INTRODUCTION:
Diagnosis of AML is lymphoblasts identification using microscope, immune phenotypic valuation and stage prediction by flow cytometry. Initially, the observation of blood samples by an expert hematologist is one diagnostic technique to distinguish different diseases. Visual inspection is not exactness, prolonged and monotonous because it depends on hematologist skill. For the spontaneous quantification of blood cells there occur many approaches that count the numbers of different types of cells within a blood smear.
According to the literature, there are limited examples of automated systems that can examine and categorize WBCs from microscopic images, and the present systems are only partially automated. In particular, a significant amount of work has been performed to achieve leucocytes segmentation. Putzu et al. proposed methods which study the blood cells automatically via image processing procedures, and it represents a medical instrument to avoid the abundant drawbacks associated with labor-intensive observation. This method can also be used for counting, as it provides tremendous performance and allows for primary diagnostic suspicion, which can then be confirmed by a hematologist [1].
Madhloom et al. developed an arithmetic and threshold operations which automatically localize and segment WBC nuclei [2]. Sinha for segmenting white blood corpuses used k-means clustering on the HSV color space and different classification models for cell differentiation [3]. Kovalev and et al. Proposed a method to separate the five types of leucocytes in cell images. In his method he identified the nuclei first and then detected the entire membrane by region growing techniques [4].
Ananthi and Balasubramaniam proposed a novel automatic segmentation technique based on interval-valued intuitionistic fuzzy similarity measure. Blood smear images in RGB color model is converted into HSI color model. The S-channel of HSI (Hue Saturation Intensity) color model is used for further enhancement technique. The S-channel is considered for analysis because it efficiently shows leukocytes in blood smear images [5]. The threshold value ‘T’ chosen for contrast enhancement is of range between [0, 255]. The image is further segmented with the determined threshold.
Chinwaraphat et al. applied fuzzy c-mean clustering for clustering of WBC components. The change in the conventional FCM was based on an iteration of false scattering color substitution by using a neighboring color data. The drawback of this method was physical cropping had been required for the test images, and the performance of this method is not compared with other existing methods.
Jie Su proposed an automatic segmentation of AML by initially performing segmentation using k-means cluster, then builds cell image representing model by HMRF (Hidden-Markov Random Field), estimates model parameters through probability of EM (expectation maximization), carries out convergence iteration until optimal value [7].
A novel method based on Dictionary Learning (DL) and Sparse Representation (SR) is proposed for classification of different sub-types of Acute Myelogenous Leukemia (AML). This method also includes a pattern recognition system which usually consists of segmentation, feature extraction and classification [8]. A fuzzy based approach is used for segmentation of WBCs in color bone marrow images which takes into account color components stability degrees [9]. A texture based approach is proposed for recognition of leukocyte in blood cells [10].
Many tries were made within the past to assembled systems that useful resource in acute leukemia segmentation and category. Segmentation plays an important step medical image processing which is used for extracting different objects in an image especially for pathological images. According to pathological studies, blood cell parameters such and platelets are very essential to detect many diseases such as anemia, leukemia, cancer and any other infections. There exists numerous algorithms by using image processing for the automatic detection of blood cells. A white blood cell (WBCs) plays an important role in diagnosing leukemia (type of cancer) in patients than red blood cells [11]. Fig 1 shows the various types of leukocytes present in the blood cell.
A)Basophil b) Esinophil c) Neutrophil
d) Lymphocyte e) Monocytes
Fig 1. Types of leukocytes
The organization of this paper is as follows, section 1 discusses the literature survey conducted with respect to the related work. The preprocessing algorithm applied on the raw image is explained in Section 2. Local Binary Pattern (LBP) algorithm is discussed in section 3. The segmentation algorithm fuzzy c mean is in section 4. The result obtained is discussed in section 5. Last section 6 deals with the conclusion.
PREPROCESSING:
Preprocessing technique is used for improving the contrast and removal of noise present in the biomedical image. Noise in the blood cell image may cause the change in brightness level even thought there is no detail in the image. Sometimes it reduces the image quality of the blood cell images. Noise may occur in microscopic images during acquisition of the image or while transmission of data by electrical means through the channels. Segmentation of noisy microscopic images will produces incorrect segmentation results. This paper utilizes Adaptive Weighted mean filter (AWMF) for removing noise and histogram equalization for improving the contrast of the image. In weighted mean filter the pixel value is replaced by the average weighted value of the neighbouring pixel. After filtering edges get blurred due to this noise removal technique. The edges can be enhanced by applying suitable contrast enhancement technique [12][13]. In Adaptive Weighted Mean Filter (AWMF) the window size is automatically adjusted according to the noise level. The image is separated into Base and Detail layer and then filtering technique is applied. For improving the contrast and to enhance the blurred edges, Adaptive Histogram Equalization (AHE) algorithm is applied on the output image obtained after filtering. Fig 2 shows the data base images considered for analyzing AML. The output obtained after preprocessing is shown in Fig.3.
(a) (b) (c)
(d) (e) (f)
Fig 2. Blood samples from AML patient (a-d). and healthy cells from non AML patient (e & f)
LOCAL BINARY PATTERN:
Local Binary Pattern (LBP) is applied on the AML image for extracting the blood cell features. The LBP operator compares the center pixel value with the neighbouring pixel value. If the neighbourhood pixel is greater than a central pixel, then its bit is recorded as 1, otherwise 0. The value of the central point is the weighted sum of each bit. The basic LBP operator is given as [14]
(1)
(2)
Where Pn is the pixel value for the neighborhood pixel and Pc is the pixel value of the center pixel for a 3 X 3 image. This linear binary pattern is weighted, and the values are added up to get the LBP code which contains information about the local features of the texture of the image. The output obtained after applying linear Binary Pattern is shown in Fig.4
FUZZY C-MEAN CLUSTERING:
Clustering is a process for classifying objects or patterns in such a way that samples belong to same group is more similar to one another than samples belonging to different groups. There are many clustering technique available, such as hard clustering and fuzzy clustering. In hard clustering method date is restricted exclusively to one cluster [17-19]. As a result, with this method the segmentation results are often very crisp, in which each pixel of the image belongs to exactly just one class. Fuzzy clustering is a soft segmentation method among this Fuzzy C Mean (FCM) is commonly used method for segmentation of AML images.
|
INPUT IMAGES |
PREPROCESSING OUTPUT |
|
|
AWMF output |
AHE output |
|
|
Input image 1 |
(a) |
(e) |
|
Input image 2 |
(b) |
(f) |
|
Input image 3 |
(c) |
(g) |
|
Input image 4 |
(d) |
(h) |
Fig. 3 Preprocessing outputs for four different types of AML images.
Fig 3.(a)-(d) output image after applying AWMF
Fig 3. (e)- (h) output image after applying AHE
|
INPUT IMAGES |
LBP output |
FCM output |
|
Input image 1 |
(a) |
(e) |
|
Input image 2 |
(b) |
(f) |
|
Input image 3 |
(c) |
(g) |
|
Input image 4 |
(d) |
(h) |
Fig 4.(a)-(d) output image after applying LBP
Fig 4. (e)- (h) output image after applying FCM
FUZZY C-MEAN CLUSTERING:
Fuzzy C mean algorithm[15] groups the
blood cell pixel value into k clusters with centers c1, …ck
, …cK in the training space. A real valued vector is assigned to
each pixel of the blood cell image is 
is the membership value. The membership
vector values 
and cluster centroids ck is
obtained by minimizing
Where K is the number of clusters and N is the number of pixel in the image database. is the membership function. Corresponds to the Euclidian distance, is the identity matrix.
RESULTS AND ANALYSIS:
The acute myeloid leukemia image considered in this study is manually cropped using image cropping tool in MATLAB. The steps adapted for preparation of leukaemia image are first the image is enlarged to an optimum extends, and cropping the Region of Interest (ROI) from the image. The image was collected from American Society of Hematology (ASH) image bank (http://www.hematology.org). On the raw image preprocessing algorithm such as Adaptive histogram based contrast enhancement and noise removal by Adaptive Weighted mean filter (AWMF) is applied whose result is shown in Fig 3. On the preprocessed image, feature extraction technique Local Binary Pattern (LBP) is applied and then the blood cell is segmented using fuzzy c mean clustering whose output is shown in Fig 4. Though the analysis of acute myeloid leukemia detection is done by three steps such as preprocessing, feature extraction and segmentation, this approach has achieved satisfactory results in visual evaluation of blood cell segmentation.
CONCLUSION:
This paper discusses about the automatic detection of acute myeloid leukemia from the blood sample images using Fuzzy c mean algorithm. Accurate cell image segmentation is of great significance for feature extraction classification and recognition. The experimental results show the effectiveness of this method compared with the existing methods. This segmentation algorithm will benefit clinicians and patients. They also provide important information for 3-D visualization, surgical planning, and early disease detection.
REFERENCES:
1. Lorenzo Putzua, Giovanni Caoccib, Cecilia Di Rubertoa. Leucocyte classification for leukaemia detection using image processing techniques. Artificial Intelligence in Medicine vol no 62 (2014) pp 179–191.
2. Madhloom HT, Kareem SA, Ariffin H, Zaidan AA, Alanazi HO, Zaidan BB. An automated white blood cell nucleus localization and segmentation using image arithmetic and automated threshold. J Appl Sci 2010;10(11):959–66.
3. Sinha N, Ramakrishnan AG. Automation of differential blood count. In: Chockalingam A. Proceedings of the conference on convergent technologies for the Asia-Pacific region, October 15–17. IEEE Publisher; 2003. p. 547–51.
4. Kovalev VA, Grigoriev AY, Ahn H. Robust recognition of white blood cell images. In: Kavanaugh ME, Werner B. Proceedings of the 13th international conference on pattern recognition, August 25–29. Vienna, Austria: IEEE Publisher; 1996. p. 371–5.
5. V.P. Ananthi, P. Balasubramaniam. A new thresholding technique based on fuzzy set as an application to leukocyte nucleus segmentation. computer methods and programs in biomedicine 134, 2016, pp 165-177.
6. S. Chinwaraphat, A. Sanpanich, C. Pintavirooj, M. Sangworasil, P. Tosranon. A modified fuzzy clustering for white blood cell segmentation, in: 3rd International Symposium on Biomedical Engineering, 2008, pp. 356–359.
7. Jie Su, Shuai Liu , Jinming Song. A segmentation method based on HMRF for the aided diagnosis of acute myeloid leukemia. Computer Methods and Programs in Biomedicine 152 (2017) 115–123.
8. Omid Sarrafzadeh, Hossein Rabbani, Alireza Mehri Dehnavi, Ardeshir Talebi. Detecting different sub-types of acute myelogenous leukemia using dictionary learning and sparse representation. IEEE International Conference on Image Processing (ICIP), 2015, pp 3339 – 3343.
9. E. Montseny, P. Sobrevilla, and S. Romaní. A fuzzy approach to white blood cells segmentation in color bone marrow images. Fuzzy Systems, 2004. Proceedings. 2004 IEEE International Conference on, 2004, pp. 173-178.
10. D. M. U. Sabino, L. da Fontoura Costa, E. Gil Rizzatti, and M. Antonio Zago. A texture approach to leukocyte recognition. RealTime Imaging, vol. 10, pp. 205-216, 2004.
11. N. Theera-Umpon, S. Dhompongsa. Morphological granulometric features of nucleus in automatic bone marrow white blood cell classification. IEEE Trans. Inf. Technol. Biomed. 11 (3) (2007) 353–359.
12. Xuming Zhang and Youlun Xiong. Impulse Noise Removal Using Directional Difference Based Noise Detector and Adaptive Weighted Mean Filter. IEEE signal processing letters, vol. 16, no. 4, pp.295-298, April 2009.
13. Yiqiu Dong and Shufang Xu. A New Directional Weighted Median Filter for Removal of Random-Valued Impulse Noise. IEEE signal processing letters, vol. 14, no. 3, pp. 193-196 , March 2007.
14. T. Ojala, M. Pietikainen, and T. Maenpaa. Multiresolution gray-scale and rotation invariant texture classification with local binary patterns. IEEE Trans. Pattern Anal. Mach. Intell., vol. 24, no. 7, pp. 971–987, Jul. 2002.
15. J. C. Bezdek. Pattern Recognition With Fuzzy Objective Function Algorithms. New York, NY, USA: Plenum, 1981.
16. Yuchun Tang, Yan-Qing Zhang. FCM-SVM-RFE Gene Feature Selection Algorithm for Leukemia Classification from Microarray Gene Expression Data. The IEEE International Conference on Fuzzy Systems, 2005, pp 99-101.
17. P.Chitra. Quantitative Characterization of Radiographic Weld Defect Based on the Ground Truth Radiographs Made on a Stainless Steel Plates. Advances in Intelligent Systems and Computing, Springer publication, Volume 433, pp 157-166,2016.
18. Melissa, S., Srilatha, K. A novel approach for pigmented epidermis layer segmentation and classification. International Journal of Pharmacy and Technology. March-2016, Vol. 8, Issue No.1, 10449-10458.
19. S. Sheela, M. Sumathi. Study and Theoretical Analysis of Various Segmentation Techniques for Ultrasound Images. ICRTCSE 2016, Elsevier – Procedia Computer Science, no.87, pp. 67 – 73, 2016.
Received on 18.05.2018 Modified on 10.07.2018
Accepted on 28.08.2018 © RJPT All right reserved
Research J. Pharm. and Tech. 2019; 12(4):1717-1720.
DOI: 10.5958/0974-360X.2019.00286.5