Field Based Application of AutomatedImage Processing Using Windows Phone  Guided Application in Diagnosis of              ...
INTRODUCTION      Malaria is a serious mosquito-borne disease that has already been eradicatedfrom North America, Europe a...
P. falcifarum or a parasite-specific lactate dehydrogenase detection assays or thedevelopment of nucleic acid amplificatio...
absolute diagnosis is the requisite of effective empirical therapy. Confirmatory diagnosisbefore treatment initiation rece...
open problems will be addressed and a perspective of the future work for realization ofautomated diagnosis of malarial par...
Although fast and simple in concept, RDT performance in practice requires well-trainedoperators that are able to interpret...
able to implement. It depends on good laboratory management to oversee processessuch as documentation, audit cycles, quali...
correctly identify individuals who have a given disease or disorder. For example, acertain test may have proven to be 90 p...
The premise is straightforward. Apply a blood sample to a slide with a dye thatonly malaria parasites can absorb. Using a ...
It must be noted that a human expert will require more time to go through a slideand focus the microscope to observe 30,00...
In addition to the necessity of reducing these variations for the local process, if exchangeof images and training samples...
however, this method also performs global thresholding and is probably negativelyaffected by uneven illumination.        R...
Science. UK: Cambridge University Press; 2005]. Additionally, the reference colorpatches (as proposed by Grana C et.al 200...
steps to provide an effective color correction. However, the method is not directlyapplicable to thick film analysis due t...
A common practice is to estimate average cell size with the peak index of thegranulometry (which can be an area or radius ...
can be followed. In some studies first the stained objects were identified by theirintensity and color properties; then on...
regions marked by these pixels. However, they identified the WBCs, platelets, andschizonts by comparing their size to the ...
"stained pixels are darker/brighter" definition. Using the detected stained pixels asmarkers, they located the objects by ...
contained 12 images. Rao et al 2004 proposed a rule-based scheme (area andhaemozoin existence) to differentiate five life-...
parasites. However, a multi-class joint classification scheme will treat each species andlife-stages as separate and provi...
automated diagnosis of malaria. Despite being very specialized, if the fatality figures areconsidered their results may be...
Another study that this endeavor dwells into is the research done by Minh-TamLe. et. al 2006. The researchers have found o...
smartphone application to address child mortality rates caused by the lack of detectionand availability of treatment for m...
Sensitivity is not the same as the precision or positive predictive value (ratio of truepositives to combined true and fal...
ParasitemiaParasitemia is the quantitative content of parasites in the blood.It is used as ameasurement of parasite load i...
requirements and thus the applicability of the proposed solutions to the problem. Here,these differences are addressed; th...
27 | P a g e
References 1. Korenromp E, Miller J, Nahlen B, Wardlaw T, Young M. Tech rep. World Health     Organization, Geneva; 2005. ...
Upcoming SlideShare
Loading in...5
×

Field Based Application of Automated Image Processing Using Windows Phone Guided Application in Diagnosis of Malaria

1,226
-1

Published on

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
1,226
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
43
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Field Based Application of Automated Image Processing Using Windows Phone Guided Application in Diagnosis of Malaria

  1. 1. Field Based Application of AutomatedImage Processing Using Windows Phone Guided Application in Diagnosis of Malaria A paper Presented to The Faculty of the Graduate School University of the East Ramon Magsaysay Memorial Medical Center In Partial Fulfillment Of the Requirements for the Course Epidemiology and Control of Vector Borne Diseases By: Melvin B. Marzan RN MSc Tropical Medicine II
  2. 2. INTRODUCTION Malaria is a serious mosquito-borne disease that has already been eradicatedfrom North America, Europe and Russia, but still prevalent in Africa, Central and SouthAmerica, the Middle East, the Indian subcontinent, and Southeast Asia including thePhilippines. Experts estimate that one to three million people die from malaria everyyear and that more than 500 million people are affected by it. Malaria is still a major health problem in the Philippines although the number ofreported cases has been decreasing since 1990. The World Health Organization (WHO)figures show that in 1990 more than 86,200 new cases of the disease were reportedcompared to only 37,005 in 2002 and 43,644 in 2003. The improvement can beattributed to the anti-malarial program of the government that has been receivingsupport from international agencies. The common method of malarial diagnosis in third world countries is bydetection of malarial parasite either by morphologic analysis or detection inidentification of antigens products in the blood of the patient by direct microscopy orusing the Rapid Diagnostic Test. When executed properly, 60 to 70% of all adults withmalaria can be identified using the Rapid Diagnostic Test Procedure, followed bymicroscopic examination. However, in areas of endemicity, laboratories are oftenoverloaded with samples for smear examination. It is also a common scenario that mostareas with malaria-affected population do not have laboratory or laboratory personnelto man the contemporary diagnostic procedure. Therefore, there is an urge for a newsimple and rapid diagnostic that should alternatively or totally replace microscopy notcompromising the high specificity and sensitivity. In the past, research was mainlyamplified on the development of either antibody/antigen for histidine-rich protein-2 of2|Page
  3. 3. P. falcifarum or a parasite-specific lactate dehydrogenase detection assays or thedevelopment of nucleic acid amplification reactions. Against this background, the current technology (Lifelens) investigated thefeasibility of a simple windows phone and a specialized lens to detect malarial species insamples of blood. With the (LifeLens) application and a tiny lens attachment, an aidworker with very little training could perform a finger prick blood sample, and take apicture of the blood cells at 350x zoom. The app then utilizes edge detection to count thecells in the image, and identify any discolored cells to count the cells in the image, andidentify any discolored cells that would suggest the presence of Malaria, ultimatelyresulting in immediate diagnosis. Current global malaria control relies on the diagnosis of cases, followed byadequate treatment. The available laboratory methods for the detection of malaria donot fully meet the need in environments with high malaria prevalence’s. The complexityof the procedures would sometimes make the end users reluctant to follow theprescribed protocol, thus causing discrepancies on case finding reports. Abovementioned scenarios would likely cause spontaneous and perennial problems of malariacases in the country. The main aim of the study is to conduct community and field-basedwide scale research to test the efficacy, sensitivity and reliability of the new application.Currently, the innovators have tested the device in purely clinical setting and haveachieved stunning 94.4% level of accuracy. Lifelens diagnosis and treatment also offereda cost effective way of managing malaria, it just cause an average of only $0.56, versus$3.40 using current methods. Empirical therapy of malaria is vital to avoid adverse and virulent effects, tomitigate or totally stop resistance, and to save cost on alternative drugs. Precise and3|Page
  4. 4. absolute diagnosis is the requisite of effective empirical therapy. Confirmatory diagnosisbefore treatment initiation recently regained attention, partly influenced by the spreadof drug resistance and thus the requirement of more expensive drugs unaffordable toresource-poor countries. This research aims to focus on the accuracy of the Smartphoneapplication (Lifelens), the malaria diagnostic that shows potential to have the largestimpact on malaria control today. Thus the research hopes to fill in the gaps regardingthe current diagnosis and the new diagnosis to yield recommendations on the effectiveuse of the Lifelens in community and field settings.PURPOSE The paper discussed to substantially identify the sensitivity and specificity of thenew diagnostic device (Lifelens) in the diagnosis of malaria. The research warrantsaddressing also the practicality and stability of the innovated device. A comparativeanalysis will be performed to find out if the (Lifelens) device could perform betteragainst the existing diagnostic methods use in the malarial infection management suchas the Microscopy and Rapid Diagnostic testing to develop a ground for the extensiveuse of (Lifelens) device in community and filed based settings.OBJECTIVES: This paper scrutinizes a prospect of lifelens device (a windows phone guided malarialdetection device) and image analysis studies aiming at automated and fast diagnosis orscreening of malaria infection in thin blood film smears. In addition, a general patternrecognition framework to perform diagnosis, which includes image acquisition, pre-processing, segmentation, and pattern classification components, will be described . The4|Page
  5. 5. open problems will be addressed and a perspective of the future work for realization ofautomated diagnosis of malarial parasite with lifelens device will be provided. The main objective of the study is to determine the sensitivity and accuracy of thenew and simple diagnostic tool in the diagnosis of malaria in community and field-based settings. the present study hopes to assess the reliability and accuracy of(Lifelens) in a field based setting of the Philippines, where health workers have limitedtraining and compromised knowledge of basic microscopy and support diagnosis andcompares it with the existing Rapid Diagnostic Testing and Microscopy. The studyendures to adhere with the following objectives in the course of study: 1.) Measure and assess the sensitivity and specificity of the device. 2.) probe the practicality, cost effectiveness, and accuracy of the proposed diagnostic tool (Lifelens) and the its abilty to perform the actual diagnostic task in community and field based settingsCURRENT STATE OF MALARIAL DIAGNOSIS/MALARIAL DIAGNOSTICRESEARCH Rapid diagnostic tests (RDTs) are frequently used as an adjunct to microscopy inthe diagnosis of malaria [Wongsrichanalai et.al 2007] and even as a point-of-carediagnostic tool [Wiese L, Bruun B et.al 2006]. In settings where high quality microscopyis not available, the detection of Plasmodium infections is often based on RDTs alone[Chilton D, Malik AN, Armstrong M et. Al 2006]. World Health Organization (WHO)recommends the use of RDTs as part of parasite-based diagnosis and supports the broadimplementation of RDTs for malaria diagnosis in areas where malaria isprevalent [World Health Organization: World Malaria Report 2009. Geneva; 2009].5|Page
  6. 6. Although fast and simple in concept, RDT performance in practice requires well-trainedoperators that are able to interpret results correctly and record them properly. Atpresent, there is no widely accepted way of assessing the quality of RDTs at the end-userlevel and both microscopy and PCR could be used as reference method [Bell D,Wongsrichanalai et.al 2006] The PATH Organization’s 2010 Report reveals that malaria diagnosis,particularly in remote areas lacking laboratory support, frequently relies on the patient’ssymptoms. The first symptoms of malaria (fever, chills, sweats, headaches, musclepains, nausea, and vomiting) are not specific to malaria. While clinical diagnosis isinexpensive and can be effective, clinicians often misdiagnose malarial infection.Misdiagnosis often leads to the unnecessary prescription of malaria medications whichare becoming increasingly expensive as drug resistance grows globally and newmedicines are required for effective treatment. Thus, increasing the accuracy of malariadiagnosis is becoming more important and will continue to be so in the future[http://www.path.org/annual-report/2010/index.php]. Clinical diagnosis is imprecise but remains the basis of therapeutic care for themajority of febrile patients in malaria endemic areas, where laboratory support is oftenout of reach. Scientific quantification or interpretation of the effects of malariamisdiagnosis on the treatment decision, epidemiologic records, or clinical studies hasnot been adequately investigated. Despite an obvious need for improvement, malariadiagnosis is the most neglected area of malaria research, accounting for less than 0.25%($700,000) of the U.S.$323 million investment in research and development in 2004.Establishing and maintaining an accurate and reliable laboratory service is a complex,expensive and technically demanding process, which very few poor countries have been6|Page
  7. 7. able to implement. It depends on good laboratory management to oversee processessuch as documentation, audit cycles, quality assurance and external validation, safetypractices, and supervisory and accountability structures [Hanscheid T. et al2003].Microscopy remains the gold standard in malaria diagnosis, and allows the use ofRapid Diagnostic Test (RDT) only in certain situations[Zhang W, Wang L]. According to Ruiz A, Priotto G. et al 2002 Rapid and accurate diagnosis ofmalaria is not only crucial for patient treatment, but also important for disease control,especially during attempts at elimination, as P. vivax infections are often found at lowparasite densities, and any missed cases of malaria could be a potential source of localtransmission. Microscopic examination of blood films is the most wildly used diagnosticapproach in the field and still remains the gold standard. However, this method islabour-intensive, requires well-trained experts and may result in therapeutic delays.Recently developed lateral flow-based malaria rapid diagnostic tests (RDTs) haveproved useful in P. falciparum-endemic countries, as the sensitivity of RDTs against P.falciparum histidine-rich protein II (PfHRP-II) and P. falciparum lactatedehydrogenase (PfLDH) is high [WHO: Malaria rapid diagnostic test performance,Results of WHO product testing of malaria RDTs (2009)]. In contrast, RDTs for P.vivax are currently not as sensitive as those for P. falciparum, due to the lowparasitaemia and lack of abundantly expressed specific antigens [Notomi T, OkayamaH et. Al 2000]. The UNICEF’s Guideline for Malaria Diagnosis suggest the following criteria forselecting Rapid Diagnostic Test. Accuracy was subdivided into two criteria namelySensitivity and specificity were defined and was used as evaluation criteria to determinethe efficacy of the Rapid Diagnostic Device. Sensitivity means the ability of a test to7|Page
  8. 8. correctly identify individuals who have a given disease or disorder. For example, acertain test may have proven to be 90 per cent sensitive if 100 people known to have acertain disease are tested with that method, the test will correctly identify 90 of those100 cases of disease. The other 10 people who were tested will have the disease, but thetest will fail to detect it. For that 10 per cent, the finding of a "normal" result ismisleading false-positive result. The UNICEF’s Guideline for Malaria Diagnosis suggestthe following criteria for selecting Rapid Diagnostic Test. Sensitivity means the ability ofa test to correctly identify individuals who have a given disease or disorder. Forexample, a certain test may have proven to be 90 per cent sensitive if 100 people knownto have a certain disease are tested with that method, the test will correctly identify 90 ofthose 100 cases of disease. The other 10 people who were tested will have the disease,but the test will fail to detect it. For that 10 per cent, the finding of a "normal" result ismisleading false-positive result. Early diagnosis and prompt treatment of high quality is central to thereduction of malaria. The inability to diagnose malaria correctly and early enough toprevent the case from progressing to severe and complicated conditions poses a bigchallenge to the overall achievement of the vision of a malaria-free Philippines by 2020.More importantly, treatment failure has always been identified as a prominenthindrance to the control of malaria. It is therefore important for health managers andservice providers to be guided on the new directions towards quality malaria diagnosis[Department of Health (Philippines): Manual of Operations Malaria Program (2010)Manila: Philippines].LIFELENS DEVICE8|Page
  9. 9. The premise is straightforward. Apply a blood sample to a slide with a dye thatonly malaria parasites can absorb. Using a specialized lens with 350x magnification,image that slide to get a cellular-level view of blood cells. The teams algorithm thendetects which, if any, cells--and how many--are infected with the malaria parasite.PARAMETERS TO MEASURE THE ACCURACY OF COMPUTER VISIONFOR MICROSCOPIC DIAGNOSIS OF MALARIAImage acquisitionAccording to Wetzel A, Feineigle P, Gilbertson J et. Al 2002 the required number ofimages to capture a 2 cm2 region of specimen at 20× magnification is calculated to benearly 1,300 images using a 1,300 × 1,030 pixel 2/3 inch charge coupled device (CCDsensor) camera. Diagnosis of malaria requires 100× objective magnification(recommended for manual examination), so the number of captured images would be25 times higher. Hence, it roughly corresponds to over 30,000 slide movements, focus,and CCD sensor shutter operations which require a very fast technique. In order toreduce the time requirements, Wetzel et al proposed to capture the images while theslide is continuously moving, which introduced the problem of image blurring.9|Page
  10. 10. It must be noted that a human expert will require more time to go through a slideand focus the microscope to observe 30,000 fields. Hence, the number of fields theexpert would examine is usually smaller. In the WHO malaria microscopy tutorial,examination of only 100 fields is recommended before giving a negative decision.Additionally, in thick films, if a parasite is observed in a field, 100 more fields (or 200WBCs, 0.025 μl of blood) would be sufficient to calculate the parasitaemia. Since it isless sensitive, routine examination of thin blood films is not recommended for thepositive/negative type of diagnosis. However, if parasites are found, examination of 50fields (average 200 per field yields 10,000 RBCs in total) would be sufficient to calculatethe parasitaemia in thin films. Thus, the speed requirements of the image acquisitionsystem can be relatively easy to achieve. In addition, recently emerging fast focusingsolutions and dedicated commercial slide scanning machines (e.g. US Patent No.563437 filed on 2000-05-03) are promising to solve this important practical obstacle.Image variationsAn image acquired from a stained blood sample (thick or thin) using a conventionallight microscope can have several conditions which may affect the observed colors of thecells, plasma (background), and stained objects. These conditions may be due to themicroscope components such as: the different color characteristics of the light source,intensity adjustments, or color filters. They may be due to the use of different camerasor different settings in the same camera: exposure, aperture diagram, or white balancesettings. The differences in specimen preparation can cause variations as often as theimaging conditions [Fujii K, Yamaguchi M et. 2002]. For example, acidity (pH) of thestain solution can seriously affect the appearance of the parasites. Addressing thesevariations can simplify the main analysis and contribute to the robustness of the system.10 | P a g e
  11. 11. In addition to the necessity of reducing these variations for the local process, if exchangeof images and training samples could be made possible, then the different diagnosislaboratories which may employ the system in the future may benefit from a uniformdiagnosis expertise.Illumination and thresholding Most microscopes are equipped with (calibration) components to provideuniform or relatively uniform illumination. A common illumination calibration standardis Kohler Illumination named after its inventor August Kohler. In this method,transmitted illumination from the light source is aligned and focused for a parallel anduniform illumination. This is often neglected by microscopists since the human visionsystem is adaptive to local illumination changes, however for an image analysisalgorithm variations can cause serious problems. Uneven illumination can be simply dealt with by acquiring a separate image ofillumination to subtract from images later. However, for a particular test image comingfrom an external source, the imaging system may not be accessible to record a referenceimage of illumination. An alternative method is to filter the images to remove thevariation in the illumination. In the case of a smooth varying illumination, as in mostmicroscope images, a filtering operation may reduce the potential effects. This may beperformed by applying a Gaussian filter or morphological image filtering method. Halim et al 2006 proposed to correct uneven illumination by calculatinggradients in the polar coordinates (r, θ coordinate system) of the background imagewhich was calculated by simple thresholding. However, in some cases the illuminationcan be excessively uneven and hinder a thresholding operation. Ross et al employedOtsus thresholding method to obtain a binary foreground-background representation;11 | P a g e
  12. 12. however, this method also performs global thresholding and is probably negativelyaffected by uneven illumination. Rao et al. 2004 proposed the use of mathematical morphology to produceforeground binary masks in the presence of uneven illumination. The proposed methodperforms an initial rough thresholding to separate foreground and backgroundhistograms from which two separate threshold values are found. In the final step, themorphological double threshold operation is employed to obtain a refined binaryforeground mask. However, it was shown in that due to the final global thresholdoperation even this method is not immune to uneven illumination, and that theillumination must be corrected prior to any global (thresholding) operation.Color The different Plasmodium species are distinguishable from each other andregular blood components and artefacts by their characteristic shapes (morphology) andcolor properties [WHO: Basic malaria microscopy Part I. Learners Guide. WorldHealth Organization; 1991]. If the color-based properties of the images are used thencolor variations must be addressed. The difference with microscope imaging is that calculations based on theLambertian surface model and use of the reference color charts are not appropriatebecause the sensor (or human eye) does not receive the light reflecting from a surface.The light reaching the sensor is the attenuated light which is left after the objects (i.e.specimens) absorption. In fact, image formation of the stained slides with lightmicroscopes are more appropriately modelled with the "Beer-Lambert Law" whichstates that there is a linear relationship between the concentration, thickness ofilluminated media, and the "absorbance" [Lee HC: Introduction to Color Imaging12 | P a g e
  13. 13. Science. UK: Cambridge University Press; 2005]. Additionally, the reference colorpatches (as proposed by Grana C et.al 2005 for other medical imaging applications,e.g.), are not practical for microscopes. Even though it was possible to manufacturethem; there is still the human factor in preparation of the blood film slides which resultsin non-standard and non-homogeneous staining concentrations and appearances. The problem of non-standard preparation of the blood film slides (specimen) wasaddressed by Abe T, Yamaguchi et.al 2004. To correct under/over staining conditionsof the slide, they obtained the spectral transmittance by a multispectral camera (acamera equipped with different filters to capture the spectral reflectance on separatebands). They mathematically modelled the relation between the transmittance and theamount of stain (dye) for each pixel using the Beer-Lambert Law and Wiener inverseestimation. The research done by Abe T, Yamaguchi et.al 2004 is an important studyproviding a mathematical model of the staining concentration-transmittance relation,which enables digital correction of non-ideal stain concentrations. However, thevariations due to the different camera parameters and light sources were not addressedwhich leaves the imaging side of the problem fuzzy. Nevertheless, the malaria diagnosissystem may not have the luxury of adding the cost of a multispectral camera; it is notpractical to capture many different bands of the same field to estimate the amount ofdye. In the study done by Ohyanma et.al 2002 the authors proposed a practicalmethod which exploits the special characteristics of the peripheral thin blood filmimages that are easily separable into the foreground and background regions. Afterseparation, the method employs the simple grey world assumption in two consecutive13 | P a g e
  14. 14. steps to provide an effective color correction. However, the method is not directlyapplicable to thick film analysis due to the assumption of an expected foreground scene.Scale and granulometry According to Hughes-Jones N. et al 2004 in healthy human peripheral blood, theaverage diameter of an RBC and platelet is between 6–8 μm and 2–3 μm, respectively.WBC size can vary between 8–20 μm depending on the type . The CCD pixel resolutionand magnification (i.e. field of view) can be used to calculate expected sizes of the bloodcells that are present in the image. Moreover, this information can be used to calculatethe image pixel scale in physical units. However, the magnification information may notbe accessible or the imaging set-up may not be present. Additionally, there are someconditions (e.g. anaemia) which result in abnormal cell shapes and sizes . Almost none of the methods which aim at diagnosis of malaria or relatedprocessing tasks are concerned about the actual physical scale of the objects in theprocessed images, but the size of the cells in the image plane to enable scale-independent processing since the cell size information used as a parameter in manyalgorithms. The granulometry of mathematical morphology (pattern spectrum) can providethe size distribution of an input image. It is computed via a family of openings whichhave increasing, anti-extensive, idempotence properties. Though the definition ofgranulometry does not suggest any special type of opening operation, in practice it isusually implemented via a set of increasing-width structuring elements of a fixedpattern (e.g. square, disk, and hexagon).Average cell size estimation14 | P a g e
  15. 15. A common practice is to estimate average cell size with the peak index of thegranulometry (which can be an area or radius index). This assumes that the thin bloodfilm image is covered by resolvable individual RBCs of similar size. However, the RBCsize variation in normal blood and the disorders which cause abnormal RBC sizes areneglected. In addition, the thickness of the thin film varies through a slide and thisresults in varying focus depths, which can also change the calculated average cell area.Existing malaria diagnosis methods concentrate only on using size or areagranulometries. However, the granulometry concept has more potential to explore,which may be applicable to blood film image analysis.Breen and Jones extended thedefinition of granulometry to be calculated with any set of attribute openings or non-increasing opening-like operations: thinnings. Urbach et al proposed animplementation of shape pattern spectrum which was later extended to the calculationof 2D granulometries (Shape × Area) in and to the vector granulometries in.Segmentation Probably one of the most common shared tasks in image analysis systems issegmentation. Segmentation aims to partition the image plane into meaningful regions.The definition of the meaningful regions and partitioning method is usually applicationspecific. For example, the methods can be aimed at separating foreground-background,moving-still regions or objects with specific properties from the scene. Thesegmentation strategy can be a hierarchical partitioning that operates deductively todefine first a higher level of object plane, then the objects, and then sub-objectcomponents. The inductive approaches define first the objects of interest with a specificproperty then perform higher levels of partitioning(s) if necessary. In order to localizehighlighted (stained) objects, either inductive or deductive segmentation approaches15 | P a g e
  16. 16. can be followed. In some studies first the stained objects were identified by theirintensity and color properties; then only the RBC regions containing the stained objectswere segmented from the image. On the other hand, in some studies, by Rao et.al 2004a deductive strategy was followed: the image was first separated into foreground andbackground regions; then foreground regions were segmented to obtain individual RBCregions; then these were further analyzed to detect the presence of staining. The globalsegmentation procedure is applied usually if a deductive approach is proposed.Stained pixels and objectsThe staining process highlights the parasites, platelets, WBCs, and artefacts in a thinblood (peripheral) film image. In order to analyze the highlighted bodies it is essential toidentify the pixels and thence locate the object regions. However, it must be noted thatother blood parasites and some disorders of blood, e.g. iron deficiency are alsohighlighted by the Giemsa-stain. Some methods of the literature name and describe this step as "ParasiteDetection" (or parasite extraction). This results in over-simplistic solutions which arenot applicable to diagnosis of malaria, because diagnosis must be performed on actualperipheral blood specimens of the patients which are certain to contain other stainedbodies: WBCs, platelets and artefacts and may be infected by other parasites or mayhave other disorders (e.g. iron deficiency). This may be related to the use of in vitrosamples as for the experimental data. Usually in vitro culture images consist of samplesgrown in a laboratory environment. Hence, they are cleaner of artefacts and do notcontain platelets or WBCs. Di Ruberto et al 2001 employed morphological regional extrema to detect (i.e.marked) the stained pixels, then used morphological opening to extract the object16 | P a g e
  17. 17. regions marked by these pixels. However, they identified the WBCs, platelets, andschizonts by comparing their size to the average cell size obtained from granulometryand exclude these from further processing. Hence, their method can be regarded asaddressing the detection issue. However, detection of stained pixels with regionalextrema is error prone because it will locate some pixels even if the image does notcontain any stained pixels. Moreover, eliminating WBCs and platelets with respect tothe average area value can eliminate some parasite species which enlarge the RBCs thatthey occupy. For example, Plasmodium vivax infected cells can enlarge up to 2.5 times.Ross et al 2006 used a similar approach: they have used a two level thresholding (globaland local) to locate stained pixels, then used morphological opening to recover theobject binary masks. Both of the methods rely on opening and disk shaped structuringelements which creates problems because the cells are rarely perfect and flat circles. Rao et al 2004 used thresholding to detect stained pixels, however they pre-processed the images to remove a global bias color value that is caused by staining,which is to prevent false pixel detections if the image do not contain any stained pixels.Since they use global segmentation to locate individual RBCs, the stained objects aredefined by the regions which contain stained pixels. As stated in the previous sectionglobal segmentation is error prone, unless examined fields are limited to the lightlyconcentrated fields. In addition, it must be noted that employing a thresholdingoperation to detect stained pixels assumes an ordered relation between stained and un-stained pixels, e.g. "stained pixels are darker than others". Tek FB et. Al 2006 proposed to detect stained pixels according to their likelihoodwhere a pixels red-green-blue color triple was used as the features and stained and un-stained classes were modelled using 3-d histograms. This removes the limitation of the17 | P a g e
  18. 18. "stained pixels are darker/brighter" definition. Using the detected stained pixels asmarkers, they located the objects by using morphological area top-hats andreconstruction. This approach prevented over-segmenting of stained bodies, whichcould be caused by employing global segmentation based on area heuristics. Detection of stained pixels is not a very complex problem especially with the useof color correction algorithms. However, as pointed out in Tek FB et al 2007, one of thebiggest problems of thin blood film analysis is to locate the stained objects and definetheir boundaries, because the stained pixels which are used as markers may be due to avariety of objects, e.g. to an artefact which can be any size or shape.Classification There are only few studies which propose a classification procedure Ross et.al2006 to differentiate between parasites and other stained components or artefacts. Themethod described by Halim S et. al 2006 also proposes a classification to differentiatebetween a healthy RBC and an "infected" RBC. However, from the diagnosis point ofview the essential task is to identify parasites in the presence of other stained structures,artefacts, and then finally identify the species. As in Di Rubertos research in 2001, theapproach to the classification task in a recent work also was also limited to detectionwhite blood cells and gametocytes by area information, for the purpose of excludingthese from parasitaemia calculation. However, although they do not address the parasite/non-parasite differentiation,some automated diagnosis of malaria studies rather focused on the life-cycle stageclassification. Di Ruberto et al 2001 proposed to use the criteria of circularity (measuredby the number of morphological skeleton endpoints and color histogram to classify thelife-stages into two categories: immature and mature trophozoites. Their test set18 | P a g e
  19. 19. contained 12 images. Rao et al 2004 proposed a rule-based scheme (area andhaemozoin existence) to differentiate five life-stages. They experimented on a set ofPlasmodium falciparum in vitro samples which contain immature-mature trophozoite,early-mature schizont but no gametocyte class or other types of stained object. Ross et al 2006 proposed a consecutive (detection-species recognition) two-stages classification for the problem. They proposed to use two different sets of featuresfor parasite detection and species recognition. The initial feature sets were comprised ofmany color- and geometry-based features. For example, they have used averageintensity, peak intensity, skewness, kurtosis and similar abstract calculations from thered green blue channels together with the same calculations from the hue-saturation-intensity channel images. For geometrical features, they have identified roundness ratio,bending energy, and size information, i.e. area, in their feature set. For parasitedetection and following species recognition tasks, the initial feature sets were comprisedof 75 and 117 features, respectively. For the species recognition task the SE-PPV resultswere: P. falciparum 57%–81%, P. vivax 64%–54%, P. ovale 85%–56%, P. malariae29%–28%. The life-stage recognition problem was not investigated. Their experimentsused a training set comprised of 350 images containing 950 objects and in the similartest set. Nevertheless, the joint classification scheme, removing the necessity for a binarydetection (parasite/non-parasites classification), may improve the expandability andscalability of a diagnosis system by preventing a narrow reference to "parasite" and"non-parasite" classes. For example, if restricted to perform a binary detection, amalaria diagnosis system will have a different notion of "parasites" than a diagnosissystem for Babesiosis or Trypanosomiasis which are examples of other peripheral blood19 | P a g e
  20. 20. parasites. However, a multi-class joint classification scheme will treat each species andlife-stages as separate and provide other parasites or conditions to be handled by thesystem. This should be supported by the use of generalized features instead of theoptimized features.CONCEPTUAL FRAMEWORK: Malaria is a serious infectious disease caused by a peripheral blood parasite of thegenus Plasmodium. According to the World Health Organization (WHO), it causes morethan 1 million deaths arising from approximately 300–500 million infections everyyear . Although there are newer techniques , manual microscopy for the examination ofblood smears (invented in the late 19th century), is currently "the gold standard" formalaria diagnosis. Diagnosis using a microscope requires special training andconsiderable expertise . It has been shown in several field studies that manualmicroscopy is not a reliable screening method when performed by non-experts due tolack of training especially in the rural areas where malaria is endemic . An automatedsystem aims at performing this task without human intervention and to provide anobjective, reliable, and efficient tool to do so. The study patterns it framework from the paradigms adopted by F.Boray et al,2009. As describe by F. Boray et.al 2009 an automated diagnosis system can bedesigned by understanding the diagnostic expertise and representing it by specificallytailored image processing, analysis and pattern recognition algorithms. Although it isnot a popular research topic, a noticeable number of vision studies directly address the20 | P a g e
  21. 21. automated diagnosis of malaria. Despite being very specialized, if the fatality figures areconsidered their results may be considered more important than some other popularcomputer vision applications. From the computer vision point of view, diagnosis of malaria is a multi-partproblem. A complete system must be equipped with functions to perform: imageacquisition, pre-processing, segmentation (candidate object localization), andclassification tasks. Hence, the complete diagnosis system also requires some functionssuch as a prototype of microscope slide positioning, an automated, fast, and reliablefocus, and image acquisition. Studies concerning image acquisition are examined insection Image acquisition. Usually, the acquired images from a microscope have severalvariations which may affect the process. These are usually addressed by pre-processingfunctions. An important step in automated analysis is to obtain/locate possibly infectedcells (i.e. candidates) which are the stained objects in the images. In order to perform a better diagnosis on peripheral blood samples, the systemmust be capable of differentiating between malarial parasites, artefacts, and healthyblood components. The majority of existing malaria-related image analysis studies donot address this requirement. This results in the over-simplified solutions, which are notapplicable to diagnosis directly. Existing works on malaria commonly use mathematicalmorphology for image processing since it suits well to the analysis of blob-like objectssuch as blood cells. On the other hand, to differentiate between observed patternsstatistical learning based approaches are very popular.21 | P a g e
  22. 22. Another study that this endeavor dwells into is the research done by Minh-TamLe. et. al 2006. The researchers have found out that novel automatic image processingapproach for determining malarial parasitemia in thin blood smear images can bepresented presented by. Firstly, the nucleated components (including parasites andleukocytes) can be identified using adaptable spectral information. In an independentstep, solid matters, i.e. cells and parasites, can be isolated from the background, bycomparing the input image with an image of an empty field of view. The range oferythrocyte sizes is then determined by examining user inputs of isolated erythrocyteregions. Leukocytes and malarial gametocytes (if present) can be detected by size andremoved accordingly. Reducing the problem of erythrocyte segmentation to a peakselection problem in a transformed image space, the next stage identifies the positionsof individual erythrocytes by finding regional maxima with area-suppression. Finally,the derived parasite and erythrocyte maps are overlaid and assessed concurrently todetermine the parasitemia. The gaps presented by F.Boray et al, 2009 and Minh-Tam Le. et. al 2006 hasalready been resolved on the algorithms that was prepared by the team who createdlifelens. The app then utilizes edge detection to count the cells in the image, and identifyany discolored cells that would suggest the presence of Malaria, ultimately resulting inan immediate diagnosis. Lifelens devices are equipped with proprietary image analysisalgorithm written in .NET with Visual Studio. The software is built in Visual Studio for aWindows Phone 7 using Microsoft Silverlight. Diagnosis is conducted using proprietarycomputer vision algorithms, written in C#, which can detect the presence of a malarialparasite within a patient’s blood cells. Lifelens introduces an innovative point-of-care22 | P a g e
  23. 23. smartphone application to address child mortality rates caused by the lack of detectionand availability of treatment for malaria. The solution has immense potential to reducethe cost of diagnosis and enable children around the world to be treated with the currentamount of funding. The following conceptual paradigm has been deduced from an extensive source ofliterature study: USE OF LIFELENS Better Diagnosis leading DEVICE FOR to better management of DIAGNOSIS OF Malarial Infection in the MALARIA Community Figure 1.1DEFINITION OF TERMSSensitivitySensitivity relates to the tests ability to identify positive results.Again, consider the example of the medical test used to identify a disease. The sensitivityof a test is the proportion of people who have the disease who test positive for it. Thiscan also be written as:If a test has high sensitivity then a negative result would suggest the absence of disease..23 | P a g e
  24. 24. Sensitivity is not the same as the precision or positive predictive value (ratio of truepositives to combined true and false positives), which is as much a statement about theproportion of actual positives in the population being tested as it is about the test.The calculation of sensitivity does not take into account indeterminate test results. If atest cannot be repeated, the options are to exclude indeterminate samples from analysis(but the number of exclusions should be stated when quoting sensitivity), or,alternatively, indeterminate samples can be treated as false negatives (which gives theworst-case value for sensitivity and may therefore underestimate it).SpecificitySpecificity relates to the ability of the test to identify negative results.Consider the example of the medical test used to identify a disease. The specificity of atest is defined as the proportion of patients who do not have the disease who will testnegative for it. This can also be written as:If a test has high specificity, a positive result from the test means a high probability ofthe presence of disease.From a theoretical point of view, a bogus test kit which always indicates negative,regardless of the disease status of the patient, will achieve 100% specificity. Thereforethe specificity alone cannot be used to determine whether a test is useful in practice.24 | P a g e
  25. 25. ParasitemiaParasitemia is the quantitative content of parasites in the blood.It is used as ameasurement of parasite load in the organism and an indication of the degree of anactive parasitic infection. Systematic measurement of parasitemia is important in manyphases of the assessment of disease, such as in diagnosis and in the follow-up of therapy,particularly in the chronic phase, when cure depends on ascertaining a parasitemia ofzero. The methods to be used for quantifying parasitemia depend on the parasiticspecies and its life cycle. For instance, in malaria, the number of plasmodia can becounted using an optical microscope, on a special thick film (for low parasitemias) orthin film blood smear (for high parasitemias).SIGNIFICANCE: Accurate and prompt diagnosis is a requisite for successful management ofmalarial infection in the areas of endemecity. This papers hopes to address the urgencyand need for the better diagnosis of malarial infection in the periphery by making aresearch ground on establishing the sensitivity and specificity of the innovativediagnostic device (Lifelens). The finding of the study could serve as a baseline data onthe feasibility of integrating the (Lifelens) diagnostic tool on the Malarial Infectionmanagement since microscopy and rapid diagnostic tool shares limitations and numberof disadvantages. Filling in the gaps of the current problem could then be tackled. Thisstudy also provides an overview of (Lifelens) for malaria diagnosis and intends to fill agap in this area by doing so. There are some different interpretations of the25 | P a g e
  26. 26. requirements and thus the applicability of the proposed solutions to the problem. Here,these differences are addressed; the practicality and accuracy of the proposed solutionsand their applicability to perform the actual, diagnosis task are questioned.26 | P a g e
  27. 27. 27 | P a g e
  28. 28. References 1. Korenromp E, Miller J, Nahlen B, Wardlaw T, Young M. Tech rep. World Health Organization, Geneva; 2005. World Malaria Report 2005. 2. Hanscheid T. Current strategies to avoid misdiagnosis of malaria. Clin Microbiol Infect.2003;9:497–504. doi: 10.1046/j.1469-0691.2003.00640.x. [PubMed] [Cross Ref] 3. WHO Basic malaria microscopy Part I Learners Guide. World Health Organization; 1991. 4. Kettelhut MM, Chiodini PL, Edwards H, Moody A. External quality assessment schemes raise standards: evidence from the UKNEQAS parasitology subschemes. J Clin Pathol. 2003;56:927–932. doi: 10.1136/jcp.56.12.927. [PMC free article] [PubMed] [Cross Ref] 5. Coleman RE, Maneechai N, Rachaphaew N, Kumpitak C, Miller R, Soyseng V, Thimasarn K, Sattabongkot J. Comparison of field and expert laboratory microscopy for active surveillance for asymptomatic Plasmodium falciparum and Plasmodium vivax in Western Thailand. Am J Trop Med Hyg. 2002;67:141–144. [PubMed] 6. Bates I, Bekoe V, Asamoa-Adu A. Improving the accuracy of malaria-related laboratory tests in Ghana. Malar J. 2004;3:38. doi: 10.1186/1475-2875-3-38. [PMC free article] [PubMed][Cross Ref] 7. Mitiku K, Mengistu G, Gelaw B. The reliability of blood film examination for malaria at the peripheral health unit. Ethiopian J of Health Dev. 2003;17:197–204. 8. Rao KNRM. PhD thesis. University of Westminster; 2004. Application of mathematical morphology to biomedical image processing. 9. Rao KNRM, Dempster AG, Jarra B, Khan S. Automatic scanning of malaria infected blood slide images using mathematical morphology. Proc IEE Semin Med Appl of Signal Process, London, UK. 2002. 10. Di Ruberto C, Dempster A, Khan S, Jarra B. Analysis of infected blood cell images using morphological operators. Image and Vis Comput. 2002;20:133–146. doi: 10.1016/S0262-8856(01)00092-0. [Cross Ref] 11. Di Ruberto C, Dempster AG, Khan S, Jarra B. Proc Int Workshop on Visual Form. Capri, Italy; 2001. Morphological image processing for evaluating malaria disease. 12. Tek FB, Dempster A, Kale I. Proc Med Image Underst and Anal Conf. Manchester, UK; 2006. Malaria parasite detection in peripheral blood images.28 | P a g e

×