[ieee 2012 international conference on computing, communication and applications (iccca) - dindigul,...

Identify Malaria Parasite Using Pattern Recognition Technique

Isha Suwalka

Geetanjali Institute of Technical Studies

Dabok, Udaipur, India [email protected]

Ankit Sanadhya



Anirudh Mathur



Mahendra S Chouhan



Abstract— Moving with the technological advancement in the field of bio imaging processing this paper presents an innovative digital technique detecting parasitic protozoa of the genus Plasmodium. Traditionally, in clinical laboratories the direct microscopic prediction of the parasite on the thick and/or thin blood smears are used for the diagnosis of malaria. It involves collection of a blood smear, its staining with Romanowsky stains and examination of the Red Blood Cells for intracellular malarial parasites. In blood samples, if the red corpuscles of vertebrates are infected by malarial parasites, they will have a specific shape which can identify their presence. Recent research has suggested that the shape of the affected red blood cells can be detected using the 2D moments of the image of the infected cell. In this paper an algorithm is implemented to identify the type of parasite through their color and shape. This technique can detect the existence of malaria parasite within seconds and thus can replace the conventional methods of detection of malaria in bio-medical applications and medical science. The proposed method can automatically diagnose the parasite’s presence using the algorithm mentioned which consumes less time and man power as compare to conventional methods.

Keywords-pattern recognition; malaria; parasite.

I. INTRODUCTION

The protozoan parasites transmitted by the Anopheles mosquito from one person to another causes Malaria. In humans, the parasites (called sporozoites) travel to the liver, where they mature and release another form, the merozoites [1]. The infection is caused by minute parasitic protozoa of the genus Plasmodium, which infect human liver cells first, then the red cells, and then the insect hosts alternatively. The detection techniques today include manual laboratory identification by visual scrutiny of the infected blood slides. A false interpretation in identification due to human indolence may lead to other secure complications. In some cases, due to late and improper diagnosis parasites develop resistance to some antibiotics which led to difficulty in controlling the rate of infection and spread of this disease. Thus, proper care is required while diagnosing for the presence of malaria in the blood sample. Generally, there are three different kinds of blood cells- RBC, WBC and blood platelets [2]. Their distinct dimensions and color distinguish them from one another. The red corpuscles of vertebrates are infected by malarial parasites

when they enter the blood stream. The malarial parasite exists in a variety of different forms, which have been successfully adapted to different cellular environments, in both the vertebrate host and the mosquito vector. The parasite develops in a highly regulated manner through distinct cycles in the vertebrate host. During the life cycle of Plasmodium, it has stages of development and growth where it exhibits several shapes. Since they exist in the form of slide, the blood has to be analyzed in image format, in order to diagnose protozoan presence for recognition. The image of the cell surface resolves the Plasmodium in a range of 6-8�m with nearly spherical, cylindrical and elliptical shapes over the cell surface [3].

The aim of this paper is to present a method to detect the

parasites using image processing by evaluating the curved

structure of the parasite. This approach makes the detection

more accurate as compared to prediction done using

conventional manual process.

II. MALARIA PARASITES

The blood sample taken from the infected person may include any type of parasite. There are 4 types of malaria parasite that infect humans: Plasmodium vivax, Plasmodium malariae, Plasmodium ovale, and Plasmodium falciparum [4].

A. Plasmodium Vivax

Plasmodium vivax is a protozoal parasite and a human pathogen and can be of different forms such as- Rings, Trophozoites, Schizonts and gametocytes. Vivax rings have large chromatin dots and cytoplasm can become ameboid as they develop, size 2.5 µm. Trophozoites have compact cytoplasm and a large chromatin dot. Vivax schizonts are large, have 12 to 24 merozoites, yellowish-brown, coalesced pigment, and may fill the RBC, size 9-10 µm. P. vivax gametocytes are round to oval with scattered brown pigment and may almost fill the RBC. A slide containing Plasmodium vivax is shown in figure 1.

Figure. 1 Plasmodium Vivax

B. Plasmodium Malariae

Plasmodium malariae is a parasitic protozoan that causes malaria in humans. It is present in two forms- Rings and Trophozoites. Rings have sturdy cytoplasm and a large chromatin dot, size 6.5-7 µm and the Trophozoites has compact cytoplasm and a large chromatin dot. A slide containing Plasmodium vivax is shown in figure 2.

Figure.2 Plasmodium Malariae

C. Plasmodium Ovale

Plasmodium Ovale is a species of parasitic protozoa that causes tertian malaria in humans. It is also present in four forms: Rings, Trophozoites, Schizonts and gametocytes in which rings have sturdy cytoplasm and large chromatin dots 6.2 µm. Trophozoites have sturdy cytoplasm, large chromatin dots, and can be compact to slightly irregular 2.5 µm. The schizonts have 6 to 14 merozoites with large nuclei, clustered around a mass of dark-brown pigment. The gametocytes are round to oval and may almost fill the red blood cells. Pigment is brown and coarser in comparison to P. vivax. A slide containing Plasmodium vivax is shown in figure 3.

Figure. 3 Plasmodium Ovale

D. Plasmodium Falcipuram

Plasmodium falciparum is known to be the most deadly form of the plasmodium parasite. It is also present in the above four forms. P. falciparum rings have delicate cytoplasm and one or two small chromatin dots. Trophozoites are rarely seen in peripheral blood smears. Older, ring stage parasites are re-ferred to as trophozoites. The cytoplasm of mature

trophozoites tends to be denser than in younger rings. Schizonts are seldom seen in peripheral blood. Gametocytes are crescent or sausage shaped. The chromatin is in a single mass (macrogamete) or diffuse (microgamete). A slide containing Plasmodium vivax is shown in figure 4.

Figure. 4 Plasmodium Falciparum

III. CURVED SHAPE ANALYSIS

The diagnosis of malaria is confirmed by blood tests which can be microscopic and non-microscopic [6]. This test involves the use of digital microscope for obtaining the stained image of the blood sample. This image is processed using two algorithms of digital image processing. The first algorithm is used to check the intensity level of image and further finds whether the slide is positive or negative with respect to malarial parasite. The second algorithm is used to find the respective shapes in the image and finds type of the parasite. Thus, detection of malaria involves color intensity test and shape detection test.

A. Color Intensity Test

In this technique, the RGB image is converted into a black and white image first. The threshold is set such that luminance values is lower than the threshold value on gray scale and will be considered as dark while the rest will be considered as white. Now this image is converted into a binary image in which the dark pixels are represented as zero while the white pixels are represented as 1.

Algorithm rgb2gray (I = rgb2gray(RGB)) converts RGB values to grayscale values by forming a weighted sum of the R, G, and B components:

I = 0.2989 * R + 0.5870 * G + 0.1140 * B (1)

If it is found positive, then it will be taken for further processing. This test decides presence of parasite.

B. Shape Detection Test

To extract a particular shape from the image, its edges are determined. This value is determined with respect to its thresholds. These edges are determined with locations of discontinuity in grey –level, color, texture, etc. Now a range is determined for various shapes. The criteria to identify the shape is based on equations of perimeter and area of different geometric structures such as for round shaped blood cells it has to satisfy following condition:-

(perimeter)2/4*pi* area = 1 (2)

The shape is extracted according to the range in which it lies.�Each pixel location, designated by the coordinates, (xi, yj), contains the gray-scale value within the image at that point

[9]. The values are on a scale of 0 to 255, whereby 0 corresponds to white and 255 corresponds to black. The gray-scale value at the lattice point, (xi, yj), is designated by f(xi, yj), gray-scale values between neighboring pixel locations. This will be determined by introducing the partial derivative formulas,

(3)

(4) The distance between pixel locations is normalized to be 1, so all of the increments in the partial derivative formulae will be equal to one. This then gives

(5)

(6) The first derivative assumes a local maximum at an edge.

For a continuous image f(x,y), where x and y are the row and column coordinates respectively, we typically consider the two directional derivatives �xf(x,y) and �yf(x,y) [10]. The gradient magnitude is defined by

(7)

and the gradient orientation is given by

(8)

Local maxima of the gradient magnitude identify edges in f(x,y). When the first derivative achieves a maximum, the second derivative is zero [11]. For this reason, an alternative edge-detection strategy is to locate zeros of the second derivatives of f(x,y). The differential operator used in these so-called zero-crossing edge detectors is the Laplacian

(9) The remaining is only the contour of the image. The large

areas with black color are all eliminated, thus a significant reduce in the size of image counted in bits is achieved. The data compression and image data size decrease are important aspects of edge detection operations. This produces a monochrome image with its edge detected.

C. Matching Criteria

This is basically comparing the exact copy of pattern of interest with the produced image. This is basically achieved with correlation of exact pattern and achieved pattern. Let h be the exact pattern and f is the processed image then for good and matching optimality criteria should be satisfied describing match between f and h located at the position b(u,v):

(10)

(11)

(12)

For all (i, j) � V, the correlation between h and f can be determined by taking first the Fourier transform of both the image say F and complex conjugate of H* and then applying the inverse transform [12]. Now values of boundaries are matched with stored data and test revel the type of parasite.

IV. METHODOLOGY

The stained image includes only WBC as they are only affected by parasite, RBCs are removed. The original circular shape of WBC is changed when attacked by parasite. The color intensity of this shape also changes. So the parasite can be easily detected by its intensity and shape. Figure 5 illustrates the algorithm used to accomplish the above process.

Read slide image RGB to Gray Set Threshold=0.3010

Trace the shape of parasite Convert into Binary Image

Matrix of Image Plot Boundaries Determine Shape

BLOB analysis image

Figure. 5 Algorithm chart

V. PATTERN RECOGNITION

For the detection of the type of malaria parasite, ranges are defined in the algorithm so that program can check that which type of parasite it is. The ranges for different parasites are defined according to the value of the area of the object in a stained image. Table 1 summarizes the ranges for different parasites.

TABLE I. RANGE OF DIFFERENT PARASITES [4]

Type of Parasite Range

P. malariae 6 to 7

P. ovale 0.60 to 0.75

P. vivax 1.2 to 2

P. falciparum 0.40 to 0.60

VI. RESULT

The experiment of malaria parasite detection when done by applying the algorithm on a simple PC, show the window as shown in figure 6. The window contains the buttons which facilitates to view the original stained image of slide, diagnose

positive or negative malaria and the type of parasite causing the malaria.

Figure. 6 Original slide image

Diagnosis of malaria is done by applying the algorithm employing the curved shape analysis in the blood sample which will determine whether infected or not. A sample window of blood sample in which parasite is detected is shown figure 7.

Figure. 7 Parasite detected

The end step of the diagnosis is to determine the type of parasite. This is done by pressing the “type of parasite”, which provides the parasite name by comparing the image sample with the stored samples of different parasites. Figure 8 illustrates the result with the type of parasite causing the malaria.

Figure. 8 Structure and type of parasite detected

VII. CONCLUSION AND FUTURE SCOPE

Diagnosis of malaria involves identification of malaria parasite or its antigens/products in the blood of the patient. In reality, the efficiency of the diagnosis is subjected to many factors such as different forms of the four malaria species; the different stages of erythrocytic schizogony; the endemicity of different species; the population movements; the inter-relation between the levels of transmission, immunity, parasitemia, and the symptoms; the problems of recurrent malaria, drug resistance, persisting viable or non-viable parasitemia, and sequestration of the parasites in the deeper tissues. Further the use of chemoprophylaxis or even presumptive treatment on the basis of clinical diagnosis can all have a bearing on the identification and interpretation of malaria parasitemia on a diagnostic test.

The algorithm presented in this paper is limited to the basic

diagnosis of malaria with the types of parasite causing the

malaria. This technique is able to detect the existence of

malaria parasite within seconds and thus can replace the

conventional methods of detection of malaria in bio-medical

applications. Further a user friendly environment is also

provided with the help of simple GUI which produces results

consuming lesser time and man power as compare to

conventional methods.

The application of this algorithm can be extended to be

used for the detection of other objects like various other

parasites and viruses. Errors in results can be reduced to a

great extent by using this project.

ACKNOWLEDGMENT

Authors give special acknowledgement to 1. Dr. Medha Shette, Parle Diagnostic Centre, Mumbai. 2. Dr. N. K. Gupta, M.B.B.S., Udaipur. 3. Mr. Anil Sharma, Central Lab., Govt. Hospital,

Nathdwara. 4. Mr. Rajiv Mathur, Head Department of Electronics

& Communication, GITS, Udaipur. 5. Prof. Pradeep Chhawchharia, Director, Techno India

NJR Research Centre, Udaipur. 6. Mr. Sourabh Porwal, Ass. Professor, Department of

Electronics & Communication, GITS, Udaipur.

REFERENCES

[1] Raviraja, S, Gaurav Bajpai, Sudhir Kumar Sharma, “Analysis Of Detecting The Malarial Parasite Infected Blood Images Using Statistical Based Approach”, International Conference On Biomedical Engineering (Biomed), Malaysia, 2006.

[2] Smyth, J.D., Introduction to Animal Parasitology, Cambridge University Press, Cambridge, 1994.

[3] Ruberto, C.Di., Dempster, A. Shahid Khan, Bill Jarra, "Automatic Thresholding of Infected Blood Images Using Granulometry and Regional Extrema," ICPR, p.3445, 15th International Conference on Pattern Recognition (ICPR'00) - Volume 3, 2000.

[4] Ruberto, C.Di., Dempster, A. Shahid Khan, Bill Jarra, "Segmentation of Blood Images Using Morphological Operators", ICPR, p.3401, 15th International Conference on Pattern Recognition (ICPR'00) - Volume 3, 2000.Alasdair McAndrew, “Introduction of Digital Image Processing with MATLAB, Cengage Learning, 2nd Indian reprint 2011.

[5] Royal Perth Hospital, Western Australia, Malaria, An on-line Resource Website: http://rph.wa.gov.au/malaria.html

[6] Dempster, A. Ruerto, C.Di. Morphological Processing of Malarial Slide Images, Matlab DSP Conference 1999, Nov, 16-17 1999, Espoo, Finland.

[7] Sonka, Hlavac, and Boyle, Digital Image Processing and Computer Vision, Cengage Learning, Fourth Indian reprint 2011.

[8] Hu M.K., “Visual Pattern Recognition by Moment Invariants”, IRE Transactions Information Theory, 8(2):179-187, 1962.

[9] Gonzalez R.C. and Woods R.E., Digital Image Processing, Pearson Prentice-Hall, First Impression, 2009.

[10] Pratt W.K., Digital Image Processing, John Wiley and Sons, New York, 2nd edition, 1991.

[11] Jayaraman S, Esakkirajan S., and Veerakumar T., Digital Image Processing, Tata McGraw Hill, Fourth Indian reprint 2011.

[12] http://upload.wikimedia.org/wikipedia/commons/ thumb/7/70/Plasmodium_vivax_01.png/240px-Plasmodium_vivax_01.png

[13] Savini. M, “Moments in Image Analysis,” Alta Frequenza, 57(2):145-152, 1988. Cash, G.L. and Hatamian, “M. Optical character recognition by the method of moments, Computer Vision, Graphics, and Image Processing”, 39 (3): 291–310, 1987.

[ieee 2012 international conference on computing, communication and applications (iccca) - dindigul,...

Documents