HISTOPATHOLOGICAL IMAGES FOR THE DETECTION OF BRAIN …ijirae.com/volumes/Vol4/iss03/09.MRAESP10088.pdf · DETECTION OF BRAIN TUMORS USING CLUSTERING ... Abstract- A tumor is a mass

International Journal of Innovative Research in Advanced Engineering (IJIRAE) ISSN: 2349-2163 Issue 03, Volume 4 (March 2017) Special Issue www.ijirae.com

___________________________________________________________________________________________________ IJIRAE: Impact Factor Value – SJIF: Innospace, Morocco (2016): 3.916 | PIF: 2.469 | Jour Info: 4.085 |

ISRAJIF (2016): 3.715 © 2014- 17, IJIRAE- All Rights Reserved Page - 48

HISTOPATHOLOGICAL IMAGES FOR THE

DETECTION OF BRAIN TUMORS USING CLUSTERING TECHNIQUES

A.ANCY CHRISTY, ANU JOSE PG Scholar Assistant Professor

Department of Medical Electronics Engineering Department of Medical Electronics Engineering Sengunthar College of Engineering, Tiruchengode Sengunthar College of Engineering, Tiruchengode Abstract- A tumor is a mass of tissue formed by an accumulation of abnormal cells. Usually, the cells in your body will die, and will be replaced by new cells. With cancer and other tumors, something bothers this cycle. Image processing is an active research area in which medical image processing is an extremely demanding field. Brain tumor analysis is performed by physicians, but its gradation gives various conclusions that can vary from one physician to another. In traditional work, brain cancer is one of the leading cancers that have increasing cases recently. Computer Aided Diagnosis (CAD) is a useful tool to help doctors make better choices in cancer diagnosis and treatment. Accurate cell detection is often an essential prerequisite for subsequent cell analysis. The major challenge of robust brain tumors / cell detection is to cope with significant changes in the cell image and separate the cells. The other state-of-the-art cell detection algorithms are not applicable to our data because a local patch does not contain sufficient information to detect a real cell boundary and an edge in the background, especially when contrast changes occur. The proposed method of brain histopathological images for the detection of tumors using clustering techniques. K-Means clustering algorithm is a non-supervised clustering algorithm that classifies the input data points into several classes based on their inherent distance to one another. The algorithm assumes that the data features form a vector space and tries to find natural clustering in it. The scans of the human brain form the input images for our system, where the hemotoxylin and eosin (H&E) histopathological input images are given as input. The preprocessing stage converts the RGB input image into the space-to-L * a * b * color space. The preprocessed image is given for image segmentation using the K-Means clustering algorithm. Finally, the dark blue is separated from the light blue by the 'L *' layer in the L * a * b * color space. The proposed method achieves the best cell recognition accuracy with efficient clustering.

Keywords- Histopathological image, K- means clustering.

I. INTRODUCTION 1.1 GENERAL Brain Tumor is one of the leading cancers that have increasing cases recently. Successful diagnostic and prognostic layering, therapy outcome prediction and therapy planning are based on a reproducible and accurate analysis of digitized histopathological samples. The current manual analysis of histopathological films is not only tedious but also subject to the variability of observers. Computer-Aided Diagnosis (CAD) systems have attracted wide interests. Cells and stand-alone cells are classified on the basis of geometrical characteristics. Then, the conventional water-cutting method is improved by a subregion-merging mechanism and a Laplace-of-Gaussian (LOG) filters. LoG-based methods are typically sensitive to large variations in cell size and the absence of apparent cell boundaries. The proposed system of brain histopathological images for the detection of tumors using clustering techniques. A cluster can be defined as a group of pixels, all pixels in a particular group being defined by a similar relationship. Clustering is also known as unattended classification technology. K-Means The clustering algorithm is a non-supervised clustering algorithm that classifies the input data points into several classes based on their inherent distance to one another.

1.2 K-MEANS CLUSTERING A color-based segmentation method using the k-means clustering technique is to track the tumor objects in the histopathological brain images. K-means is a widely used clustering algorithm to partition data into k clusters. Clustering is the process of grouping data points with similar feature vectors into a single cluster and grouping data points with dissimilar feature vectors into different clusters.

http://www.ijirae.com




Segmentation is an essential process to extract information from complex medical images. The main focus of image segmentation is to separate an image into generally exclusive and exhausted regions so that each region of interest is spatially adjacent and the pixels within the domain are homogeneous with respect to a predefined criterion. A cluster is a collection of objects that are similar between them and do not belong to the objects of other clusters. Clustering is an unsupervised learning process that involves finding a structure in a collection of unmarked data.K means clustering is an algorithm to group of objects that is based on attributes / features into k number of groups, where k is a positive integer. The clustering is done by minimizing the Euclidean distance between the data and the corresponding cluster center of gravity. The function of k-means clustering is thus to clutter the data.

II. RELATED WORK

SCALE-SPACE AND EDGE DETECTION USING ANISOTROPIC DIFFUSION

Author: Pietro Perona and Jitendra Malik in the year of 2011. The scale-space technique introduced by Witkin involves generating coarser resolution images by convolving the original image with a Gaussian kernel. This approach has a major drawback: it is difficult to obtain accurately the locations of the “semantically meaningful” edges at coarse scales. It is shown that the “no new maxima should be generated at coarse scales” property of conventional scale space is preserved. As the region boundaries in our approach remain sharp, we obtain a high quality edge detector which successfully exploits global information. The algorithm involves elementary, local operations replicated over the image making parallel hardware implementations feasible.

ROBUST TRACKING USING LOCAL SPARSE APPEARANCE MODEL AND K-SELECTION

Author: Baiyang Liu , Junzhou Huang, Lin Yang in the year of 2014. Online learned tracking is widely used for it’s adaptive ability to handle appearance changes. The recent literature demonstrates that appropriate combinations of trackers can help balance stability and flexibility requirements. We have developed a robust tracking algorithm using a local sparse appearance model (SPT). A static sparse dictionary and a dynamically online updated basis distribution model the target appearance. A novel sparse representation-based voting map and sparse constraint regularized mean-shift support the robust object tracking. Besides these contributions, we also introduce a new dictionary learning algorithm with a locally constrained sparse representation, called K-Selection. Based on a set of comprehensive experiments, our algorithm has demonstrated better performance than alternatives reported in the recent literature.

FEATURE CODING IN IMAGE CLASSIFICATION: A COMPREHENSIVE STUDY

Author: Yongzhen Huang, Zifeng Wu, Liang Wang, and Tieniu Tan in the year of 2013 Image classification is a hot topic in computer vision and pattern recognition. In this paper, we firstly make a survey on various feature coding methods, including their motivations and mathematical representations, and then exploit their relations, based on which a taxonomy is proposed to reveal their evolution. Finally, we choose several representatives from different kinds of coding approaches and empirically evaluate them with respect to the size of the codebook and the number of training samples on several widely used databases Experimental findings firmly justify our theoretical analysis, which is expected to benefit both practical applications and future research.

TRAINING-FREE, GENERIC OBJECT DETECTION USING LOCALLY ADAPTIVE REGRESSION KERNELS

Author: Hae Jong Seo and Peyman Milanfar in the year of 2012. We present a generic detection/localization algorithm capable of searching for a visual object of interest without training. The proposed method operates using a single example of an object of interest to find similar matches, does not require prior knowledge (learning) about objects being sought, and does not require any preprocessing step or segmentation of a target image. Our method is based on the computation of local regression kernels as descriptors from a query, which measure the likeness of a pixel to its surroundings. Salient features are extracted from said descriptors and compared against analogous features from the target image. This comparison is done using a matrix generalization of the cosine similarity measure. The approach is extended to account for large variations in scale and rotation. High performance is demonstrated on several challenging data sets, indicating successful detection of objects in diverse contexts and under different imaging conditions.

III. SYSTEM DESIGN

The System design describes about the detection of tumors using clustering techniques. A cluster can be defined as a group of pixels, all pixels in a particular group being defined by a similar relationship. Clustering is also known as unattended classification technology. K-Means clustering algorithm is a non-supervised clustering algorithm that classifies the input data points into several classes based on their inherent distance to one another. The algorithm assumes that the data features form a vector space and tries to find natural clustering in it.





Figure 3.1 System Design

IV. PROPOSED SYSTEM

The proposed system of brain histopathological images for the detection of tumors using clustering techniques. A cluster can be defined as a group of pixels, all pixels in a particular group being defined by a similar relationship. Clustering is also known as unattended classification technology. K-Means clustering algorithm is a non-supervised clustering algorithm that classifies the input data points into several classes based on their inherent distance to one another. The algorithm assumes that the data features form a vector space and tries to find natural clustering in it. The scans of the human brain are the input images for our system, where the histopathological input images of haemotoxylin and eosin (H & E) are given as an input signal. The preprocessing stage converts the RGB input image into the space-to-L * a * b * color space. The preprocessed image is given for image segmentation using the K-Means clustering algorithm. Finally, the dark blue is separated from the light blue by the 'L *' layer in the L * a * b * color space. The cell nuclei are dark blue.

V. METHOD IMAGE ACQUISITION Images are obtained using a histopathological image, and these scanned images are displayed in a two-dimensional matrix with pixels as their elements. These matrices are dependent on the matrix size and their facial field. The images are stored in the image file and displayed as haemotoxin and eosin image (H & E). A simple approach for determining the gray value threshold T is to analyze the histogram for peak values and to find the lowest point typically between two successive peak values of the histogram.

PRE-PROCESSING Hematoxylin and eosin (H & E) stained histological images are originally displayed in the red-green-blue (RGB) color space. We convert the RGB image into the CIE L * a * b color space, and the luminance component L is hereinafter referred to as the gray level image I, which is further processed to obtain the segmentation. The L * a * b * space consists of a luminosity layer L *, a chromaticity layer a * which indicates where the color falls along the red-green axis, and a color layer b * indicating where the color falls along the line Blue-yellow axis. The translation formula first calculates the tri-stimulus coefficients W=0.4303R + 0.3416G + 0.1784B, Y= 0.2219R + 0.7068G + 0.0713B, Z= 0.0202R + 0.1296G + 0.9393B.

The CIELab color model is calculated as L* = 116(h(Y/YS))-16, a* = 500(h(W/Ws))-h(Y/YS) b* = 200(h(Y/YS)-h(Z/ZS)), h(q) = � q > 0.008856 = 7.787q+16/116 q ≤ 0.008856, Where YS, WS, and ZS are the standard stimulus coefficients. COLOR BASED SEGMENTATION USING K-MEANS CLUSTERING The converted L * a * b * space image is indicated for image segmentation using the K-means clustering algorithm. Because there are chances for the occurrence of mismatched areas after the application of K-medium clustering algorithm. K-means is a widely used clustering algorithm to partition data into k clusters. Segmentation is an essential process to extract information from complex medical images. The main focus of image segmentation is to separate an image into generally exclusive and exhausted regions so that each region of interest is spatially adjacent and the pixels within the domain are homogeneous with respect to a predefined criterion.





The feature vectors derived from l cluster data are X = {xi│i = 1,2 ...., L}.The generalized algorithm initiates the K cluster centroids C = {cj│j = 1,2, .... K} by randomly selecting k feature vectors are grouped into k clusters using a selected distance measure, such as the Euclidean distance, d= ││xi-cj││

THE K-MEANS ALGORITHM OPERATES AS FOLLOWS: 1. Assigning document vectors, ie 2D, to a cluster using an initial seed. 2. Initialize cluster centers C from the original document mappings. 3. For each document d 2 D

(a) Calculate the distances from documents di to centroids (C1, C2, ..., Ck), and find the closest center of gravity Cmin. (b) Move the document d from the current cluster Ck to the new cluster Cmin and calculate the center of gravity for Ck

and Cmin. 4. Repeat step 3 until either the maximum epoch limit is reached or an epoch where no changes are made to the document assignments.

NUCLEI DETECTION IN HISTOPATHOLOGICAL BRAIN TUMOR IMAGES Hematoxylin dyes cores in dark blue color while eosin dyes other structures (cytoplasm, stroma, etc.) with a pink color. The nuclei are prone to a wide variety of patterns (relative to the distribution of chromatin, prominent nucleolus) that are diagnostically significant. The germ count is then divided into epithelial nuclei and stromal cells using the epithelial mask. The separate epithelial nuclei (light blue) from the stromal cells (dark blue). Among the different types of nuclei, two types are typically the subject of particular interest: lymphocyte and epithelial nuclei. The nuclei can look very different according to a number of factors such as nuclei, malignancy of the disease and nucleus life cycle.

VI. EXPERIMENTAL RESULTS

We have conducted experiments based on a data set containing 59 brain tumor images. The images are captured at 4ox magnification. 27 images are randomly selected as training images from which N = 2000 patches with a centralized single cell are manually cropped. K = 1400 of the 2000 patches are picked out by K-selection. The above figure shows the experimental results of this project and find out the brain tumor identification also established.

Figure 6.1: Converting RBG to L*a*b* color space





Figure 6.2: Image Labeled by clustering index

Figure 6.3: Clustering images





Figure 6.4: Cancer cells

Figure 6.5: Outlines





Figure 6.6: Detected cancer cells area

VII. CONCLUSION

The segmentation of the brain image is essential in surgical planning and therapy planning in the field of medicine. In this proposed system of brain histopathological images for the detection of tumors using clustering techniques. A cluster can be defined as a collection of pixels, all pixels in a particular group being defined by a similar relationship. Clustering is also known as unattended classification technology. K-Means The clustering algorithm is a non-supervised clustering algorithm that classifies the input data points into several classes based on their inherent distance to one another. The scans of the human brain are the input images for our system, where the histopathological input images of haemotoxylin and eosin (H & E) are given as an input signal. The preprocessed image is given for image segmentation using the K-Means clustering algorithm. We were able to segment tumor from different brain histopathological images from our database.

VIII. FUTURE WORK In this future work, preprocessing with clustering technique with a new approach, with this we have successfully established the brain tumor.

IX. REFERENCES

[1]. J. Ferlay, I. Soerjomataram, M. Ervik, R. Dikshit, S. Eser, C. Mathers, M. Rebelo, D. M. Parkin, D. Forman, and F. Bray, “Globocan 2012 v1.0, cancer incidence and mortality worldwide: Iarc cancerbase no. 11,” accessed: 07-11-2014. [Online]. Available: http://globocan.iarc.fr

[2]. L. S. Hu, L. C. Baxter, K. A. Smith, B. G. Feuerstein, J. P. Karis, J. M. Eschbacher, S. W. Coons, P. Nakaji, R. F. Yeh, J. Debbins, and et al., “Relative cerebral blood volume values to differentiate high-grade glioma recurrence from posttreatment radiation effect: direct correlation between image-guided tissue histopathology and localized dynamic susceptibility-weighted contrast-enhanced perfusion mr imaging measurements,” American Journal of Neuroradiology, vol. 30, no. 3, pp. 552–558, 2009.

[3]. C. F. Lucchinetti, B. F. G. Popescu, R. F. Bunyan, N. M. Moll, S. F. Roemer, H. Lassmann, W. Bruck, J. E. Parisi, B. W. Scheithauer, C. Giannini, and et al., “Inflammatory cortical demyelination in early multiple sclerosis,” New England Journal of Medicine, vol. 365, no. 23, pp. 2188–2197, 2011.

[4]. F. Ghaznavi, A. Evans, A. Madabhushi, and M. Feldman, “Digital imaging in pathology: whole-slide imaging and beyond,” Annual Review of Pathology: Mechanisms of Disease, vol. 8, pp. 331–359, 2013.

[5]. M. Veta, J. P. Pluim, P. J. van Diest, and M. A. Viergever, “Breast cancer histopathology image analysis: A review,” IEEE Trans. On Biomedical Engineering, vol. 61, no. 5, pp. 1400–1411, May 2014.

[6]. S. B. Wharton, E. Maltby, D. A. Jellinek, D. Levy, N. Atkey, S. Hibberd, D. Crimmins, K. Stoeber, and G. H. Williams, “Subtypes of oligodendroglioma defined by 1p, 19q deletions, differ in the proportion of apoptotic cells but not in replication-licensed nonproliferating cells,” Acta neuropathologica, vol. 113, no. 2, pp. 119–127, 2007.

[7]. D. Scheie, M. Cvancarova, S. Mørk, K. Skullerud, P. A. Andresen, I. Benestad, E. Helseth, T. Meling, and K. Beiske, “Can morphology predict 1p/19q loss in oligodendroglial tumours?” Histopathology, vol. 53, no. 5, pp. 578–587, 2008.


http://globocan.iarc.fr




[8]. N. Malpica, C. Ortiz de Solorzano, J. J. Vaquero, A. Santos, I. Vallcorba, J. M. Garcia-Sagredo, and F. d. Pozo, “Applying watershed algorithms to the segmentation of clustered nuclei,” Journal of Cytometry, vol. 28, pp. 289 –297, 1997.

[9]. H. Ancin, B. Roysam, T. Dufresne, M. Chestnut, G. Ridder, D. Szarowski, and J. Turner, “Advances in automated 3-d image analysis of cell populations imaged by confocal microscopy,” Journal of Cytometry, vol. 25, no. 3, pp. 22 –234, 1996.

[10]. V. Grau, A. U. J. Mewes, M. Alcaniz, R. Kikinis, and S. Warfield, “Improved watershed transform for medical image segmentation using prior information,” IEEE Trans. on Medical Imaging, vol. 23, no. 4, pp. 447–458, 2004.

[11]. X. Yang, H. Li, and X. Zhou, “Nuclei segmentation using markercontrolled watershed, tracking using mean-shift, and kalman filter in time-lapse microscopy,” IEEE Trans. on Circuits and Systems I: Regular Papers, vol. 53, no. 11, pp. 2405–2414, 2006.

[12]. O. Schmitt and M. Hasse, “Radial symmetries based decomposition of cell clusters in binary and gray level images,” Journal of Pattern Recognition, vol. 41, no. 6, pp. 1905 –1923, June 2008.

[13]. C. Jung and C. Kim, “Segmenting clustered nuclei using h-minima transform-based marker extraction and contour parameterization,” IEEE Trans. on Biomedical Engineering, vol. 57, no. 10, pp. 2600–2604, October 2010.

[14]. S.Taheri, S.H.Ong, V.F.H. Chong, ―Level-set segmentation of brain tumors using a threshold-based speed function, ‖Elsevier, 2009.

[15]. Brijesh Shah, Satish Shah, Y P Kosta,‖Novel Improved Fuzzy C-Mean Algorithm MR-images Segmentation,‖IJSCE,2010and L. Avila, Eds., (ser.Lecture Notes in Computer Science), vol. 6453 Heidelberg, Germany: Springer, 2010, pp. 190–198.


HISTOPATHOLOGICAL IMAGES FOR THE DETECTION OF BRAIN …ijirae.com/volumes/Vol4/iss03/09.MRAESP10088.pdf · DETECTION OF BRAIN TUMORS USING CLUSTERING ... Abstract- A tumor is a mass

Documents