Classification of Distinct Plasmodium Species in Thin Blood Smear Images using Kapur Segmentation Strategy

 

T. Abimala¹, R. Joylin Rini²

^1,2Assistant Professor, ICE Department, St. Joseph’s College of Engineering, Chennai

 

ABSTRACT:

Malaria is a mosquito-borne irresistible chronic sickness of humans and other creatures brought about by parasitic protozoans which belong to Plasmodium type. Malaria causes side effects that incorporate fever, fatigue, vomiting and cerebral pains. If not properly treated, it can bring about yellow skin, unconsciousness and even death. Malaria is induced by five species of plasmodium- P. Falciparum, P. Vivax, P. Malariae, P. Ovale and P. Knowlesi.  In this paper, a venture has been formulated to develop an automated diagnosis strategy for classifying the malarial parasites. The blood smear images obtained from CDC database were segmented by utilizing Fuzzy C Means (FCM) and kapur segmentation strategies. The segmented image has been further utilized to extract features and the extracted measurements have been utilized for classifying the plasmodium species using SVM (Support Vector Machine) classification technique.

 

 

 

1. INTRODUCTION:

Image Segmentation is used to partition an image into numerous segments called pixels.^[1] The main objective of image segmentation is to elucidate an image or modify the image into a form which is easier to inspect and analyze.^[1] Malaria is an communicable infection which can affect red blood cells and is caused by plasmodium parasites. Malaria can be life menacing if it is not treated legitimately. It can be analyzed by provoke and successful recognition of malarial parasites in blood smear.^[3] Conventional strategies like Quantitative Buffy Coat Parasite Detection scheme which has been utilized  to detect malarial parasites appears to be lumbering, tedious and the results are prone to be influenced by operator's exhaustion. Malaria is usually caused by five distinct type of plasmodium species- P. Falciparum, P. Vivax, P. Malariae, P. Ovale and P. Knowlesi.

 

 

 

 

Sri Widodo et al had put forward a novel texture analysis strategy to detect malaria tropica in blood smear images using Support Vector Machine technique.^[4] Detecting tropical malaria by analyzing the shape and pattern of blood smears is an efficient technique than conventional segmentation strategies. The proposed strategy involves two essential steps-Initially segmentation of blood smear images is done by utilizing active contour method. Finally, classification of blood smear images that are affected by plasmodium species is carried out by using Support Vector Machine (SVM) technique. Experimental results illustrates that the proposed strategy produces an accuracy of 93% for normal blood cells, 92.5% for trophozoite, 98.3% for schizont and 100% for gametozyte. The author concluded by stating that the proposed segmentation strategy performs better only for images whose blood preparations do not overlap thereby making the proposed strategy less capable to describe the characteristics of blood images.

 

A. S. Abdul Nasir et al had put forward a unsupervised color image segmentation scheme which detects malarial parasites which utilizes Moving K-means clustering algorithm.^[5] The author has utilized this technique to detect p. vivax species. The author has initially enhanced the segmented red blood cells by utilizing partial contrast stretching technique. Then, MKM clustering algorithm has been utilized to segment the infected cells. Then, median filter and seeded region growing extraction algorithm has been utilized to eliminate unwanted information from the image. Finally, morphological reconstruction algorithm has been applied to fill the holes inside the affected cells. Experimental results illustrates that MKM scheme produced an accuracy of 99.49% than intensity component image.

 

Zazilah May et al a novel strategy which automatically quantifies and classifies erythrocytes which are affected by Plasmodium vivax at trophozoite stage.^[6] The proposed strategy has been divided into three distinct stages- Image processing, Segmentation and Classification. Preprocessing involves conversion into Lab color space initially followed by removal of noise. Segmentation is carried out by utilizing Otsu segmentation technique. Finally, dilation and erosion techniques has been utilized to eliminate background elements completely. Experimental results illustrates that the proposed technique provides higher potential for detecting malarial parasites.

 

The rest of the paper is organized as follows. Section II provides a overview about the proposed system. Experimental results have been provided in Section III and finally conclusion is provided in Section IV.

 

2. DETECTION OF MALARIAL PARASITES IN BLOOD SMEAR IMAGES:

2.1 IMAGE ACQUISTION:

Thin blood smear images of five different plasmodium species- Plasmodium Falciparum, Plasmodium Vivax, Plasmodium Malariae, Plasmodium Ovale and Plasmodium Knowlesi has been acquired from Centre for Disease Control (CDC).

 

2.2 IMAGE PREPROCESSING:

The RGB images are initially converted into gray scale images. The images acquired may have poor illumination. To overcome the above drawback, the contrast of the image is enhanced by balancing the histogram of the image.

 

2.3 SEGMENTATION:

Segmentation is the strategy of isolating an image into multiple components. It is utilized to categorize area of concern. In this paper, Fuzzy C Means clustering (FCM) and Kapur segmentation techniques has been utilized to segregate the relevant information. Fuzzy C Means is a soft clustering algorithm which has been widely utilized to solve medical imaging issues.^[7] We have applied FCM algorithm to extort affected plasmodium parasites from thin blood smear images. The number of clusters is assumed as two. FCM works by allocating membership function to individual data which relates to each cluster center depending on the distance between the centre of the cluster and the allocated data. Kapur is an entropy based thresholding method which maximizes the entropy criterion in order to determine the optimal threshold value. Kapur is a most commonly used bi-level thresholding technique which is an m level optimization problem. On considering the uniformity and shape assessments entropy based segmentation techniques outperforms than other segmentation strategies.

 

2.4 FEATURE EXTRACTION:

Geometric feature extraction technique is a strategy which combines machine learning and computer vision in order to resolve visual errands. The major objective of this strategy is to determine a group of delegate features which are of geometric form in order to characterize an object. They utilize genetic algorithms for learning the features and recognizing objects. Several features such as area, centroid, major and minor axis length, eccentricity, orientation, Euler number, solidity, extent, perimeter has been extracted from distinct plasmodium species using GLCM approach.

 

2.5 CLASSIFICATION:

Image classification breaks down the numerical properties of various image features and organizes data into categories. Classification algorithms ordinarily utilize two phases- training phase and testing phase. Training phase isolates the distinctive features of an image and forms a training class. In testing phase, the extracted feature space partitions are utilized to classify the image. In this paper, Support Vector Machine (SVM) classifier has been utilized to extract and classify image features. Support Vector Machines are supervised learning models which are coupled with certain algorithms to scrutinize the data for classification and deterioration analysis.  A SVM model is a depiction of points in space which are mapped in such a way that different classes are separated by a reasonable space that is as wide as possible.SVM has been utilized to detect five distinct species of plasmodium-Falciparum, Vivax, Malariae, Ovale and Knowlesi.

 

3. RESULTS AND DISCUSSIONS:

The results of the proposed segmentation technique summarized in Figure 1 was carried out for 100 blood samples which incorporated both malaria infected and healthy blood samples. In this paper, an attempt has been made to prove the significance of Kapur segmentation technique than FCM technique which has been proved to be the best segmentation technique among Sobel, Prewitt, Roberts, Log, Zero cross, Canny and Watershed algorithms^[8]. A comparison analysis between Fuzzy C Means and Kapur techniques has been presented below in Table 1. Experimental results illustrate that Kapur segmentation technique outperforms than FCM strategy since FCM fails to segment the images in which the infected red blood cells overlaps with the healthy red blood cells. After segmentation, SVM classifier has been trained using features suchas mean, SD, Entropy, RMS, Variance, Kurtosis and Skewness which has been shown below in Table 2. Each test sample has underwent 500 iterations thereby produced an accuracy of 88%. During the classification process, it is inferred that the SVM strategy is able to classify P. falciparum, P. knowlesi, P. malariae, and P. vivax species accurately but failed to classify P. ovale species.

 

Fig 1: Segmented images using Fuzzy C Means (FCM) and Kapur techniques

 

 

Table 1: Feature extraction using GLCM approach

 

 

 

 

 

 

 

 

 

 

 

Original image	Classified image
Original image	Classified image
Original Image	Classified image

Fig 2: Classification of segmented images using Support Vector Machine (SVM) technique

 

4. CONCLUSION:

In this work, an attempt has been made to segment the thin blood smear images using kapur and Fuzzy C Means (FCM) segmentation strategies.  Classification of blood smear images plays a vital role in diagnosing the disease since the anti malarial drugs to be injected varies based on the plasmodium species. Hence, Support Vector Machine (SVM) has been utilized to classify the distinct species of malarial parasites- P. Falciparum,
P. Vivax, P. Malariae, P. Ovale and P. Knowlesi. Experimental analysis demonstrates that the proposed work successfully classified P. Falciparum, P. Vivax, P. Malariae, and P. Knowlesi species but failed to classify P. Ovale species accurately. Further, this work can be extended to accurately classify P. Ovale species using different segmentation and classification techniques.

5. REFERENCES:

1.       N. Siva Balan et al, “Optimal Multilevel Image Thresholding to Improve the Visibility of Plasmodium sp. in Blood Smear Images”, Proceedings of the International Conference on Soft Computing Systems, Advances in Intelligent Systems and Computing , Springer (2016).

2.       K. Manickavasagam et al, “Development of Systems for Classification of Different Plasmodium Species in Thin Blood Smear Microscopic Images”, J. Adv. Microsc.Vol. 9, No.2Res. 2014.

3.       Feminna Sheeba et al, “Detection of Plasmodium Falciparum in Peripheral Blood Smear Images”, Proceedings of Seventh International Conference on Bio-Inspired Computing: Theories and Applications, Springer 2013.

4.       Sri Widodo et al, “Texture Analysis To Detect Malaria Tropica In Blood Smears Image Using SUPPORT VECTOR MACHINE”, International Journal of Innovative Research in Advanced Engineering (IJIRAE) ISSN: 2349-2163, Volume 1 Issue 8 (September 2014).

5.       A.S.Abdul Nasir et al, “Segmentation Based Approach for Detection of Malaria Parasites Using Moving K-Means Clustering”, 2012 IEEE EMBS International Conference on Biomedical Engineering and Sciences, Langkawi ,17th - 19th December 2012

6.       Zazilah May et al, “Automated Quantification and Classification of Malaria Parasites in Thin Blood Smears”, 2013 IEEE International Conference on Signal and Irnage Processing Applications (ICSIPA).

7.       J. Somasekar et al, “Segmentation of erythrocytes infected with malaria parasites for the diagnosis using microscopy imaging”, Computers and Electrical Engineering xxx (2015).

8.       Abimala.T et al, “Optimal Multi-level Thresholding for RGB Images using Kapur’s Entropy and Firefly Algorithm”, International Journal of Applied Engineering Research, ISSN 0973-4562 Vol. 10 No.87 (2015).

9.       Pal, N. and Pal, S. A review on image segmentation techniques, Pattern Recognition, 26(9) : 1277-1294, 1993.

10.     Sathya, P.D. and Kayalvizhi, R. Optimal multilevel thresholding using Bacterial for aging algorithm, Expert Systems with Applications, 38:15549–15564, 2011.

11.     Sezgin, M and Sankar, B. Survey over Image Thresholding Techniques and Quantitative Performance Evaluation, Journal of Electronic Imaging. 13(1): 146 –165, 2004.

12.     Kapur, J. N., Sahoo, P. K., and Wong, A. K. C. A new method for gray-level picture thresholding using the entropy of the histogram. Computer Vision Graphics Image Processing, 29: 273–285, 1985.

13.     Yang, X.S. Nature-Inspired Metaheuristic Algorithms, Luniver Press, Frome, UK, 2nd edition, 2011.

14.     Fister, I., Yang, X. S., Fister, D., and FisterJr, I. (2014). Firefly algorithm: a brief review of the expanding literature. In Cuckoo Search and Firefly Algorithm (pp. 347-360). Springer International Publishing.

15.     Rajinikanth V, Sri Madhava Raja N, Latha K. Optimal Multilevel Image Thresholding: An Analysis with PSO and BFO Algorithms. Aust. J. Basic and Appl. Sci., 8(9): 443-454, 2014.

16.     Sarkar S, Das S. Multilevel image thresholding based on 2D histogram and maximum Tsallis entropy – A Differential Evolution Approach. IEEE T. on Image Processing, 22(12): 4788-4797, 2013.

17.     Papamarkos, N., Strouthopoulos, C and Andreadis, I. Multithresholding of color and gray-level images through a neural network technique, Image and Vision Computing, 18(3): 213–222, 2000.

18.     Rougemont M, Van Saanen M, Sahli R, Hinrikson HP, Bille J et al. (2004) Detection of four plasmodium species in blood from humans by 18s rrnagenesubunit-based and species-specific real-time pcr assays. J ClinMicrobiol 42:5636-5643.

19.     Snow RW, Guerra CA, Noor AM, Myint HY, Hay SI (2005) The global distribution of clinical episodes of plasmodium falcipar ummalaria, Nature 434: 214-217.

20.     H. Reyburn, (2010) New who guidelines for the treatment of malaria. BMJ 28:340.

21.     Toha S, Ngah U (2007) Computer aided medical diagnosis for the identification of malaria parasites, in: Signal Processing, Communications and Networking, ICSCN’07. International Conference on, IEEE.521-522.

22.     Frean J (2010) Microscopic determination of malaria parasite load: role of image analysis. Micrsocopy: Science, Technology, Applications, and Eductaion862-866.

23.     Somasekar J, Reddy B, Reddy E, Lai C (2011) Computer vision for malariaparasite classification in erythrocytes, International Journal on ComputerScience and Engineering 3: 2251-2256.

24.     Edison M, Jeeva J, Singh M (2011) Digital analysis of changes by Plasmodiumvivax malaria in erythrocytes. Indian Journal of Experimental Biology 49: 11-15.

25.     Anggraini D, Nugroho AS, Pratama C, Rozi IE, Iskandar A A, et al. (2011) Automated status identification of microscopic images obtained from malariathin blood smears. In: Electrical Engineering and Informatics (ICEEI), 2011International Conference on, IEEE. 1-6.

26.     Tek FB, Dempster AG, Kale I (2010) Parasite detection and identification for automated thin blood film malaria diagnosis, Computer Vision and Image Understanding 114: 21-32.

27.     Elter M, Hasslmeyer E, Zerfass T (2011) Detection of malaria parasites in thickblood films. Conf Proc IEEE Eng Med Biol Soc. 2011. 5140-5144.

28.     Mandal S, Kumar A, Chatterjee J, Manjunatha M, Ray A (2010) Segmentation of blood smear images using normalized cuts for detection of malarial parasites. in: India Conference (INDICON), 2010 Annual IEEE1-4.

29.     Makkapati VV, Rao RM (2011) Ontology-based malaria parasite stage and species identification from peripheral blood smear images. Conf. Proc. IEEEEng Med BiolSoc 2011 2011:6138-41.

 

 

 

 

Received on 04.10.2018          Modified on 05.12.2018

Accepted on 27.02.2019        © RJPT All right reserved