSo,wearehereusingtheConvolutionNeuralNetworkoneofdeeplearningmodels,andtheMulticlassSVMasInwhichweconsiderabalanceddatasetwhichappxof3352imagedatasetswhichwehavemadeofHBW,LBW,NORMALwhereeachfolderthesetofultrasoundimagedatasetswhichwillbetoRGBtoGRAYscaleimagesafterwhichthefeatureextractionisdonewhichisdoneusingglcmextractionwithmulticlassSVMalgorithm,finallyweandclassifythem.Whenaninputisgiven,foreverynputitestimatestheweightofthefetusanddisease.Alongthiswehavemaderealclinicalresearchandcollectedsampleultrasoundreportsofpregnantwomenwhointhe26thweakoftheirpregnancyasthefetusstartinthe26thweak,weareconsideringBPD,HC,AC,FLthefetusasthesearethejudgementalfeaturesofafetusisthefetusishealthyornot?wearegivingthereadingstheseparticulardataforinputimagesconcerningtheofthefetuswhicheventuallydependsontheHCindatasets.Thishelpsthemothertobeknowledgeableherfetusinabsenceofhergynecologist.Theresultsvary+10%to10%withgivenoutputastheyarevaluesonly.Themodel'simplementationbeginswiththeloadingbalanceddatasetsintothesystem,followedbytheoffeaturesusingglcmfeatureextraction.Thedatathenstoredinthetrainmatfile,whereeachfeaturevaluestored,andweusethesetrainmatfilestotrainourmodelgmulticlasssvmandcnnalgorithms.Theoutputisaftertheinputhasbeentrainedandgives.Wefinallygettheaccuracyofbothalgorithmsin
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 08 Issue: 12 | Dec 2021 www.irjet.net p ISSN: 2395 0072 © 2021, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page310 8pounds,13ounces.Whenababy'sbirthweightexceedsa certainthreshold,thechancesoffoetalmacrosomiaincrease dramatically. i.e. more than 9 pounds 15 ounces (4,500 grams) Fetal macrosomia can complicate vaginal delivery and put the baby at risk of injury during birth. Fetal macrosomiaalsoincreasesthebaby'sriskofpostnatalhealth problems. Foetal macrosomia can be difficult to detect duringpregnancy.
2. DATA COLLECTION Inanystudythemainfactorisdatacollection.Hereinour modelwehavecollectedadatabaseofultrasoundimagesof theembryofromthedefaultsizeoftheembryo'sheadusing 2D ultrasound images | Zenodo. The database contains appx.3352imagesofwhich90%areundertrainingand10% for test. During pregnancy, ultrasound imaging is used to measurefetalbiometric.Oneofthesemeasurementsisfetal headcircumference(HC).HCcanbeusedtomeasuretheage ofpregnancyandtomonitorfetalgrowth.HCismeasuredat acertainpointinthecrossofthefetalhead,whichiscalled thenormalplane.AccordingtoaHCstudyatweek26(mid term)ofpregnancyvaluessuchas,ifthefetusHCis<18.0cm consideredtobeunderweight,>24.0cmisoverweightand between18.0to24.0cmisconsiderednormal(valuescan vary + 10% to 10%) in this analysis, considering training_set.csv, we have divided HC values into three different folders. similarly by weight we have five folders diseaseforwith(cardiacanomolies,RDS,sezieures,Macrosomia,Nodisease)adiseasedatasetinbetween.Nowourdatasetisreadytraininganddistinguishestheeffectsofweightandonaparticulartestimage.
3. PROPOSED METHDOLOGY
tobeagooddiscriminator Asaresult,thesearchforthebest imagequalitymetricisongoing.Afewexamplesofcommon statistics applied to co occurrence probabilities are discussedahead.
6. Concentrate the GLCM highlights for picture GLCM components such asEnergy, Entropy, Contrast, Variance, Homogeneity, Correlation, Sum Average, Sum Entropy, Sum Variance, Difference Variance, Difference MaximumEntropy,CorrelationCoefficient,InformationMeasuresof Correlation1andInformationMeasuresofCorrelation2are separated
1)Energy: This metric is also known as consistency or precisesecond.Itassessesthetexturalconsistencyofpixel pair redundancy, recognizes surface peculiarities. Energy canjusthaveagreatestworthofone.Atthepointwhenthe dimleveldispersionhasasteadyorintermittentstructure, highenergyesteemshappen.Thereisastandardizedreach forenergy.TheGLCMofthelesshomogeneouspicturewill beenormousoflittlesections.
GLCMisanothernameforGrayLevelDependency Matrix.Itisdefinedas"atwo dimensionalhistogramofgrey levels for two pixels separated by a fixed spatial relationship."Whenstudyingdifferentimages,GLCMproves
6)Correlation:diminishes.
3)Difference: This measurement, which is the distinction snapshot of GLCM, measures the spatial recurrence of a picture. It is the distinction between an adjacent arrangementofpixels'mostelevatedandleastqualities.It talliesthequantityofnearbyvarietiesinthepicture.Alow difference picture has low spatial frequencies and presentations the GLCM focus term around the primary 4)Variance:askew.Thismeasurementestimatesheterogeneityand is exceptionally corresponded with first request factual factorslikestandarddeviation.Atthepointwhenthedark level qualities stray from their mean, the fluctuation 5)Homogeneity:increments.
A. Classification Therearemanyclassificationtechniques.Hereweusethe Multiclass SVM and Convolution Neural Network as classifiersindividually.Themainagendahereistocomeup with better classification techniques when compared to other simple regression techniques like random forests, linear regression, logistic regression analysis which have beenusedinpreviouswork StepsforMultiClassSVMworking
1. Picture Acquisition: First we need to choose the fetal weightwhichisinfluenced bythe infectionandafterward gatherthefetalultrasoundpictureandburdenthepicture intotheframework.
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 08 Issue: 12 | Dec 2021 www.irjet.net p ISSN: 2395 0072 © 2021, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page311
2)Entropy: This metric surveys a picture's problem or intricacy. At the point when the picture isn't texturally uniform, the entropy is high, and numerous GLCM components have tiny qualities. Complex surfaces have a high entropy. Entropy has a solid however converse relationshipwithenergy.
2. Division: It implies portrayal of the picture in more significant and simple to dissect way. In division an advanced picture is divided into various fragments can characterizedassuper pixels.
Themeansandstandarddeviationsofgxand gyThe relationship highlight estimates the picture's dark tonestraightconditions.
4. Difference Enhancement: Thefirstpictureisthepicture giventotheframeworkandtheyieldoftheframeworkafter contrast upgrade is Enhanced Image, this is the picture subsequenttoeliminatingthesharpedges.
This measurement is likewise called as OppositeContrastSecond.Itestimatespicturehomogeneity as it accepts bigger qualities for more modest dim tone contrasts in pair components. It is more delicate to the presenceofclosetoslantingcomponentsintheGLCM.Ithas mostextremeworthwhenallcomponentsinthepictureare same. In terms of identical dispersion in the pixel sets populace, GLCM difference and homogeneity are unequivocally yet conversely related. It implies that if contrastincrementswhileenergystayssteady,homogeneity
3. Differentiation: Picturepixelesteemsareconcentrated almostalimitedreach.
5. Changing RGB over to HSI: TheRGBpictureisinthesize of M by N by 3,where the three measurements represent three picture planes(red, green, blue).if every one of the threepartsareequivalentthenchangeisvague.Byandlarge thepixelscopeofRGBis[0,255]inhisthepixelrangeis[0, 1]. Transformation of pixel reach should be possible by computingoftheparts;Hue,Saturation,Intensity.
terms of percentage by which we can know which outperformsthebest. Finally,wehavetheaccuracyofbothalgorithmsin percentageterms,allowingustodeterminewhichalgorithm outperforms the other. One of the most important characteristics used in identifying objects or regions of interest in an image is texture. The texture contains vital information about the surface's structural arrangement. Texturalfeaturesbasedongray tonespatialdependencies havebroadapplicabilityinimageclassification.
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 08 Issue: 12 | Dec 2021 www.irjet.net p ISSN: 2395 0072 © 2021, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page312 7. Train the Dataset features Model=Svmtrain(glcmfeatures,label); 8 Classify the input image Classifyoutput=Predict(Model,testimagefeature) Generally,thepixelrangeofRGBis[0,255]inhisthepixel range is [0, 1]. Conversion of pixel range can be done by calculatingthecomponents;Hue,Saturation,Intensity. Steps for CNN working 1. Load the dataset 2. Train the Model by adding following layers classificationsoftmaxfullyrelubatch=imageInputLayer(imageSize,'Name','input')whereimageSize[64643];convolutional2dLayer(3,8,'Paddding','saame')NormalizationLayerLayerConnectedLayer(3)LayerLayer]; 3. Build aCNN Net Model net= 'Plots','train'Verbose',true,'InitialLearnRate',1e'MaxEpochs',100,...wheretrainNetwork(datastore,layers,options);options=trainingOptions('sgdm',...4,......ingprogress'); 4. Classify the Input Image YPred=classify(net,imdsTest_rsz); Where,imdsTest_rszisatestimage Fig-2. LayersofCNN 4. RESULTS AND DISCUSSION To compare the work to previous defined methodologies, previously the work was only based on traditional techniques such as the SMOTE algorithm, which lacks variance and generalisation, and some used simple regressionformulas(Quantileregression,Randomforest), whichdonotprovideaccuratemodelresults.Insomecases, thegoalwassimplytoimproveweightaccuracy.Asaresult, justifying our model is that it will overcome all of the shortcomingsofpreviousworkbyanalysingtheweightof the foetus and disease accordingly, allowing patients and doctors to diagnose as early as possible in order to avoid Somerisks.ofthekeyissuesaddressedare Ourmodelpredictsbothfetalweightanddiseaseseverityin relationtoweight. Thismodel'suserinterfacegivestheuseranideaoffoetus parameterssuchasheadcircumference,foetallength,BPD, and abdominal circumference. In the absence of a gynaecologist, these parameters assist patients in an emergency.Thisgoesbeyondbasic regressionformulasand employs the most recent machine learning methodology, GLCM FeatureExtraction. Becausethedatasetismuchclearerforalgorithmicaction, themodel'saccuracyisaround95%. Userfriendlymodel. Becauseofthismodel,thecaregivercanbemindfulofboth herbabyandherhealthmetrics. In the method of tracking down the best characterization methodologyandhighlightextractiondependentonpicture descriptorsandAIstrategiesthemulticlasssvmandcnnasa calculation the approx of 3352 pictures dataset are been considereddependentonthedataset90oftheinformation has been considered for preparing and 10 for testing the datasetpicturesdependontheirhcvalues.
Fig. 3.TrainiginSVM
From,Figure3,wetrainthedataset,each picture's highlights are been extricated utilizing the GLCM method which is removed and put away in Train.mat which is additionallyutilizedforcharacterization.
Fig.4 Outputof MultiClassSVM
Fig 5.TraininginCNN Fig 6.AnalysisresultofCNN Fig 7. OutputofCNN As shown in Figures 5,6,7, we first trained the datasets; featuresareextractedbylayers,i.e.pixelbypixel,andstored intheTrain.matfile,whichisthenusedforclassification.The CNNanalyzerdisplaystheinternalalgorithmicstructureof the model, allowing us to understand the steps of the algorithm.Women'ssafety,aspreviouslystated,iscriticalin terms of medical analysis. The goal of this project is to proposeamachinelearningsolutiontoimprovetheaccuracy of fetal weight estimation and to assist clinicians in identifyingpotentialrisksbeforedelivery.Followingthese techniques,weclassifiedthefetus'sweightasOverweight, UnderweightandNormal,aswellasthediseasesmentioned above. Graphical Analysis of Accuracy for Fetal Health
Fig 8.InputImage
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 08 Issue: 12 | Dec 2021 www.irjet.net p ISSN: 2395 0072 © 2021, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page313
5. CONCLUSION AND FUTURE SCOPE classificationmethodhasbeendevelopedforestimating fetalweightanddiseaseiffetusesareaffected.InSVMs,the GLCMtechniqueisusedtoextractimage InCNN, layersofimagesareconsideredandtakenintoaccount.
Neuralcompared6.ToovercomethiswehavetheConvolutionNetworkswith
we have used latest algorithms
have
friendly
4.Inthismodelaccordingto the weight of fetus model gives an average information on HC,BPD,FL,AC of fetus which helps to know the fetalhealthconditions.
5.We don’t oftenly find a model which have both fetal relationweightanddiseaseseverityintoweight.
4.There was no information on fetal (HC,BPD,FL,BPD).parameters
A
Table 1:Atableofcomparisonwithpaststudies.
and Convolution Neyral Networks. 2.They
1.Inpreviouswork,they come regrsession formula I.e,ANN,Logistic this model I.e,Multiclass SVM were not that user model. 2.We here came up with a userfriendlymodel where apatientcanalsomake gynecologist.
3.Lessaccurateresults 3. We have achived the accuracy around 95% in MulticlassSVMalgorithm.
For efficientbelievetheandtheaccuracyoutperformingSVMclassification,theproposedmethodemploysbothMulticlassandCNNalgorithms,withMulticlassSVMtheothers.Asaresult,MulticlassSVMrangesbetween93and94percent,dependingoninputimage,whereasCNNaccuracyrangesbetween6870percent,dependingontheimageinputdata.Despitefactthatthesystem'sperformanceisadequate,wethatitcanbeimprovedfurther.Incorporatingfeaturescanhelptoimprovetheoverallquality.
6.TherestudiesmadeonCNN didn’tmeetthegoodpercent ofaccuracy. Multiclass SVM’s,we found overMulticlassoutperformswelltheNeuralnetworks.
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 08 Issue: 12 | Dec 2021 www.irjet.net p ISSN: 2395 0072 © 2021, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page314 Fig 9.AccuracyinConvolutionNeuralNetwork Fig 10.AccuracyinMulticlassSVM Accuracy achieved with the iterations for both the algorithmshasbeen showedabove UsingBothAlgorithms >CalculatetheaccuracywithttheTrainDataandtestData usingformulaasshownbelow >ByMulticlassSVM accuracy= (Correct(TestDataOutput)/ Total TrainDataoutput)*100>ByConvolutionNeuralNetwork accuracy= (Correct(EachDataOutput)/ TotalTrainDataoutput)*100>Graphhasbeenplottedwiththexyplot >Where x axis represnets the iteration count and y axis reprsensttheaccuracyvaluegenretaed. AswiththeaccuracygrpahwecansaythatmulticlassSVM hasbetteraccuracythanCNN. Previous Results Present Results
5.Butherewehavemadea way where both fetal weigth weight and disease severity in relation to weight.
use of this in absence of their
features.
Regression. 1.In
up with simple
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 08 Issue: 12 | Dec 2021 www.irjet.net p ISSN: 2395 0072 © 2021, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page315 Sofar,wehaveusedtheNeuralNetworkconcepts andsupportvectorshereinfutureitcanbebuiltonlatest machine learning algorithms in mixture of Artificial intelligenceandthedatacollectioncanbestillmoreaccurate sothatthemodelsaccuracygetshiced. [1]REFERENCESAshleyI.Naimi , Robert W. Platt,and Jacob C. Larkin, Department of Epidemiology, University of Pittsburgh, Health,DepartmentofEpidemiology,Biostatistics,andOccupationalMcGillUniversityand Magee Women’s Research Institute, University of Pittsburg Epidemiology.Machine LearningforFetalGrowthPrediction. 2018March;29(2): 290 298.doi:10.1097/EDE.0000000000000788. [2]Jia Zheng, Xin Hua Xiao, Qian Zhang, Li Li Mao, Miao Yu, Jian Ping Xu and Tong Wang,Department of ChinaPekingResearchHealthEndocrinology,KeyLaboratoryofEndocrinology,Ministryof,PekingUnionMedicalCollegeHospital,DiabetesCenterofChineseAcademyofMedicalSciences&UnionMedicalCollege,Beijing100730, Correlation of placental microbiota with fetal macrosomia and clinical characteristics in mothers and newborns.Oncotarget, 2017, Vol. 8, (No. 47), pp: 82314 82325. [3]JiaZheng,Xin HuaXiao,QianZhang,Li LiMao,MiaoYu, Jian Ping Xu, and Tong Wang collaborated on this study. Fetal Weight Estimation Using the Evolutionary Fuzzy Support Vector Regression for Low Birth Weight Fetuses, Jinhua Yu, Yuanyuan Wang, and Ping Chen, IEEE Transactions on Information Technology in Biomedicine neural[8]g_Nodule_Detection_based_on_Faster_Rhttps://www.researchgate.net/publication/347697194_LunEstimationWong,[7]at:ondoi:10.1186/1471BMCbirthweightPayne,[6]2019.publicationWeightXIAORONG[5]MIAOweightDemographicNorlaili[4]pp.(Volume:13,Issue:1,Jan.2009).Oncotarget,Vol.8,(No.47),8231482325,2017.MaznahDahlu,NazarAzahar,OcheMansurOche,andAbdulAziz.Evidencefromthe2013NigeriaandHealthSurveyonriskfactorsforlowbirthinNigeria.Citation:GlobHealthAction2016,9:28822http://dx.doi.org/10.3402/gha.v9.28822FENG,LIWAN,ZHILI,LINBOQING,ANDQIUsingMachineLearningtoEstimateFetalviaUltrasound.AcceptedonJune22,2019,dateofJuly2,2019,dateofcurrentversionJuly18,SareerBadshah,LindaMason,KennethMcKelvie,RogerandPauloJGLisboaidentifiedriskfactorsforlowinpublichospitalsinPeshawar,NWFPPakistan.PublicHealth2008,8:197(June4,2008).24588197.14thofJune,2007AcceptedJune4,2008.Thisarticlecanbefoundhttp://www.biomedcentral.com/14712458/8/197.YuLu,XiZhang,XianghuaFu,FangxiongChen,KelvinK.L.,EnsembleMachineLearningforFetalWeightatDifferentGestationalAges.CNN_FrameworkEfficientfetalsizeclassificationcombinedwithartificialnetworkforestimationoffetalweightYuehChin Chenga, Gwo Lang Yanb, Yu Hsien Chiuc, Fong Ming Changd,Chiung HsinChangd,*,Kao ChiChung,Accepted12 July 2012. Efficient fetal size classification combined with artificial neural network for estimation of fetal weight COREReader [9]Low Birth Weight Infants | Reiter & Walsh [10](abclawcenters.com)Newbornrespiratory distress syndrome NHS [11](www.nhs.uk)Automated measurement of fetal head circumference using2Dultrasoundimages|Zenodo [12]Fetalmacrosomia Symptomsandcauses MayoClinic [13]InfantSeizures|BirthInjuryGuide