Detection of Malaria Parasite Using K-Mean Clustering Avani Patel, Zalak Dobariya Electronics and Communication Department Silver Oak College of Engineering and Technology, Ahmedabad I. INTRODUCTION Malaria is a most popular life-threatening parasitic disease, and it s transmitted inside human body through female Anopheles mosquito. It s caused by the genus Plasmodium the protozoan parasites. This parasite grows and reproduces to the complex life cycle. During whole process, hosts are red blood cells (RBCs) which are destroyed afterwards. Hence, the ratio of total number of red blood cells to infected parasite cells decreases. Malaria is one of those diseases which are caused due to blood. However, they are rarely used in developing countries because of the high cost, specialized infrastructure needs and very much handling difficulties. Low cost method is Rapid diagnostic test; Rapid Diagnostic Test (RDTs) detects specific antigens derived from malaria parasites in blood [1] and conventional microscopy. In malaria diagnosis, RTD is relatively fast and can be administered by unskilled personnel. II. METHODOLOGY 2.1 Image Acquisition The first step is to acquire the images of malaria samples. In this study, the malaria images of ring, Trophozoite and gametocyte stages have been captured from the thin blood smears of P.Vivax samples P.oval samples and other samples. The malaria slides are collected from Supratech Micro path pathology and research institute Ahmadabad. Each slide has been stained by using the Giemsa staining. 2.2 Contrast Enhancement Using Partial Contrast Stretching The malaria images captured through the microscope may have their own weaknesses such as blurred or low contrast of the images. Thus, a contrast enhancement technique namely partial contrast stretching (PCS) is utilized to improve the quality of images and contrast of malaria images. This technique is based on the most popular linear mapping function that is used to increase the contrast and brightness levels of the malaria image. Partial contrast is a linear mapping function that is used to increase the contrast level and brightness level and also quality of image of the image. The technique is based on the original brightness and contrast level of the images to be adjusted so we can see clearly. First the system will be find the range of the majority input pixels converge for the each colour space. Since the input image is in RGB colour space, so that it is necessary to find the pixels range between the red color, blue color and green color intensities. Then, the average of these three colour space will be calculated to using the upper and lower colour values by using the following formula [2]: maxth = (maxred + maxblue + maxgreen)/3 minth = (minred + minblue + mingreen)/3 output x y = in(x, y) minth NminTH NmaxTH NminTH (in x, y fmin) maxth minth in x, y ( maxth NmaxTH) Where, in(x,y) : color level for the input pixel out(x,y) : color level for the output pixel minth : lower threshold value maxth : upper threshold value NminTH : new lower stretching value NmaxTH : new upper stretching value fmin: Minimum colour level values in the input image IJIRT 142465 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 42
2.3 Detection of Malaria Parasites Based on RGB, HSI and C-Y color Models Since the differences in smear preparation and also in the imaging condition can cause the variations in malaria images, selection of color component is very important step, in this step may ease the parasite detection and segmentation process in malaria images[3]. Thus, the current study investigates three types of color models which are red green blue (RGB), hue saturation intensity (HIS) and chrominance-luminance C-Y color models. The RGB is the best of the known color model and is widely used for acquiring and displaying the color digital images. Each color pixel is represented most impotent three components which are red (R), green (G) and blue (B). As for the HSI and C-Y color models, the two color models have been chosen because both color models are very important and attractive color models for image processing applications as they can represent the color similarly as how the human eye senses color [4]. 2.4 Image Segmentation Using k-means Clustering Colour models, the next and most important step in image segmentation are to extract the meaningful region from the malaria images. The malaria slides are usually stained to highlight the region of interest (ROI) or mining full region which is referred to the parasite or infected cell in the images of malaria. However, segmenting the parasite or infected cell in an image is not an easy task or process due to the inconsistency intensity or different intensity of these two regions as it may appear lighter or darker part of the image depending in the ph of the buffer also used [5]. 2.5 Image Filtering Using Median Filter Algorithm After done segmentation k-mean, we can see than there might be some unwanted regions or noise that is give not batter results of the image. Thus, median filter is used to remove the noise in order to obtain a noise-free image or give nose less image. Due to its good smoothing effect of the image, it can also be used to fill the small holes that might appear on the segmented infected cell in the image. Here, the neighborhood of n n (n = 5) pixels is used because of the large neighborhoods produce more severe smoothing. 2D Median filtering example using a 3 x 3 sampling window 2.6 Seeded Region Growing Area Extraction Algorithm In the last stage, a modified version of conventional seed based region growing algorithm it s name is seeded region growing area extraction (SRGAE) algorithm has been applied on the segmented image. This algorithm is chosen due to its capability to choose the Region of Interested according to their order in the image as well as extracting the size of the segmented region in the image. Since the segmentation using k-means clustering is based on the only on colour information of the pixels in the malaria image, some artefacts and the unwanted regions which share the same colour of image as the infected cell are still appeared on the segmented image of malaria. So that, the SRGAE algorithm is applied for the two main purposes. First one is to calculate the total area in pixels for the region of interested (ROI). Second one is to remove any unwanted regions or noise that are bigger in size in which cannot be cleaned by using the 5x5 pixels median filter [6]. 2.7Canny edge detection The Process of Canny edge detection algorithm given 5 steps show below: 1. Apply Gaussian filter to smooth the image in order to remove the noise into image. 2. Find the intensity gradients of the image 3. Apply non-maximum suppression to get rid of spurious response to edge detection 4. Apply double threshold to determine potential edges 5. Track edge by hysteresis: Finalize the detection of edges by suppressing all the other edges that are weak and not connected to strong edges. III. RESULTS 3.1 Image Acquisition The first step is to acquire the images of malaria samples. In this study, the malaria images of ring, trophozoite and gametocyte stages have been IJIRT 142465 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 43
captured from the thin blood smears of p.vivax samples. Size of image is 300 900 pixels. Fig.3.1: Samples of the captured malaria images 3.2 Contrast Enhancement Using Partial Contrast Stretching Thus, a contrast enhancement technique namely partial contrast stretching (PCS) is utilized to improve the image quality and contrast of malaria image as show in result. Fig.3.2: contrast enhancement 3.3 Detection of Malaria Parasites Based on RGB, HSI and C-Y color Models Since the differences in smear preparation as well as the imaging condition can cause variations in malaria images, selection of color component is very important as this step may ease the parasite detection and segmentation process. Red green blue Red Green Blue Fig 3.3(a): RGB color models that have been extracted from malaria image IJIRT 142465 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 44
HIS Hue of HIS Saturation of HSI Intensity of HSI Fig.3.3 (b): HSI color models that have been extracted from malaria image R-Y of C-Y B-Y of C-Y IJIRT 142465 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 45
Saturation of C-Y Hue OF C-Y Fig.3.3(c): C-Y color models that have been extracted from malaria image 3.4 Image Segmentation Using k-means Clustering Anunsupervised pixel segmentation based on k-means clustering algorithm is applied for easily segmenting the infected cell from its complicated blood cells background. The k means is a clustering method which is one of the most popular unsupervised learning algorithms due to its simplicity. Fig.3.4: k means clustering on contrast stretching image 3.5 Median filtering Thus, median filter is used as a noise removal in order to obtain a noise-free image. Due to its good smoothing effect, it can also be used to fill the small holes that might appear on the segmented infected cell. Here, the neighbourhood of n n (n = 5) pixels is used because large neighbourhoods produce more severe smoothing. Fig.3.5: median filter image 3.6 Seeded Region Growing Area Extraction Algorithm In this study, a modified version of conventional seed based region growing algorithm namely seeded region growing area extraction (SRGAE) algorithm has been applied on the segmented image. This algorithm is chosen due to its capability to label the ROI according to their order in the image as well as extracting the size of the segmented region. Two main purposes first are to calculate the total area in pixels for the ROI. Secondly is to remove any unwanted regions that are bigger in size in which cannot be cleaned by using the 5 5 pixels median filter. In order to apply the SRGAE algorithm, the segmented malaria image will first be converted into binary image, where the ROI and background regions will be assigned to 0 and 255, respectively. IJIRT 142465 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 46
Area 5706 pixels Fig.3.7: Seeded Region Growing Area Extraction 3.6 canny edge detaction Algorithm In the last SRGAE algorithm compeer with canny detection. SRGAE is give better result but it s not good in time. Time problem is more in SRGAE. So I can use canny detection in place of region growing and give same results in short time compare to region growing.now show in fig 3.8(a) results of canny edge detection. Fig 3.8(a) show the gametocyte cell of vivax parasite. (1) Input image (2) canny edge detection (3) malaria detected (4) malaria cell detected using edge detection Fig.3.8 (a): Results of the canny edge detection apply on gametocyte image Show in table time comparison between region growing and canny edge detection. In final result I get batter result of canny edge detection and conclude that canny edge detection is batter compare to region growing. Table 1 Images Region growing time Canny edge detection time Po_gametocyte_thinE.jpg 3.587280 seconds 2.916711 seconds Pf_schizont_thinA.jpg 3.172308 seconds 2.917747 seconds Pf_gametocyte_thinC.jpg 3.150307 seconds 2.808387 seconds Po_troph_thinA.jpg 2953871 seconds 27422109 seconds IJIRT 142465 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 47
REFERENCE [1] Beadle, C., Long, G.W., Weiss, W.R., P.D. McElroy, Maret, S.M., Oloo, A.J. and Hoffman, S.L. Diagnosis of malaria by detection of p. falciparum HRP-2 an- tigen with a rapid dipstick antigen Capture assay. Lancet, 343, 564-568 1994 [9] Pallavi T. Suradkar Detection of Malaria Parasite in Blood Using Image Processing International Journal of Engineering and Innovative Technology April 2013. [2] Jaspreet Kaur, Amita Choudhary, Comparison of Several Contrast Stretching Techniques on Acute Leukemia Images International Journal of Engineering and Innovative Technology (IJEIT) Volume 2, Issue 1, July 2012 [3] Pallavi T. Suradkar Detection of Malaria Parasite in Blood Using Image Processing International Journal of Engineering and Innovative Technology April 2013 [4] Rapid diagnostic tests for Malaria Parasites by Anthony Moody Clin. Microbiol. Rev. 2002, 15(1):66. DOI: 10.1128/CMR.15.1.66-78.2002. [5] J. MacQueen Some methods for classification and analysis of multivariate observations Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability, 1967, pp. 281-297. [6] N. H. Harun, M. Y. Mashor, and H. Rosline Calculation of blast area for acute leukemia blood cells images Proceedings of the International Postgraduate Conference on Engineering, 2010. [7] S. Mandal, A. Kumar, J. Chatterjee, M. Manjunatha, and A. K. Ray, Segmentation of blood smear images using normalized cuts for detection of malarial parasites annual IEEE India conference 2010. [8] M. Ghosh, D. Das, C. Chakraborty, and A. K. Ray Probabilistic Prediction of Malaria using Morphological and Textural Information international conference on image information processing, 2011. IJIRT 142465 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 48