train.jsonl

{"id": "2d87191d9618-0", "text": "ResearchArticle\nChagas Parasite Detection in Blood Images Using AdaBoost\nV\u00edctor Uc-Cetina,1Carlos Brito-Loeza,1and Hugo Ruiz-Pi\u00f1a2\n1FacultaddeMatem \u00b4aticas,UniversidadAut \u00b4onomadeYucat \u00b4an,AnilloPerif \u00b4ericoNorte,TablajeCatastral,13615M \u00b4erida,YUC,Mexico\n2CentrodeInvestigacionesRegionalesDr.HideyoNoguchi,UniversidadAut \u00b4onomadeYucat \u00b4an,Avenida,Itz \u00b4aesNo.490x59,\nColoniaCentro,97000M \u00b4erida,YUC,Mexico\nCorrespondenceshouldbeaddressedtoV \u00b4\u0131ctorUc-Cetina;uccetina@uady.mx\nReceived17October2014;Revised20February2015;Accepted20February2015\nAcademicEditor:IriniDoytchinova\nCopyright\u00a92015V \u00b4\u0131ctorUc-Cetinaetal. This is an open access article distributed under the Creative Commons Attribution\nLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly\ncited.\nTheChagasdiseaseisapotentiallylife-threateningillnesscausedbytheprotozoanparasite, Trypanosoma cruzi. Visualdetectionof\nsuchparasitethroughmicroscopicinspectionisatediousandtime-consumingtask.Inthispaper,weprovideanAdaBoostlearning", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-1", "text": "solution to the task of Chagas parasite detection in blood images. We give details of the algorithm and our experimental setup.With this method, we get 100% and 93.25% of sensitivity and specificity, respectively. A ROC comparison with the method mostcommonlyusedforthedetectionofmalariaparasitesbasedonsupportvectormachines(SVM)isalsoprovided.Ourexperimental\nwork shows mainly two things: (1) Chagas parasites can be detected automatically using machine learning methods with high\naccuracy and (2) AdaBoost + SVM provides better overall detection performance than AdaBoost or SVMs alone. Such resultsare the best ones known so far for the problem of automatic detection of Chagas parasites through the use of machine learning,\ncomputervision,andimageprocessingmethods.\n1. Introduction\nThe Chagas disease, also known as American trypanosomi-\nasis, is a potentially life-threatening illness caused by theprotozoan parasite, Trypanosoma cruzi (T. cruzi).According\nto the World Health Organization [ 1], it is found mainly in\nLatin America, where it is mostly transmitted to humans by\nthefaecesoftriatominebugs.Morethan25millionpeopleare\nat risk of the disease and an estimated 10 million people areinfected worldwide, mostly in Latin America where Chagasdiseaseisendemic.Approximately20,000deathsattributabletoChagasdiseaseoccurannually[ 2].\nThe Chagas disease presents itself in two phases. The", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-2", "text": "The Chagas disease presents itself in two phases. The\ninitial,acutephaselastsforabouttwomonthsafterinfection.Duringtheacutephase,ahighnumberofparasitescirculatein the blood. When the Chagas disease is diagnosed earlyin this phase and a treatment is initiated, the patient can becured. During the chronic phase, the parasites are hiddenmainly in the heart and digestive muscle. In later years theinfectioncanleadtosuddendeathorheartfailurecausedbyprogressivedestructionoftheheartmuscle.Sometestscanbeusefulformakingadiagnosis,depend-\ning on the phase of the disease. According to [ 3], the most\ntypical tests used for the diagnosis of the Chagas disease\nare blood culture, chest X-ray echocardiogram, electrocar-\ndiogram (ECG), enzyme-linked immunoassay (ELISA), and\nperipheralbloodsmears.Uptodate,oneofthemosteffective\nways of detecting the Chagas disease in its initial phase isthrough the ELISA test. Another commonly used method is\ntheChagasStat-Pakrapidimmunochromatographictest[ 4],\nwhich provides a performance comparable to that obtained\nwithELISA.\nScreening blood donors for Chagas disease is of much\nconcerninallLatinAmericancountries.AlthoughtheW orldHealth Organization (WHO) expert committee and some\nguidelines recommend a single ELISA test to screen blood\ndonors[5],insomecountries,suchasBrazil,thereisamore\nrestrictiveregulation,recommendingtwosimultaneoustests\nofdifferenttechniques[ 6],performedinparallel.Oneofthe\ntests that can be performed in parallel is the inspection ofperipheralbloodsmears.\nHindawi Publishing Corporation", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-3", "text": "Hindawi Publishing Corporation\nComputational and Mathematical Methods in Medicine\nVolume 2015, Article ID 139681, 13 pages\nhttp://dx.doi.org/10.1155/2015/1396812 ComputationalandMathematicalMethodsinMedicine\nA peripheral blood smear is basically a glass microscope\ns l i d ec o a t e do no n es i d ew i t hat h i nl a y e ro fv e n o u sb l o o d .The slide is stained with a dye, usually Wright\u2019s stain, andexaminedunderamicroscope.Eventhoughvisualdetectionof the Chagas parasite through microscopic inspection ofp e r i p h e r a lb l o o ds m e a r si st h em o s tw i d e l yu s e dt e c h n i q u e\nfor parasitemia determination, it is a time-consuming and\nlaborious process. When the number of blood screeningsperformed in a laboratory increases, it becomes a problem.To cope with this problem, we introduce an automaticcomputational method for the detection of Chagas parasitesbasedonmachinelearningandimageprocessingalgorithms.Chagas detection using automatic image analysis is, to thebest of our knowledge, not yet studied as it is evidenced bythelackofpublicationsonthistopic.\nCurrently, there is only a couple of papers reporting\nresultsonChagasparasitesdetectionusingmachinelearningmethods [ 8,9]. In the former, a Gaussian discriminant\nanalysis is implemented and the resulting performance ratesare0.0167false-negatives,0.1563false-positives,0.8437true-negatives,and0.9833true-positives.Inthelatter,a \ud835\udc58-nearest", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-4", "text": "neighbors binary classifier is trained and the performancesare 0.98 and 0.80 in sensitivity and specificity terms, respec-tively. The results reported in this paper using AdaBoost +SVMhaveasignificantimprovementoverthepreviousones.\nA p a r tf r o mt h e s et w op a p e r s ,n oo t h e rs t u d yu s i n g\nmachine learning and computer vision methods has beenreported for the detection of the Chagas parasite, to the besto fo u rk n o w l e d g e .H o w e v e r ,s u c hk i n do ft e c h n i q u e sh a v ebeen extensively used for the detection of Malaria parasites\n[10\u201316]. All the approaches we reviewed utilize supervised\nand/orunsupervisedlearningmethodstodetect,classifyandq u a n t i f yt h en u m b e ro fm a l a r i ap a r a s i t e so nb l o o di m a g e s .The general process commonly reported can be divided inthreestages: (1)segmentation; (2)featuresextraction;and (3)\nclassification.\nIn the first stage, the segmentation is obtained through\ndifferentmethodsbasedonimagehistrogramcomputations.The second stage is performed through the computation ofdifferent features that can be classified in four categories[15]: texture features, color features, geometric features, and\nfeatures obtained from human expert knowledge. The thirdand final stage makes the biggest difference among themethods, some applied neural networks [ 10], some others\napplied statistical measures [ 11,13], bayesian classifiers [ 12],\n\ud835\udc58-means clustering [ 17], and other methods [ 14,16]. Brief", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-5", "text": "\ud835\udc58-means clustering [ 17], and other methods [ 14,16]. Brief\nsummariesoftheapproachesmorerelevantforourownworkarepresentednext.\nIn [10], an image classification system implementing\na two-stage tree classifier using back-propogation neuralnetworks is introduced. Such a system identifies malariaparasites present in thin blood smears and classifies themaccording to their different species. Image features based oncolor,texture,andthegeometryallowthesystemtoperform\nmorphological and threshold selection of possible parasites,\ndistinguishingthemfromsaneerythrocytesandothercells.\nSimilarly, the study presented in [ 11] provides a method\nfor quantification and classification of erythrocytes instained thin blood films infected with malaria parasites.Thisapproachiscomposedofthreemainphases.First,there\nisapreprocessingstep,whichcorrectsluminancedifferences.Second,thereisasegmentationstepthatusesthenormalizedRGBcolorspaceforclassifyingpixelseitheraserythrocyteorbackground. Third, there is a two-step classification processidentifies infected erythrocytes and diagnoses the infection\nstage, using a trained bank of classifiers. An interesting\ncharacteristic of this method is that user intervention isallowed when the approach cannot make a proper decision.Automatic identification of infected erythrocytes showed aspecificity of99.7% and a sensitivityof94%. Meanwhile, theinfectionstagewasdeterminedwithanaveragesensitivityof78.8%andaveragespecificityof91.2%.\nIn [12], the parasite detector uses a Bayesian pixel classi-", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-6", "text": "In [12], the parasite detector uses a Bayesian pixel classi-\nfier to mark stained pixels. The class conditional probabilitydensity functions of the stained and the nonstained classesareestimatedusinganonparametrichistogrammethod.Thestained pixels are further processed to extract features suchasHumoments,relativeshapemeasurements,andcolorforaparasiteornonparasiteclassifier. Finallya distanceweighted\ud835\udc58-nearest neighbor classifier is trained with the extracted\nfeatures achieving 74% of sensitivity, 98% of specificity, 88%ofpositiveprediction,and95%ofnegativepredictionvaluesforthemalariaparasitedetection.\nAnother malaria study based on pattern matching and\nparameter optimization is presented in [ 13]. In this case, the\nparasitaemia measurements are carried out by partitioningtheuninfectedandinfectedcellsusingunsupervisedmachinelearning. A comparison of the performance is done witha training-based method which improves the classification\nratesgiving92%ofprecisionand95%ofrecall.\nEven when some machine learning methods have been\nsuccessfully applied to the detection, classification, andquantification of Malaria parasites, these methods cannotbe employed in a straightforward way to detect Chagasparasites for one powerful reason: their morphology isdifferent. Malaria parasites have a ring shape ( Figure1(a) )\nwhile Chagas parasites have a curved shape, similar to ashrimp (Figure1(b) ). This difference in their morphology\nprevents us from using exactly the same features during theveryimportantstageoffeaturesselection.\nIn addition to detection algorithms, some authors have\ndeveloped complete automated systems. In [ 18], an auto-\nmated system to identify and analyze parasite species onthick blood films by image analysis techniques is presented.The system comprises two main components: (1)image\nacquisition unit and (2)image analysis module. The authors", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-7", "text": "acquisition unit and (2)image analysis module. The authors\nhave developed an image acquisition system that can beeasily mounted on most conventional light microscopes. Itautomatically controls the movement of microscope stagein 3-directional planes. The vertical adjustment (focusing)can be made in a nanometer range (7\u20139nm). Images areacquiredwithadigitalcamerathatisinstalledatthetopofthemicroscope. The captured images are analyzed by an image\nanalysis software which utilizes computer vision algorithms\ntodetectandidentifymalariaparasites.\nOtherworksrelatedtotheapplicationofpatternrecogni-\ntion methods for images taken with microscopes are those\nused for the detection of special types of cells such asComputationalandMathematicalMethodsinMedicine 3\n(a)\n (b)\n (c)\nFigure 1: (a) Malaria parasites in early form (inside green squares); (b) Chagas parasite (inside red square); (c) blood cell that looks very\nsimilartoaChagasparasite(insidebluesquare).\ncancerous cells [ 19] and cervical cells [ 20]. In the first work\nt h ea u t h o r su s e dt h es h a p e ,s i z e ,a n dt e x t u r eo ft h ec e l l sto perform a classification, meanwhile in the second worka\ud835\udc58-means clustering algorithm for designing binary tree\nclassifiers is used, along with the Bhattacharyya distancemetric.\nInthispaper,wepresentacomparisonoftwoofthemost\nrobust algorithms for binary classification. Both methodshavebeenextensivelystudiedandtestedwithdifficultpatternrecognition problems on images such as face detection [ 21,\n22]. We provide an AdaBoost and a SVM learning solution", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-8", "text": "22]. We provide an AdaBoost and a SVM learning solution\nto the task of Chagas parasite detection in blood images.Our AdaBoost solution includes the definition of a newset of Haar-like features specially designed for learning theChagas parasite\u2019s morphology pattern. We give details ofthe algorithms and our experimental setup. With the bestresulting method, AdaBoost, we get 100% and 93.25% ofsensitivity and specificity, respectively. Our experimentalwork shows mainly two things: (1)Chagas parasites can be\ndetectedautomaticallyusingmachinelearningmethodsand(2)AdaBoost+SVMprovidesbetterdetectionperformance\nthanplainAdaBoostorSVMs.\nIn the next section, we provide a brief review of the\ntwo compared machine learning methods: AdaBoost andSVM.Section3 describesthefourstagesrequiredtoanalyze\nt h ei m a g e sl o o k i n gf o rs u b w i n d o w sc o n t a i n i n gaC h a g a sparasite. The experimental work is detailed in Section4,\nwhich contains the experimental methodology we followedand the experimental results. In Section5,w ed i s c u s st h e\nresults and point out the issues that need to be addressed ifthesystemistobeestablishedanddeployedinreality.Finally,ourconclusionsareexplainedin Section6.\n2. AdaBoost and SVM\nAdaBoost is the short for Adaptive Boosting which is cur-\nr e n t l yt h em o s tu s e db o o s t i n gm e t h o d .Th eg o a lo fb o o s t i n gis to improve the accuracy of any given learning algorithm.Boosting creates an ensemble of classifiers by training andaddingonecomponentclassifieratatime.Eachnewclassifier", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-9", "text": "istrainedusingadifferentsubsetofexamples.Thenewtrain-ingsubsetcontainsexamplesthatareincorrectlyclassifiedbythe current ensemble. By doing such an iterative selection ofdifficult examples, boosting methods improve the accuracyof any supervised machine learning algorithm. Althougheach component classifier has an accuracy just greater thanaverage, the joint decision rule of the ensemble has a highaccuracywithallpreviouslyselectedtrainingexamples[ 23].\n2.1. Boosted Cascade of Simple Features. In 2001, Viola and\nJones [7] introduced an extremely rapid approach for visual\nobjectdetectionmotivatedbythetaskoffacedetection.Suchanapproachhasasitscentralcomponentanalgorithmbasedon the standard AdaBoost learning method. In addition tot h eu s eo fb o o s t i n g ,t w oa r et h ec h a r a c t e r i s t i c st h a tm a k ethis approach very fast and efficient. First, the use of theso-called integral image allows the features used by thedetector to be computed very quickly. Second, a methodfor combining the increasingly more complex classifiers in acascade,asillustratedin Figure2,allowsbackgroundregions\no ft h ei m a g et ob eq u i c k l yd i s c a r d e dw h i l es p e n d i n gm o r e\ncomputingresourcesonpromisingregions.\nThealgorithmintroducedbyViolaandJonesispresented\nnext.\n(i) Given example images (\ud835\udc65\n1,\ud835\udc661) ,...,( \ud835\udc65\ud835\udc5b,\ud835\udc66\ud835\udc5b)where\n\ud835\udc66\ud835\udc56=0 , 1for negative and positive examples, respec-\ntively.", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-10", "text": "tively.\n(ii) Initializeweights \ud835\udc641,\ud835\udc56= 1/2\ud835\udc5a, 1/2\ud835\udc59 for\ud835\udc66\ud835\udc56=0 , 1resp-\nectively, where \ud835\udc5aand\ud835\udc59are the number of negatives\nandpositives,respectively.\n(iii) For \ud835\udc61=1 ,...,\ud835\udc47 :\n(1) Normalizetheweights,\n\ud835\udc64\ud835\udc61,\ud835\udc56\u2190\udb8e\ude00\ud835\udc64\ud835\udc61,\ud835\udc56\n\u2211\ud835\udc5b\n\ud835\udc57=1\ud835\udc64\ud835\udc61,\ud835\udc57(1)\nsothat \ud835\udc64\ud835\udc61isaprobabilitydistribution.4 ComputationalandMathematicalMethodsinMedicine\nCascade of classifiers\nAll\nsubwindows\nsubwindows\nsubwindowsAcceptedTrue True True\nFalse False False\nDiscardedh1 h2 hT\u00b7\u00b7\u00b7\nFigure 2: The detection cascade. Classifiers with increasing com-\nplexity are arranged in a cascade scheme to allow background\nregions of the image to be quickly discarded while spending morecomputingresourcesonpromisingregions.\n(2) For each feature, \ud835\udc57, train a classifier \u210e\ud835\udc57which is\nrestricted to using a single feature. The error isevaluatedwithrespectto \ud835\udc64\n\ud835\udc61,\ud835\udf16\ud835\udc57=\u2211\ud835\udc56\ud835\udc64\ud835\udc56|\u210e\ud835\udc57(\ud835\udc65\ud835\udc56)\u2212\n\ud835\udc66\ud835\udc56|.\n(3) Choosetheclassifier, \u210e\ud835\udc61,withthelowesterror \ud835\udf16\ud835\udc61.\n(4) Updatetheweights:", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-11", "text": "(4) Updatetheweights:\n\ud835\udc64\ud835\udc61+1,\ud835\udc56=\ud835\udc64\ud835\udc61,\ud835\udc56\ud835\udefd1\u2212\ud835\udc52\ud835\udc56\n\ud835\udc61, (2)\nwhere \ud835\udc52\ud835\udc56=0if, for example, \ud835\udc65\ud835\udc56is classified\ncorrectly, \ud835\udc52\ud835\udc56=1otherwise,and \ud835\udefd\ud835\udc61=\ud835\udf16\ud835\udc61/(1 \u2212 \ud835\udf16\ud835\udc61).\n(iv) Thefinalstrongclassifieris\n\u210e(\ud835\udc65)={{\n{{\n{1\ud835\udc47\n\u2211\n\ud835\udc61=1\ud835\udefc\ud835\udc61\u210e\ud835\udc61(\ud835\udc65)\u22651\n2\ud835\udc47\n\u2211\n\ud835\udc61=1\ud835\udefc\ud835\udc61,\n0otherwise ,(3)\nwhere \ud835\udefc\ud835\udc61=log(1/\ud835\udefd\ud835\udc61).\n2.2. Support Vector Machines. As u p p o r tv e c t o rm a c h i n e\n(SVM)isalearningtoolthatoriginatedinmodernstatisticallearning theory [ 24]. In recent years, SVM learning has\nfound a wide range of real-world applications, includinghandwritten digit recognition, object recognition, speakeridentification, face detection in images, and text catego-rization. The formulation of SVM learning is based on theprinciple of structural risk minimization. SVM tends toperform well when applied to data outside the training setand it has been reported that SVM-based approaches areabletosignificantlyoutperformcompetingmethodsinmanyapplications. SVM achieves this by focusing on the trainingexamples that are most difficult to classify. These trainingexamplesarecalledthesupportvectors.\n3. Chagas Parasites Detection", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-12", "text": "3. Chagas Parasites Detection\nThe whole process of Chagas detection is divided into fourstages, as illustrated in Figure3. This process has been spe-\ncially designed to allow the automatization of the diagnosisas much as possible. The robustness of our methodologyrelies on the four modular stages that can be implemented\nImage\nacquisition Preprocessing\nLearning-based\npostprocessingAdaBoost\ndetection\nFigure 3: The process of parasite detection consists of four stages:\n(1)image adquisition via microscope; (2)image conversion from\nRGBcolortograyscaleformat; (3)possibleparasitesdetectionusing\nAdaBoost; and (4)amount of DNA pixels used to further discard\nfalsepositives.\na ss e p a r a t e dp r o g r a m m i n gc l a s s e s ,w h e r ee v e np a r a l l e lp r o -\ngramming,thatis, usingGraphics Processing Units(GPUs),c a na l s ob ee m p l o y e dt or e d u c et h et i m eo fd e t e c t i o n[25]. This methodology is generic enough to be applied in\notherobjectdetectionsystems,particularlyfromthebiologydomain where samples observed through a microscope arestained.\nThe first step consists in obtaining a digital image using\nac a m e r aa n dam i c r o s c o p e .O n c ew eh a v et h ei m a g ei nRGB format, in a second step we convert it to a grayscaleimagetoreducetheamountofinformationthattheAdaBoostalgorithm needs to process. In the third stage, the grayscale\nimage is scanned using our previously trained AdaBoost", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-13", "text": "image is scanned using our previously trained AdaBoost\nclassifier in order to detect subwindows of pixels containingpossibleparasites.Finally,inthefourthstep,weusethegreencomponent of the original RGB image to extract some threefeatures related to the number of pixels representing a highDNA (deoxyribonucleic acid) content. Based on the amountofDN Aitispossibletodiscardfalseparasites.ThislaststageisimplementedwithaSVManditisveryimportantbecauseitallowsustoinclude,aspartoftheclassificationprocedure,aprioriknowledgeabouttheDNAoftheparasites.\nIt is important to mention that the DNA from both\nparasitesandnonparasitessuchaswhitebloodcellstendstoabsorbdifferentamountsofthestainemployed.However,thepattern generated by the stained DNA of parasites is clearlydifferent from the pattern generated by the stained DNAof white blood cells. Such a difference is recognized by thepatternrecognitionalgorithms.\n3.1.Stage1:ImageAcquisition. Agroupofmicewereinfected\nwith an inoculation of 5 \u00d7 104blood trypomastigotes of T.\ncruzivia intraperitoneal. Once the mice were infected, the\nparasitaemia detection started in average between 11 and 15daysafterwards.Atthistime,thebloodsmearswerepreparedandstainedusingWrightstain,whichallowstheobservationofthemorphologyofdifferentbloodcells,aswellasparasitesComputationalandMathematicalMethodsinMedicine 5\n(a) (b)\nFigure4:(a)GenericHaar-likefeaturesinspiredbythoseproposedbyViolaandJones[ 7];(b)Haar-likefeaturesspeciallydesignedtocapture\nChagasparasite\u2019smorphology.", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-14", "text": "Chagasparasite\u2019smorphology.\nsuch asT. Cruzi,Leishmania sp., andPlasmodium sp,. After\nthe staining process the blood smears were placed verticallyand were left to dry. Finally, an optical Nikon Eclipse E600microscope was used to take images, first at 10x and then at100x,witharesultingsizeincreaseof1000times.\n3.2. Stage 2: RGB to Grayscale Format. The conversion from\nan RGB image to a grayscale image involves a simple mani-pulation of matrices and is performed to reduce the amountofinformationinvolvedinthelearninganddetectionprocess.The RGB color model is an additive color model in whichred, green, and blue light are added together in differentproportions to produce a wide array of colors. A color inthe RGB color model is described by indicating how muchof each of the red, green, and blue is included. The color isexpressedasatriplet (\ud835\udc5f, \ud835\udc54, \ud835\udc4f)whereeachcomponentcanvary\nfrom zero to a defined maximum value, which in our casewas256.R G Bi m a g e sa r es t o r e di nm e m o r ya s \ud835\udc64\u00d7\u210e\u00d73\nmatrices, where \ud835\udc64is the width and \u210eis the height of the\nimagemeasuredinnumberofpixels.Thethirddimensionofthe matrix which is of size 3corresponds to the 3different", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-15", "text": "color components of the image: red, green, and blue. Toconvert any RGB image to a grayscale representation of itsluminance, first one must obtain the values of its red, green,andbluecomponents.Then,weneedtoaddtogether30%oft h er e dv a l u e ,5 9 %o ft h eg r e e nv a l u e ,a n d1 1 %o ft h eb l u e\nvalue. These percentages are regarded as typical values. The\nresultingimageisstoredina \ud835\udc64\u00d7\u210eimagecontainingpixelsof\ndifferentintensitiesofthegraycolor.\n3.3. Stage 3: Parasites Detection. The Chagas detection pro-\ncess employs a set of Haar-like features and one AdaBoost\nbinary classifier. Haar-like features provide informationabout clear and dark regions in images. Such informationis very useful to detect types of objects that share thesame morphological pattern. Since Chagas parasites can beconsidered objects that share the same shrimp-like shape,t h e yc a nb ed e t e c t e du s i n gt h ea p p r o p r i a t es e to fH a a r - l i k efeatures.\nGivenadetectionwindowtakenfromaspecificlocation\nofthegrayscaleimage,aHaar-likefeatureconsidersadjacentrectangularregions,sumsupthepixelintensitiescontainedineachrectangularregionandcalculatesthedifferencebetweenthese sums. This difference is then compared to alearnedthreshold that separates nonparasites from parasites. The\nrectangular regions are defined in such a way that relevantdark pixels fall into one same region, meanwhile clearpixels fallintoanotherone. Figure4illustratestheHaar-like\nfeaturesutilizedtodetectChagasparasites.\nThe nine features in Figure4(a) were inspired by the", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-16", "text": "The nine features in Figure4(a) were inspired by the\noriginalHaar-likefeaturesproposedbyViolaandJonesin[ 7].\nThesetoffourHaar-likefeaturesin Figure4(b) weredesigned\nspecifically to represent the shrimp-like shape adopted byChagas parasites. After running ten times the AdaBoostalgorithm and examining the number of times that thealgorithmchosetheHaar-likefeaturesasweaklearners,itwaspossible to effectively discard those Haar-like features thatfailed to capture important patterns of the Chagas parasites,while keeping only the best ones. Figure5illustrates how\nspecially designed Haar-like features adapt better to themorphologyoftheparasite.\nMoreover, the position of these rectangular regions is\ndefined relative to a detection window that acts like abounding box to the target parasite. In the detection phase,a window of the target size is moved over the input image,andforeachsubsectionoftheimage,theHaar-likefeatureis\ncomputed.\nThe Haar-like features can be computed very fast using\nthe concept of integral image. Given an image \ud835\udc3cof size \ud835\udc45\u00d7\n\ud835\udc36,w h e r e \ud835\udc45is the number of rows and \ud835\udc36is the number of\ncolumns, the integral image of a point located in row \ud835\udc5fand\ncolumn \ud835\udc50isdefinedas\n\ud835\udc56\ud835\udc56(\ud835\udc5f, \ud835\udc50)=\u2211\n\ud835\udc5f\udba0\udc20\u2264\ud835\udc5f,\ud835\udc50\udba0\udc20\u2264\ud835\udc50\ud835\udc3c( \ud835\udc5f\udba0\udc20,\ud835\udc50\udba0\udc20),(4)\nwhere \ud835\udc3c(\ud835\udc5f\udba0\udc20,\ud835\udc50\udba0\udc20)is the gray intensity value of pixels located in", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-17", "text": "row\ud835\udc5f\udba0\udc20andcolumn \ud835\udc50\udba0\udc20.Notethatthenumberofrowsincreases\nintheimagefromtoptobottomandthenumberofcolumnsincreasesfromlefttoright.\nNow, given the same image \ud835\udc3cand four points\n\ud835\udc5d\n1(\ud835\udc5f1,\ud835\udc501), \ud835\udc5d2(\ud835\udc5f2,\ud835\udc502), \ud835\udc5d3(\ud835\udc5f3,\ud835\udc503),a n d \ud835\udc5d4(\ud835\udc5f4,\ud835\udc504)as illustrated\ninFigure6(a) , and using the definition of integral image of\n(4),w ec a nc o m p u t et h er e g i o n s \ud835\udc461,\ud835\udc462,\ud835\udc463,a n d \ud835\udc464as follows.\nTo calculate the sum of pixels in subwindow \ud835\udc461, denoted by\n\ud835\udf0e(\ud835\udc461),wesimplycompute\n\ud835\udf0e( \ud835\udc461)=\ud835\udc56\ud835\udc56 ( \ud835\udc5d1)=\ud835\udc56\ud835\udc56 ( \ud835\udc5f1,\ud835\udc501). (5)6 ComputationalandMathematicalMethodsinMedicine\nFigure 5: Haar-like features on top can capture more information abou tthemorphology oftheChagas parasite,given their circular shape,\nthanthegenericHaar-likefeaturesproposedbyViolaandJones,shownonthebottom.\nS1S2\nS3S4(r1,c1) (r2,c2)\n(r3,c3)( r4,c4)(1,C)\n(R,C ) (R,1)(1,1)\n(a)\nR1R2\nR3\nR4R5\n(b)", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-18", "text": "(a)\nR1R2\nR3\nR4R5\n(b)\nFigure6:(a)Genericintegralimagecomputation;(b)computationofoneChagas-specificintegralimage.\nFor\ud835\udc462and\ud835\udc463wecompute,respectively,\n\ud835\udf0e( \ud835\udc462)=\ud835\udc56\ud835\udc56 ( \ud835\udc5f2,\ud835\udc502)\u2212\ud835\udc56\ud835\udc56 ( \ud835\udc5f1,\ud835\udc501),\n\ud835\udf0e( \ud835\udc463)=\ud835\udc56\ud835\udc56 ( \ud835\udc5f3,\ud835\udc503)\u2212\ud835\udc56\ud835\udc56 ( \ud835\udc5f1,\ud835\udc501).(6)\nFinally,for \ud835\udc464,wecalculate\n\ud835\udf0e( \ud835\udc464)=\ud835\udc56\ud835\udc56 ( \ud835\udc5f4,\ud835\udc504)\u2212\ud835\udc56\ud835\udc56 ( \ud835\udc5f3,\ud835\udc503)\u2212\ud835\udc56\ud835\udc56 ( \ud835\udc5f2,\ud835\udc502)+\ud835\udc56\ud835\udc56 ( \ud835\udc5f1,\ud835\udc501).(7)\nIn general, for an image of size \ud835\udc45\u00d7\ud835\udc36,thewho lein t egral\nimage \ud835\udc56\ud835\udc56(\ud835\udc45, \ud835\udc36)c a nb ec o m p u t e di no n l yo n ep a s so v e rt h e\nimageandcanbestoredina2-dimensionalarray.Therefore,tocomputeasubwindowlike \ud835\udc46\n4inFigure6(a) ,weonlyneed\ntoaccesstheintegralimagearrayfourtimes.Furthermore,tocomputetheintegralimageofourChagasHaar-likefeatures,\nwejustneedtosegmenteachHaar-likefeatureintheminimalnumber \ud835\udc5b\n\ud835\udc5fof subwindows required to cover it, this number", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-19", "text": "\ud835\udc5fof subwindows required to cover it, this number\nbeing \ud835\udc5b\ud835\udc5f=3for the simplest one and \ud835\udc5b\ud835\udc5f=5for the most\ncomplex. The computation of one Chagas-specific integralimagewith5subwindowsisillustratedin Figure6(b) ,where\ntheHaar-likefeaturetemplatehasbeendividedinto5regions,namely, \ud835\udc45\n1,...,\ud835\udc455andinordertocom putethefea turevalue\nweneedtocalculate \ud835\udf0e(\ud835\udc451)+\ud835\udf0e ( \ud835\udc452)+\ud835\udf0e ( \ud835\udc453)+\ud835\udf0e ( \ud835\udc454)\u2212\ud835\udf0e ( \ud835\udc455).\nInconclusion,thecomputationoftheintegralimageforourspeciallydesignedHaar-likefeaturesisstillveryfast.\n3.4. Stage 4: Postprocessing. The cascade application of the\nAdaBoost algorithm together with the appropriate set ofComputationalandMathematicalMethodsinMedicine 7\nDNA\n(a)\n180\n140\n100\n60\n12\n8\n4\n004812\n(b)\nFigure7:(a)DarkspotofpixelsgeneratedbyanaccumulationofDNA;(b)aDNAspotseenasasurfaceof 11 \u00d7 11pixels.\nHaar-like features is highly effective in finding most of the\ntrueChagasparasites.However,italsodetectsasmallnumberof false parasites. The postprocessing filter described in thissectionhelpstodiscardfalseparasitesanddecreasesthefalsepositiverateofourmethod.\nThepostprocessingfilterisfocusedontheanalysisofone\ndark spot of pixels appearing in all parasites\u2019 bodies. Thisspot, indicated by an arrow in Figure7(a) , corresponds to", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-20", "text": "an accumulation of DNA of the Chagas parasite. When the\nimageisplottedasasurface,itcanbeseenthatsuchDNAspothas one particular valley-like shape, shown in Figure7(b) .\nSincetheshapeofthisdarkspotandthelowintensityvaluestakenbytheircorresponingpixelscreateapatternveryusefultodiscriminatebetweenparasitesandnonparasites,theywereusedasasourceforfeatureextractionasitisdescribednext.\nGiven that the stained pixels have low intensities, these\nvalues being not greater than 80, in the green component oftheRGBimage,thefollowingthreefeatureswereusedtotrainaSVM.\n(i) Feature 1 . Given a subwindow detected by AdaBoost, the\npercentage of pixels that have intensities at most 80 wascomputed. This percentage, which represents the size of thestainedregionoftheparasite,isthefirstfeature.\n(ii) Feature 2 . The mean of the intensities of all pixels with\nindividual intensities at most 80 was encountered in thesubwindowdetectedbyAdaBoost.\n(iii) Feature 3 . The standard deviation of the intensities of all\npixelswithindividualintensitiesatmost80wasencounteredinthesubwindowdetectedbyAdaBoost.\nThe choice of the green component over the red and the\nblue ones was decided after a careful visual examination ofa subset of images in the original RGB version and in all\n3 individual components: red, green, and blue. After such\nvisualexamination,itwasclearthatthegreencomponentwassuperiortotheothersintermsofdiscriminativeinformation.In other words, the information provided by the green\nchannelallowsustoperformamoreaccurateclassificationofparasitesandnonparasites.\n4. Experimental Work", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-21", "text": "4. Experimental Work\n4.1.ExperimentalMethodology. W etestedtheproposedalgo-\nrithm using a data collected in the Instituto de Investiga-ciones Regionales at the Universidad Aut \u00b4onomadeYucat \u00b4an,\nM\u00b4exico. We had available for our study a total of 120\nc o l o ri m a g e so fd i m e n s i o n2 5 6 \u00d7256 pixels. Sixty of these\nimages were specially selected to contain a Chagas parasite.Meanwhile, the other 60 remaining images were selectedto contain nonparasites. Machine learning was employed togenerate our basic classifier algorithm, which was then usedfor the more general Chagas parasite detection process. Theresultsweprovideinthispaperwereobtainedusingatypicalpattern recognition methodology [ 26], using training and\ntestingsetsofimages.\nA 10-fold cross-validation procedure was used for train-\ning and testing the AdaBoost and the SVM + FeatureExtraction learning methods. Since AdaBoost is sensible torotation, during the training phase, we generated a total of840 positive examples of parasites images by rotating the\noriginal images in increasing amounts of 15 degrees. For\nevery experiment, we used 756 positive images for training,leaving 84 positive images for testing. Even though we werelimited by the number of positive examples available, thenumber of negative examples could be defined as desired.All negative examples were generated by random samplingfrom the original 256 \u00d7 256 pixels images. For training,", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-22", "text": "AdaBoost used as many negative examples as possible toreducethefalsepositiverate,andthefinallearnedmodelwastested with 10800 negative images. In the case of the SVM +Feature Extraction methods, we used 1296 negative imagesfortrainingand10800fortesting.Itisworthmentioningthatonce the classifiers have been trained, the proper analysis of8 ComputationalandMathematicalMethodsinMedicine\na256 \u00d7256image,whichmeansthatthealgorithmisactually\nlookingforparasites,isperformedinrealtime.Wearetalkingabout a few milliseconds, using a commercial laptop ordesktopcomputer.Therefore,severalhundredimagescanbeanalyzedinamatterofminutes.Moreover,ourprogramwasimplementedinMatlabusing2publiclibraries:fdtool,forthe\nAdaBoostimplementation,andlibsvm,forthesupportvector\nmachineimplementation.\n4.1.1.ComparisonwithSupportVectorMachines. Theapplica-\ntionofSVMlearningrequiresthecomputationofonefeaturevector for each subwindow that needs to be classified asparasiteornonparasite.Suchfeaturesvectorsarethenpassedto the SVM classifier which makes the final decision. Thes e l e c t i o no ft h ef e a t u r e si sak e yp a r to ft h ec l a s s i fi c a t i o nprocess, simply because by using the wrong features thelearning algorithm cannot extract the patterns required tobuild a confident classifier, resulting in a high classificationerror. In order to obtain the best results with the applicationof SVMs, we reviewed the most relevant published worksinvolving SVM learning and malaria parasite detection. Sothatwecanuseforourexperimentalworkthosefeaturesthathavebeenpreviouslyreportedtoprovidehighqualityresults.\nColor,shape,andtexturefeatureswereselectedtoforma", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-23", "text": "Color,shape,andtexturefeatureswereselectedtoforma\n38-dimensional feature vector. The color and shape featuresare computed from 7 different histograms that containinformation about the red channel and green channel of theRGBimage,thehueandsaturationcomponentfromtheHSVimage, the grayscale pixel intensities and the result of the\napplication of the sobel operator horizontally and vertically\nover the RGB image. Once the 7 histograms are computed,5 values of each histrogram form the first 35 features. Thevaluestakenfromthehistogramsarethemean,thestandarddeviation, kurtosis, skewness, and entropy. Finally, 3 morefeatures known as Tamura texture parameters are computedfrom the original RGB images. Those final parameters arecoarseness, contrast, and direction. All theses features areclassified in [ 15] aspartoftheirworkinautomaticdetection\nofMalariaparasitesandtheyaredescribedin[ 27].\n4.1.2.Sensitivity,Specificity,and \ud835\udc39-Measure. Inordertoeval-\nuate the performance of our implemented AdaBoost clas-sifier, we used three different statistical metrics: sensitivity,specificity, and effectiveness. Sensitivity is the probability ofa positive test given that the patient is ill and it is computedasfollows:\nsensitivity\n=number of true positives\nnumber of true positives +number of false negatives.\n(8)\nSpecificityinturnistheprobabilityofanegativetestgiven\nthatthepatientiswellanditiscomputedasfollows:\nspecificity\n=number of true negatives\nnumber of true negatives +number of false positives.\n(9)Due to the unbalanced class distribution between sub-", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-24", "text": "(9)Due to the unbalanced class distribution between sub-\nwindowscontainingaChagasparasiteandsubwindowscon-tainingsomethingelse,theuseofstatisticalmetricsrelatedtotheeffectivenesswerealsoconsidered.Intheseexperiments,the\ud835\udc39-measureisused:\n\ud835\udc39\n\ud835\udefd=TPR\u2217PR\n\ud835\udefd\u2217TPR+( 1\u2212\ud835\udefd )\u2217 PR(10)\nwith\nTPR =TP\nTP+FN,\nPR=TP\nTP+FP,(11)\nwhere TP stands for the true positives, FN for the false\nnegatives, and FP for the false positives. The \ud835\udefdcoefficient\n(0<\ud835\udefd<1 )allowstoassignrelativeweightstoboththetrue\npositiveandprecisionrates.Intheseexperiments, \ud835\udefd = 0.05is\nused,sothesearchwasaddressedtodetectionofTP.\n4.2.ExperimentalResults\n4.2.1. AdaBoost Results. Looking at the values provided in\nTable 2, we see that AdaBoost methods are as good as SVM\nmethodsintermsofsensitivity,withmeanvalueof1andstan-\ndard deviation of 0. All five methods were trained to detect\neveryparasiteinthetestsets.Intermsofspecificity,AdaBoost+ Postprocessing is clearly the winner among the five meth-odsbeingcompared,withmeanvalueof0.9325andstandarddeviation of 0.0496, as given in Table 3. Tables4and5\nalso show that AdaBoost + Postprocessing is superior to theothers, in terms of \ud835\udc39-measures. In Figure8,w ei l l u s t r a t e", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-25", "text": "t h er e s u l to fo u rA d a B o o s ta l g o r i t h mf o rC h a g a sp a r a s i t edetection.Theseresultswereobtainedwithacascadeofweakclassifiers of 11 stages. Figure9provides the ROC curves\nfor SVMs using different degrees of polynomial kernels andFigure10 comparestheROCcurvesforAdaBoostandSVMs\nmethodsusingdifferenttypesofkernels.\n4.2.2. SVM + Feature Extraction Results. SVM results are\ngood especially using linear and polynomial kernels. Fea-tures used to train the SVM proved to contain the neededinformation to build good classifiers. Those features werecreated from color histograms and some other values thatencapsulatetextureinformationoftheimage.Thebestresultsforeachtypeofkernelwereobtainedwiththelibrarylibsvm,finding first the best pair of parameters \ud835\udefeand\ud835\udc36for our\ntrainingexamples.Thoseparametersaregivenin Table 1.The\nAdaBoost + Postprocessing method was implemented usinga linear kernel SVM with parameters \ud835\udefe = 0.002 and\ud835\udc36=\n0.5.Theseparameterswereobtainedthroughmultiplecross-\nvalidationexperimentsusingdifferentpairsofvaluesforthe\ud835\udefeand\ud835\udc36parameters.\n4.2.3.AdaBoostversusSVM+FeatureExtractionComparison.\nResults obtained with AdaBoost are more robust than thoseobtained with SVM. However, the sensitivity and specificityComputationalandMathematicalMethodsinMedicine 9\n(a)\n (b)\n (c)\n(d)\n (e)\n (f)\n(g)\n (h)\n (i)\n(j)\n (k)\n (l)", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-26", "text": "(h)\n (i)\n(j)\n (k)\n (l)\nFigure 8: Results of the Chagas parasite detection algorithm with a sample of images. Each column corresponds to a different example of\ndetection. The first row contains the color images in RGB format; the second row shows the grayscale images once the parasites have been\ndetected by the classifier; in the third row we show the result of the postprocessing stage; and the fourth row shows the final result of thedetectionsystems.10 ComputationalandMathematicalMethodsinMedicine\n0 0.005 0.01 0.015 0.02 0.025 0.03 0.03500.10.20.30.40.50.60.70.80.91\nFalse positive rateTrue positive ratePolynomial SVM ROC curves\nSVM_poly_ 2\nSVM_poly_ 3\nSVM_poly_ 4SVM_poly_ 5\nSVM_poly_ 6\nFigure 9: ROC curves with SVMs and polynomial kernels ranging\nfromdegrees2to6.\n0 0.005 0.01 0.015 0.02 0.025 0.03 0.035\nFalse positive rate\nSVM-linear\nSVM-polynomial\nSVM-RBFAdaBoost00.10.20.30.40.50.60.70.80.9True positive rate1\nAda+Postprocessing\nFigure 10: ROC curves using AdaBoost and SVMs with three\ndifferenttypesofkernel:linear,polynomial,andRBFs.\nreached by using SVM cannot be considered bad, given the\ndifficultyofthetaskandtheirperformanceiscomparabletothe performance obtained in the malaria parasite detectiontask.\n5. Discussion", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-27", "text": "5. Discussion\nB a s e do nt h er e s u l t so b t a i n e dw i t hb o t hm a c h i n el e a r n i n gmethods, AdaBoost and SVM, we can see that taking intoaccount features related to shape, color and texture seemsTable 1: Best \ud835\udefeand\ud835\udc36parametersforSVMsaccordingtothetypeof\nkernelemployedandourtrainingdata.\nKernel \ud835\udefe\ud835\udc36\nLinear 0.00003 8\nPolynomial 0.0078 0.0312\nRadialbasisfunction 0.00003 8192\nTable 2: Sensitivity by method.\nMethod Mean Std. Dev.\nAdaBoost 1 0\nAdaBoost+Postprocessing 1 0\nSVMlinear+FeatureExtraction 1 0\nSVMpolynomial+FeatureExtraction 1 0SVMRBF+FeatureExtraction 1 0\nT able3:Specificitybymethod.\nMethod Mean Std. Dev.\nAdaBoost 0.7550 0.1820AdaBoost+Postprocessing 0.9325 0.0496SVMlinear+FeatureExtraction 0.8218 0.0118SVMpolynomial+FeatureExtraction 0.6691 0.0035\nSVMRBF+FeatureExtraction 0.7111 0.0068\nTable 4: \ud835\udc39-measure( \ud835\udefd=1)bymethod.\nMethod Mean Std. Dev.\nAdaBoost 0.7550 0.1820\nAdaBoost+Postprocessing 0.9325 0.0496\nSVMlinear+FeatureExtraction 0.8218 0.0118", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-28", "text": "SVMlinear+FeatureExtraction 0.8218 0.0118\nSVMpolynomial+FeatureExtraction 0.6691 0.0035SVMRBF+FeatureExtraction 0.7111 0.0068\nTable 5: \ud835\udc39-measure( \ud835\udefd=1 0)bymethod.\nMethod Mean Std. Dev.\nAdaBoost 0.9976 0.0018\nAdaBoost+Postprocessing 0.9993 0.0004\nSVMlinear+FeatureExtraction 0.9982 0.0001SVMpolynomial+FeatureExtraction 0.9967 0.0000\nSVMRBF+FeatureExtraction 0.9971 0.0001\nto be the most robust way to go when we target a parasite\ndetection task. The main advantage in using AdaBoost withthe approach introduced by Viola and Jones is the fastcomputation of the features using the integral images trick,as opposed to any other method of image segmentation thatarecommonlyusedtogetherwiththeSVMapproach.Imagesegmentation is a time-consuming task and its use withoutparallel computation is almost impossible for applicationsthatrequiretoscanmanyimagesinashorttime.\nThe consideration of DNA pixels through a trained\nSVM in the Ada + Postprocessing method is handled as apostprocessingprocedurebecauseitisalsoatime-consumingComputationalandMathematicalMethodsinMedicine 11\n(a)\n (b)\n(c)\n (d)\nFigure11:Tworepresentativeimages.Inimage(a),thenumberofdarkpixelsmakesiteasyforthedetectionsystemtoconsideritasavalid", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-29", "text": "parasite. In image (b), the number of dark pixels is larger than average, which makes it difficult for the detection system to consider it as avalidparasite.Aneasybackgroundlookslike(c)andadifficultonelookslike(d).\nprocedure which should be applied a reduced number of\ntimes. If we attempt to interwine this counting procedurewith the typical AdaBoost algorithm using integral images,we simply nullify the major advantage of the integral imagetrick,whichisafastcomputationofthepixelsaddition.\nInFigure11, we present some examples of images that\narerepresentativecasesofeaseanddifficultyofclassification.When an image contains a high percentage of pixels stainedindarkpurple,asin Figure11(d) ,thealgorithmfindsitmore\ndifficulttodiscardfalsepositives.\nP a r a s i t e sa n do t h e rs t a i n e do b j e c t sl i k eb l o o dc e l l sa r e\nnonrigid and therefore vary in shape and size. The color isimportantbutitsonlyusetodistinguishbetweentheChagasparasitesandotherthingsisnotenough.Ontheotherhand,rawimagescannotbeuseddirectlyasafeaturevectorfortwomain reasons: first the size of such a vector would increase\nunnecesarily the time of computation and second the high\nvariationinshapeandsizewouldmakeitextremelydifficultfor any machine learning algorithm to learn the correctclassifier. In the ideal case, features must be able to capture\nshapeandcolorcharacteristicsoftheobjectstobeclassified.\nFailing to detect Chagas is certainly a more serious", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-30", "text": "Failing to detect Chagas is certainly a more serious\nsituation than having a false positive error. Diagnosing apersonsufferingfromChagasasnothavingthediseasewouldp u th e rl i f ea tr i s k .Af a l s ep o s i t i v ea l a r mo nt h ec o n t r a r ywould only incur in a labor cost for the doctor in chargeof confirming whether the parasite is present in the bloodsample.\nInpractice,thedetectionof T.cruziby analysis of digital\nimages obtained from peripheral blood smears would bringm u l t i p l ebe n e fi t sa s soc i a t e dt oth eC h a g a sd i se a sed i a g n o s i s .First,itwouldmakepossiblethediagnosisofthediseaseinitsinitialphase,whichistheacutephase,becausethepresenceoftheparasiteinthebloodwouldindicatewithhighprobabilitythat the infection is present. As a consequence of the early\ndiagnosis, the human cases could be treated with drugs\nsuch as Nifurtimox or Benznidazole, and the probability ofstopping the disease is very high. Second, given that human12 ComputationalandMathematicalMethodsinMedicine\nc a s e so fC h a g a sa r ev e r yd i ffi c u l tt ob ed i a g n o s e dd u r i n g\nthe analysis routinely performed in national laboratories incountries such as Mexico, an automated Chagas detectionsystem would contribute to estimate the local impact of thediseaseandtodeterminetheexistenceofanyendemicregion.\nMoreover, the successful implementation of this algo-\nrithm, would motivate the laboratories and hospitals to", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-31", "text": "rithm, would motivate the laboratories and hospitals to\nreconsider the usefulness of the blood smears analysis andtotrustthistestasaconfidentlaboratorydiagnosis.Further-more, the laboratory technicians would reduce the numberof hours they need to sit in front of a microscope to analyzeso many samples day after day. Situation that is currentlyoriginatingthemproblemssuchasvisualfatigueandfrequentpainintheback.\nFinally, this algorithm can be integrated as part of one\nautomated microscopic system for the adquisitionof imagesfrom blood smears. Such a device would be composed ofoneopticalmicroscopewithacontrollableelectromechanicals u r f a c es p e c i a l l yd e s i g n e dt om o v et h eb l o o ds m e a r sa sneeded by the scanning software. The microscope would beequipped with a high-resolution camera in order to take the\nimages.Imageswouldbestoredinharddrivestoallowtheir\nposterioranalysiswithourChagasdetectionalgorithm.Thisdevice is currently being built as part of one CONACYTresearch project on Chagas led by Dr. Ruiz-Pi \u02dcna in Mexico\n(projectcode:salud-2009-01-113848).\n6. Conclusions\nIn this paper, we have provided an approach to the Chagasparasite detection problem based on AdaBoost learning.\nUsingtheapproachusedbyViolaandJonesforthetaskofface\ndetection,weobtainedhighsensitivityandspecificityvalues.\nChagasdetectionapproachesbasedonmachinelearningand\ncomputer vision methods have been barely studied as it is\nevidenced by the lack of literature on the topic. The mostpromising method proposed for the detection of malaria\nparasiteswasimplementedandcomparedtoAdaBoost.This", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-32", "text": "parasiteswasimplementedandcomparedtoAdaBoost.This\nmethod consists basically in the use of different kinds ofvalues taken from color histograms in order to form the\nf e a t u r ev e c t o rt h a ta r el a t e ru s e dt ot r a i nas u p p o r tv e c t o r\nm a c h i n e .B o t hs h a p ea n dc o l o rp r o v e dt ob ei m p o r t a n ti n\nthe task of Chagas parasite detection when high level of\nsensitivityandspecificityarerequired.ApplyingSVMforthetask of Chagas parasite detection requires the computation\nof the 38-dimensional feature vector for each subwindow\nthat we want to classify as being a parasite or nonparasite.\nHowever, if we proceed in this way with every subwindow\nof the original image of size 256 \u00d7256, the whole process\nwould be highly time-comsuming. One way to reduce thenumber of subwindows that are worth checking is by meansofsegmentation.Segmentationallowsustoidentifydifferentrelevantobjectsintheimageanddiscardthosethatdoesnotlook like parasites. Given that most of the objects detectedduringthesegmentationarediscarded,onlyaminornumberof subwindows are further check by the SVM. Although theuse of segmentation is very common for this kind of tasks,it can also be a time-consuming process. AdaBoost appliedwithintegralimagesdoesnotrequiresegmentation,because\neverysubwindowcanbecheckedinconstanttime,regardlessofitssize.\nConflict of Interests\nThe authors declare that there is no conflict of interestsregardingthepublicationofthispaper.\nAcknowledgment", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-33", "text": "Acknowledgment\nThe authors thank the Universidad Autonoma de Yucatan,f o rs u p p o r t i n gt h er e s e a r c hp r o j e c to nC h a g a sp a r a s i t e sdectectionusingmachinelearningalgorithms.\nReferences\n[1] World Health Organization, \u201cChagas disease (American try-\npanosomiasis),\u201d Fact Sheet 340, WHO, 2010, http://www.who\n.int/mediacentre/factsheets/fs340/en/ .\n[2] L. V. Kirchhoff, Chagas Disease (American Trypanosomiasis) ,\neMedicine,2010.\n[3] L. V. Kirchhoff, \u201cTrypanosoma species (American trypanoso-\nmiasis,Chagas\u2019disease):biologyoftrypanosomes,\u201din Principles\nand Practice of Infectious Diseases ,G.L.M a ndell ,J .E.Bennet t,\nandR.Dolin,Eds.,ElsevierChurchillLivingstone,Philadelphia,\nPa,USA,7thedition,2009.\n[4] C. Ponce, E. Ponce, E. Vinelli et al., \u201cValidation of a rapid and\nreliable test for diagnosis of Chagas\u2019 disease by detection ofTrypanosoma cruzi -specific antibodies in blood of donors and\npatients in Central America,\u201d J o u r n a lo fC l i n i c a lM i c r o b i o l o g y ,\nvol.43,no .10,pp .5065\u20135068,2005.\n[5] World Health Organization, \u201cControl of Chagas disease:\nsecondreportoftheWHOexpertcommittee,\u201dWHOTechnical", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-34", "text": "secondreportoftheWHOexpertcommittee,\u201dWHOTechnical\nReportSeries905,WHO,2002, http://www.who.int/iris/handle/\n10665/42443 .\n[6] \u201cBrazilian consensus on Chagas disease,\u201d Revista da Sociedade\nBrasileiradeMedicinaTropical ,vol.38,no .3,pp .7 \u201329 ,2005.\n[7] P. Viola and M. Jones, \u201cRapid object detection using a boosted\ncascadeofsimplefeatures,\u201din ProceedingsoftheIEEEComputer\nSocietyConferenceonComputerVisionandPatternRecognition(CVPR\u201901) ,vol.1,pp.I-511\u2013I-518,December2001.\n[8] V. Uc-Cetina, C. Brito-Loeza, and H. Ruiz-Pi \u02dcna, \u201cChagas par-\nasites detection through Gaussian discriminant analysis,\u201d\nAbstractionandApplication ,vol.8,pp .6\u201317 ,2013.\n[9] R.Soberanis-Mukul,V .Uc-Cetina,C.Brito-Loeza,andH.Ruiz-\nPi\u02dcna,\u201cAnautomaticalgorithmforthedetectionof Trypanosoma\ncruziparasitesinbloodsampleimages,\u201d ComputerMethodsand\nProgramsinBiomedicine ,vol.112,no .3,pp .633\u2013639 ,2013.\n[10]M.I.Khan,B.Acharya,B.K.Singh,andJ.Soni,\u201cContentbased\nimageretrievalapproachesfordetectionofmalarialparasitein\nbloodimages,\u201d InternationalJournalofBiometricsandBioinfor-\nmatics,vol.5,no.2,pp.97\u2013110,2011.", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-35", "text": "matics,vol.5,no.2,pp.97\u2013110,2011.\n[11] G. D \u00b4\u0131az, F. A. Gonz \u00b4alez, and E. Romero, \u201cA semi-automatic\nmethod for quantification and classification of erythrocytes\ninfectedwithmalariaparasitesinmicroscopicimages,\u201d Journal\nofBiomedicalInformatics ,vol.42,no .2,pp .296\u2013307 ,2009 .\n[12] F. B. Tek, A. G. Dempster, and I. Kale, \u201cMalaria parasite detec-\ntion in peripheral blood images,\u201d in Proceedings of the 17th\nBritish Machine Vision Conference (BMVC \u201906) ,p p .3 4 7 \u2013 3 5 6 ,\nSeptember2006.ComputationalandMathematicalMethodsinMedicine 13\n[13] S. Halim, T. R. Bretschneider, Y. Li, P. R. Preiser, and C. Kuss,\n\u201cEstimating malaria parasitaemia from blood smear images,\u201d\ninProceedings of the 9th International Conference on Control,\nAutomation,RoboticsandVision(ICARCV\u201906) ,pp .1 \u20136,Singa-\npore,December2006.\n[ 1 4 ]C .d iR u b e r t o ,A .D e m p s t e r ,S .K h a n ,a n dB .J a r r a ,\u201c A n a l y s i s\nof infected blood cell images using morphological operators,\u201d\nImage and Vision Computing ,vol.20,no .2,pp .133\u2013146,2002.\n[15] N.E.Ross,C.J.Pritchard,D.M.Rubin,andA.G.Dus \u00b4e,\u201cAuto-", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-36", "text": "mated image processing method for the diagnosis and classifi-cationofmalariaonthinbloodsmears,\u201d Medical and Biological\nEngineeringandComputing ,vol.44,no .5,pp .427 \u2013436,2006.\n[16] S. W. S. Sio, W. Sun, S. Kumar et al., \u201cMalariaCount: an image\nanalysis-based program for the accurate determination of par-\nasitemia,\u201d Journal of Microbiological Methods ,v o l .68 ,n o .1 ,p p .\n11\u201318,2007.\n[17] Y .Purwar,S.L.Shah,G.Clarke,A.Almugairi,andA.Muehlen-\nbachs, \u201cAutomated and unsupervised detection of malarialparasitesinmicroscopicimages,\u201d MalariaJournal ,vol.10,article\n364,2011.\n[18] S.Kaewkamnerd,A.Intarapanich,M.Pannarat,S.Chaotheing,\nC. Uthaipibull, and S. Tongsima, \u201cDetection and classification\ndeviceformalariaparasitesinthick-bloodfilms,\u201din Proceedings\nof the IEEE 6th International Conference on Intelligent Data\nAcquisitionandAdvancedComputingSystems(IDAACS\u201911) ,vol.\n1,pp.435\u2013438,Prague,CzechRepublic,September2011.\n[19] J.-P.ThiranandB.Macq,\u201cMorphologicalfeatureextractionfor\nthe classification of digital images of cancerous tissues,\u201d IEEE\nTransactionsonBiomedicalEngineering ,vol.43,no.10,pp.1011\u2013\n1020,1996.\n[20] Y.K.LinandK.S.Fu,\u201cAutomaticclassificationofcervicalcells", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-37", "text": "[20] Y.K.LinandK.S.Fu,\u201cAutomaticclassificationofcervicalcells\nusingabinarytreeclassifier,\u201d Pattern Recognition ,vol.16,no .1,\npp .69\u201380,1983.\n[21] P. Viola and M. J. Jones, \u201cRobust real-time face detection,\u201d\nInternationalJournalofComputerVision ,vol.57 ,no .2,pp .137 \u2013\n154,2004.\n[22] M.R \u00a8atsch,S.Romdhani,andT.Vetter,\u201cEfficientfacedetection\nbyacascadedsupportvectormachineusinghaar-likefeatures,\u201dinProceedings of the 26th Annual Symposium of the German\nA s s o c i a t i o nf o rP a t t e r nR e c o g n i t i o n , pp. 62\u201370, Tuebingen,\nGermany,2004.\n[23] R.E.Schapire,Y .Freund,P .Bartlett,andW .S.Lee,\u201cBoostingthe\nmargin: a new explanation for the effectiveness of votingmethods,\u201din Proceedingsofthe14thInternationalConferenceon\nMachineLearning ,1997 .\n[24] V.Vapnik, StatisticalLearningTheory ,Wiley,1998.\n[25] O.Mudanyali,C.Oztoprak,D.T seng,A.Erlinger,andA.Ozcan,\n\u201cDetection of waterborne parasites using field-portable and\ncost-effective lensfree microscopy,\u201d L a bo naC h i p :M i n i a t u r i -\nsationforChemistryandBiology ,vol.10,no.18,pp.2419\u20132423,\n2010.\n[26] C.M.Bishop, PatternRecognitionandMachineLearning ,Sprin-\nger,2006.", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-38", "text": "ger,2006.\n[27] S.TheodoridisandK.Koutroumbas, Pattern Recognition ,A ca-\ndemicPress,SanDiego,Calif,USA,1999.Submit your manuscripts at\nhttp://www.hindawi.com\nStem Cells\nInternational\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014M ED IATO R S\nINFLAM MATIONof\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014Behavioural \nNeurology\nEndocrinologyInternational Journal of\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014Disease Markers\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014BioMed  \nResearch International\nOncologyJournal of\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014Oxidative Medicine and  \nCellular Longevity\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014PPAR Research\nThe Scientific  \nWorld Journal\nHindawi Publishing Corporation \nhttp://www.hindawi.com Volume 2014\nImmunology Research\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014Journal of\nObesityJournal of\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014 Computational and  \nMathematical Methods  \nin Medicine\nOphthalmologyJournal of\nHindawi Publishing Corporation", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}
{"id": "2d87191d9618-39", "text": "Mathematical Methods  \nin Medicine\nOphthalmologyJournal of\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014\nDiabetes ResearchJournal of\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014Research and TreatmentAIDS\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014Gastroenterology \nResearch and Practice\nHindawi Publishing Corporation\nhttp://www.hindawi.com Volume 2014Parkinson\u2019s \nDisease\nEvidence-Based \nComplementary and  \nAlternative Medicine\nVolume 2014Hindawi Publishing Corporation\nhttp://www.hindawi.com", "source": "https://downloads.hindawi.com/journals/cmmm/2015/139681.pdf"}