Page 47 - Fister jr., Iztok, and Andrej Brodnik (eds.). StuCoSReC. Proceedings of the 2018 5th Student Computer Science Research Conference. Koper: University of Primorska Press, 2018
P. 47
ast Histopathology Image Clustering using Cuckoo Search
Algorithm
Krishna Gopal Dhal1, Iztok Fister Jr.2, Arunita Das3, Swarnajit Ray4, Sanjoy Das5
1 Dept. of Computer Science and Application, Midnapore College (Autonomous), Paschim Medinipur, West Bengal, India.
Email: krishnagopal.dhal@midnaporecollege.ac.in
2Faculty of Electrical Engg. and Computer Sc., University of Maribor, Slovenia, Email: iztok.fister1@um.si.
3Dept. of Information Technology, Kalyani Govt. Engineering College, Kalyani, Nadia, India.
Email: arunita17@gmail.com.
4Skybound Digital LLC, Kolkata, West Bengal, India. Email: swarnajit32@gmail.com.
5Dept. of Engg. and Technological Studies,University of Kalyani, Kalyani, India, Email: dassanjoy0810@hotmail.com
ABSTRACT processing significantly assists pathologists and has attracted
many attentions in both research and clinical practice.
Breast histopathological image segmentation is exigent due to the A critical requirement in computer-aided diagnosis is
existence of imperceptibly correlated and indistinct multiple segmentation, which is typically measured as the basis of
regions of concern. Clustering based segmentation is one of the automated image analysis. It provides assistances for various
most significant approaches to perform proper segmentation of quantitative analyses such as shape, size, texture, and other
such complex images. K-means is the well-known clustering imagenomics [5, 6]. However, it is difficult to achieve stout and
techniques but very sensitive to initial cluster centers and easy perfect pathological image segmentation as these images
convergences to local optima. Therefore, researchers are frequently reveal background clutter with many noises, artifacts
employing Nature-Inspired Optimization Algorithms (NIOA) in such as blurred regions introduced during image acquisition, and
this domain. This study develops Cuckoo Search (CS) algorithm poor contrast between the foreground and the background.
based image clustering model for the proper segmentation of Second, there exist significant variations on cell size, shape, and
breast histopathology images. Experimental results show that CS intracellular intensity heterogeneity [5, 6].
provides better-quality segmented images compare to classical K-
means algorithm by considering the computational time, fitness In this study, proper segmentation of breast
values and the values of quality parameters. histopathology images is the main aim. Many efforts have been
performed to attain automated segmentation of breast
KEYWORDS histopathology images which includes thresholding [7, 8],
watershed method [9, 10], Active Contour model [8, 11], edge
Clustering, K-means, Image Segmentation, Optimization, Swarm based approach [14], neural network [15, 16] etc. A Particle
intelligence, Histopathology image. Swarm Optimization (PSO) with Otsu criterion based multi-level
thresholding technique was proposed by Jothi and Rajam [7] to
1 Introduction automatically segment the nuclei from hematoxylin and eosin
(H&E) – stained breast histopathology images. To remove noise,
Breast Cancer is the mainly widespread kind of cancer in women the input image filtered by 3x3 gaussian filter. Experimental result
worldwide [1]. Present breast cancer clinical practice and proved that this method automatically segmented the nuclei
treatment mostly depend on the evaluation of the disease’s effectively. A Localized Active Contour Model (LACM) in
diagnosis. Bloom-Richardson grading system [2] describes the conjunction with an automatic technique for optimal curve
scoring of three morphological features of the dubious tissue, placement with Krill Herd Algorithm (KHA) was developed by
which are fraction of tubule construction, amount of nuclear pleo- Beevi et. al. [8] for the segmentation of nuclei from breast
morphism, and mitotic cell count. However, the scoring had been histopathology images. Based on Hausdorff (HD) and Maximum
performed by pathologists based on the visual assessment of the Address Distance (MAD) measures segmentation performance
tissue’s biopsy sample under the microscope [3]. Therefore, was investigated. The proposed segmentation approach provided
researchers concentrate and suggest the use of image analysis superior results compared to GA, Bacterial Foraging Algorithm
methods to mitigate the said issue [4]. Digital histopathology and (BFA), Harmony Search (HS) algorithm and FCM clustering
microscopy images carry out an important role in decision making method by considering the convergence rate, objective function
in disease diagnosis, as they could provide wide information for values, computational time, sensitivity, precision and F-score.
computer-aided diagnosis (CAD), which facilitates quantitative Shen et. al. [9] developed one improved watershed algorithm by
analysis of digital images with a high throughput processing rate
[5, 6]. At present, analysis of digital histopathology through image
StuCoSReC Proceedings of the 2018 5th Student Computer Science Research Conference DOI: https://doi.org/10.26493/978-961-7055-26-9.47-54 47
Ljubljana, Slovenia, 9 October
Algorithm
Krishna Gopal Dhal1, Iztok Fister Jr.2, Arunita Das3, Swarnajit Ray4, Sanjoy Das5
1 Dept. of Computer Science and Application, Midnapore College (Autonomous), Paschim Medinipur, West Bengal, India.
Email: krishnagopal.dhal@midnaporecollege.ac.in
2Faculty of Electrical Engg. and Computer Sc., University of Maribor, Slovenia, Email: iztok.fister1@um.si.
3Dept. of Information Technology, Kalyani Govt. Engineering College, Kalyani, Nadia, India.
Email: arunita17@gmail.com.
4Skybound Digital LLC, Kolkata, West Bengal, India. Email: swarnajit32@gmail.com.
5Dept. of Engg. and Technological Studies,University of Kalyani, Kalyani, India, Email: dassanjoy0810@hotmail.com
ABSTRACT processing significantly assists pathologists and has attracted
many attentions in both research and clinical practice.
Breast histopathological image segmentation is exigent due to the A critical requirement in computer-aided diagnosis is
existence of imperceptibly correlated and indistinct multiple segmentation, which is typically measured as the basis of
regions of concern. Clustering based segmentation is one of the automated image analysis. It provides assistances for various
most significant approaches to perform proper segmentation of quantitative analyses such as shape, size, texture, and other
such complex images. K-means is the well-known clustering imagenomics [5, 6]. However, it is difficult to achieve stout and
techniques but very sensitive to initial cluster centers and easy perfect pathological image segmentation as these images
convergences to local optima. Therefore, researchers are frequently reveal background clutter with many noises, artifacts
employing Nature-Inspired Optimization Algorithms (NIOA) in such as blurred regions introduced during image acquisition, and
this domain. This study develops Cuckoo Search (CS) algorithm poor contrast between the foreground and the background.
based image clustering model for the proper segmentation of Second, there exist significant variations on cell size, shape, and
breast histopathology images. Experimental results show that CS intracellular intensity heterogeneity [5, 6].
provides better-quality segmented images compare to classical K-
means algorithm by considering the computational time, fitness In this study, proper segmentation of breast
values and the values of quality parameters. histopathology images is the main aim. Many efforts have been
performed to attain automated segmentation of breast
KEYWORDS histopathology images which includes thresholding [7, 8],
watershed method [9, 10], Active Contour model [8, 11], edge
Clustering, K-means, Image Segmentation, Optimization, Swarm based approach [14], neural network [15, 16] etc. A Particle
intelligence, Histopathology image. Swarm Optimization (PSO) with Otsu criterion based multi-level
thresholding technique was proposed by Jothi and Rajam [7] to
1 Introduction automatically segment the nuclei from hematoxylin and eosin
(H&E) – stained breast histopathology images. To remove noise,
Breast Cancer is the mainly widespread kind of cancer in women the input image filtered by 3x3 gaussian filter. Experimental result
worldwide [1]. Present breast cancer clinical practice and proved that this method automatically segmented the nuclei
treatment mostly depend on the evaluation of the disease’s effectively. A Localized Active Contour Model (LACM) in
diagnosis. Bloom-Richardson grading system [2] describes the conjunction with an automatic technique for optimal curve
scoring of three morphological features of the dubious tissue, placement with Krill Herd Algorithm (KHA) was developed by
which are fraction of tubule construction, amount of nuclear pleo- Beevi et. al. [8] for the segmentation of nuclei from breast
morphism, and mitotic cell count. However, the scoring had been histopathology images. Based on Hausdorff (HD) and Maximum
performed by pathologists based on the visual assessment of the Address Distance (MAD) measures segmentation performance
tissue’s biopsy sample under the microscope [3]. Therefore, was investigated. The proposed segmentation approach provided
researchers concentrate and suggest the use of image analysis superior results compared to GA, Bacterial Foraging Algorithm
methods to mitigate the said issue [4]. Digital histopathology and (BFA), Harmony Search (HS) algorithm and FCM clustering
microscopy images carry out an important role in decision making method by considering the convergence rate, objective function
in disease diagnosis, as they could provide wide information for values, computational time, sensitivity, precision and F-score.
computer-aided diagnosis (CAD), which facilitates quantitative Shen et. al. [9] developed one improved watershed algorithm by
analysis of digital images with a high throughput processing rate
[5, 6]. At present, analysis of digital histopathology through image
StuCoSReC Proceedings of the 2018 5th Student Computer Science Research Conference DOI: https://doi.org/10.26493/978-961-7055-26-9.47-54 47
Ljubljana, Slovenia, 9 October
