ADVANCES in NATURAL and APPLIED SCIENCES

Transcription

1 ADVANCES in NATURAL and APPLIED SCIENCES ISSN: Published BYAENSI Publication EISSN: July 11(9):pages 297- Open Access Journal Survey On Identification Of Brain Tumor Using Medical Image Segmentation And Edge Detection 1 S. Shanmuga Priya and 2 Dr.A.Valarmathi 1 Assistant Prof, Department of CSE, MIET Engineering College, Trichy, Tamil Nadu 2 Assistant Prof & Head, Department of Computer Application Anna University Trichy, Tamil Nadu Received 2017; Accepted 2017; Available online Address For Correspondence: S. Shanmuga Priya, Assistant Prof, Department of CSE, MIET Engineering College, Trichy, Tamil Nadu Copyright 2017 by authors and American-Eurasian Network for ScientificInformation (AENSI Publication). This work is licensed under the Creative Commons Attribution International License (CC BY). ABSTRACT Biomedical images are an important feature in the field of medical science. MRI- Magnetic Resonance Imaging is one of the biomedical images that captures the fine details of the internal structure of human brain and also used to picture other parts of the body. In case of brain tumor, MRI data are captured manually by medical experts and it s a tedious and cumbersome task to be performed. MRI used for brain tumors use the concept of edge detection in order to produce high definition images. Several researches and techniques have been proposed for automatic brain tumor detection from MRI scan image. Few of these techniques like thresholding and artificial neural network (ANN) algorithms, region based, K-means, fuzzy clustering means (FCM), sobel edge and canny edge have been discussed in this paper. The different segmentation methods are reviewed and a comparison study is evaluated on the basis of performance and application of the technique in the MRI image. Finally we have also discussed on the culmination and future aspects of brain tumor segmentation methods. KEYWORDS: Brain tumor, Image Segmentation, Threshold based, Region based, Artificial Neural Networks, FCM, edge detection, canny, sobel. INTRODUCTION Tumors can happen in any parts of the body. Brain tumor can be considered as one of the severe and lifethreatening tumors. It is created by the uncontrolled and irregular cell division inside the brain or from cancers that is present in other parts of the body. In most cases, Tumors are divided depending on the locality of their origin and its malignancy. Brain tumor division is one the focused task to analyze the characteristics of tumor in planning for the medical treatment. In medical terms, brain tumor referred to as Intracranial Neoplasm, is caused due to the abnormal development of brain tissues. Brain tumors are of two types: Primary brain tumors and Metastatic brain tumors. The previous develops in the brain and stay there only, the last which start as a cancer in any parts of the body and finally invades the brain. Brain tumor s death rate has increased and studies demonstrate that around 90% of tumors are observed to be glial tumors over 20 years. Brain tumors can change based on the components like location, shape, size and image intensities. In the field of biomedical imaging, the division of tumor from the human brain has turned into a new range of research and hence different literature is accessible in the field. For the diagnosis and treatment of the patient suffering from brain tumor, specialists take the assistance of the MRI scans of the brain. In any case, the analysis of the MRI scan is done manually by the specialist who is tedious and the precision of the outcome ToCite ThisArticle: S. Shanmuga Priya and Dr.A.Valarmathi., Survey On Identification Of Brain Tumor Using Medical Image Segmentation And Edge Detection. Advances in Natural and Applied Sciences. 11(9);Pages: 297-

2 298 S. Shanmuga Priya and Dr.A.Valarmathi., 2017/Advances in Natural and Applied Sciences. 11(9) July 2017, Pages: 297- depends on the experience of the specialist. The conclusions may differ from one doctor to another. Thus, there is a need to overcome these issues and to automate or robotize the investigative procedure of brain tumor in MRI images. For this, biomedical image processing techniques are applied to the MRI scans. Thus, the segmentation and further characterization of brain tumor from the MRI scans remains a broad range of research in the field of medical science. Alongside the progress of medical imaging, imaging modalities serves as critical part in the assessment of patients with brain tumors and have an impact on patient care. Recent years, the developing new imaging modalities, for example, X-ray, Ultrasonography, Computed Tomography (CT), Magneto Encephalo Graphy (MEG), Electro Encephalo Graphy (EEG), Positron Emission Tomography (PET), Single-Photon Emission Computed Tomography (SPECT), and Magnetic Resonance Imaging (MRI), not just demonstrate the complete aspects of brain tumors, additionally enhance clinical specialists to concentrate the instrument of brain tumors at the point of better treatment. Clinical specialists play a vital role in brain tumor assessment and treatment. Once a brain tumor is clinically suspected, radiologic assessment is required to locate the area of infection, the degree of the tumor and its relationship to the surrounding structures. This data is essential and critical in deciding between distinctive types of therapy, like surgery, radiation, and chemotherapy. Therefore, the assessment of brain tumors with imaging modalities is now one of the key issues of radiology divisions. Edges of an image contain significant variation of intensity. Edges demonstrate the limits between an object and the background of an image, which helps in division and object recognition. Edge detection is useful in image segmentation which distinguishes whether a line or an edge is present or not and depicts them in a suitable way. It is defined as the process of recognizing and locating sharp discontinuities, limits of objects or textures depicts in an image (i.e. edges) These discontinuities are rapid variations in pixel intensity which describe objects limits in an image. The aim of edge detection is to extract the important elements like lines, corners, curves etc. from the edges of an image. The main motto of edge detection is to eliminate the unwanted information and preserves the essential information and thus it decreases the amount of data needs to be processed. The important features of edges are position of subarea, amplitude and direction. Based on these features, the detector has to decide whether each of the inspected pixels is an edge or not. In this paper, different existing strategies for identification and division of brain tumor from MRI image i.e. thresholding based, edge based, region based and grouping based division have been discussed. Segmentation: Image segmentation is the process of dividing a digital image into multiple segments (sets of pixels, also known as super pixels). The aim of segmentation is to simplify and/ or change the representation of the image into something that is more meaningful and easier to investigate. Image segmentation is used to locate objects and boundaries (lines, curves, etc.) in images. More precisely, image segmentation is the process of assigning a label to every pixel in an image such that pixels with similar label share certain visual qualities. In case of medical image segmentation the aim is to Study anatomical structure Identify Region of Interest i.e. locate tumor, lesi on and other abnormalities Measure tissue volume to measure growth of tumor (also decrease in size of tumor with treatment) Help in treatment planning prior to radiation therapy; in radiation dose calculation Using segmentation in therapeutic images is a vital task for detecting the abnormalities, study and tracking progress of diseases and surgery planning. Brain Tumor Segmentation Methods: Intuitionistic Fuzzy Approach: In the Fuzzy mathematical morphology has used Hamacher t-norm and t-co norm using the fuzzy set theory. The objective of this approach is to convert an image into intuitionistic fuzzy image and then use the morphological operators from the Sugeno's fuzzy complement or intuitionistic fuzzy generator [1]. K means clustering: The K-means algorithm initiates the mean vector iteration that is specific to each K clusters. Secondly the class which holds the pixel vector closest to the mean vector is considered. Based on the above criteria each pixel will be placed in a training set. The pixel set will now form the decision boundaries. The iteration will repeat again where the clusters will be formed from the new pixel group and then the cluster mean vector is computed. As the iterations continue finally k-means will concentrate the data in nearby regions of the feature space. The number of iterations depends on the state where there are no further changes in the pixel movement or assignment. Finally there will be a stage where the centroid of each cluster no longer changes and is said to be stable. Hence the threshold value is computed from the final value of centroid [2].

3 299 S. Shanmuga Priya and Dr.A.Valarmathi., 2017/Advances in Natural and Applied Sciences. 11(9) July 2017, Pages: 297- Fuzzy Inference System: Fuzzy Inference System (FIS) is used for edge detection in an image. While we have other edge detection techniques like Rothwell, Nalwa Binford, Canny, Iversopn, Bergholm, etc it is observed that these methods depend on the choice of the input parameters. These input parameters may be the threshold values as well. Whereas Fuzzy logic based edge detection is an adaptive thresholding method where the threshold values are decided based on certain fuzzy rules [2]. Hybrid Segmentation Approach: Hybrid segmentation is used for image segmentation. In case of tumor it helps in identifying the targeted brain tumor image segmentation. The problem of identifying single seed region growing hole is solved using Hybrid segmentation. The results can be combined using both single threshold segmentation and single seed region growing [6]. Seed Region Growing: This method is to identify the seed region growing and it is performed using single seed point or pixel. The single seed point or the pixel is taken and all the pixels surrounding this region is considered and forms the region called r. The following points are considered for region growing segmentation. 1. The seed point must be selected automatically. 2. There should be minimum distance between the seed point and the neighbor pixels. 3. There must be at least one seed generated in the region in order to predict that region. 4. Seeds from different regions must be disconnected. The minimum pixel distance is taken as a default process. The small region will then grow iteratively. This is done by evaluating all non-distributed neighboring pixels. The variation considered between the mean of the region and the value of pixel intensity is used as a measure of similarity. The pixel that holds the minimum difference is allocated to the particular region. The iteration will stop or end when the variation in the intensity between the region mean and new region become larger than a certain seed. The final output is produced by combining these regions. So the single seed region is formed as a binary image [6]. Otsu's thresholding: Ostsu's threshold method is used for the selection of threshold values. This method helps in deciding if the pixels fall in the foreground or background by considering all possible threshold values. It uses the concept of variance to measure the regions homogeneity. It minimizes the within-class variance of the two groups of pixels that are separated by a threshold value. This method is proved to be minimum error method [12]. Watershed segmentation: Watershed segmentation is used to partition the pixels based on the intensity of the pixel. It is the best method to group the pixels as each pixel has a different intensity level. Pixels that share close similarity are grouped together and in this way the tumor in the image is separated from the brain MRI. It exactly separates the part that is affected with tumor from the rest of the brain. It is a morphological operating tool where the pixels are grouped based on intensities and is the best way to separate the tumor region from an image [12]. Artificial Neural Networks: Artificial Neural Network (ANN) is a system model that is created based on the human brain. It is a network of multiple units having a local memory attached to each unit. It trains itself by learning from data and does not use any rules to perform classification. ANN performs well even in case of non linear, difficult, and multivariate and noise domains like brain tumor segmentation. The classifier inputs the features through a series of nodes and mathematical operations are applied on the input nodes after which the classification happens [20]. Edge based segmentation: This is a most common edge detection method used in medical images. It helps in creating a boundary that separates distinct regions from the rest of the image. This method is based on the markings of discontinuities in color, gray level, etc. These markings make the boundary between the objects. There are several methods for edge detection in brain tumor like prewitt edge detection operator, sobel edge detection and canny edge detection operator. Cellular Automata Based Edge Detection: Cellular Automata is based on a component called cell. We all are aware of the fact that each cell only has two states namely 0 and 1. The entire computation involved in this method is based on these states of cells. A cell is usually covered with a set of neighbors and initially all cells are to be in the state 0. Only after meeting

4 300 S. Shanmuga Priya and Dr.A.Valarmathi., 2017/Advances in Natural and Applied Sciences. 11(9) July 2017, Pages: 297- certain criteria the state of the cell will change to 1. These conditions are specified by different rule numbers. In this method they also use the 2D cellular automata based on moore neighborhood model. According to the moore neighborhood model the state of the cell in the next generation depends on its 8 neighbors and the cell itself [3]. Sobel edge detection: This method is based on an endeavor of 3X3 convolution portions. One bit of this portion is turned to 90 degree. These bits will respond to the edges running vertically and on level plane considering the pixel cross section. These areas can be connected to frame the information picture to approximate estimation of the slant part if the presentation (Gx and Gy). Finally these can be put together to calculate the aggregate level of edge at each point and the presentation of the slant [15]. Canny Edge Detection: In order to detect the edges in the picture by eliminating the clamor is done through Canny edge detection. It is a superior technique that detects the edges without exasperating the components of edges. It follows algorithmic steps that involves Convolve picture f(r,c) and a Gaussian capacity to achieve a smooth image [15]. Prewitt edge detection: This method uses a 3X3 kernel that represents the original image in order to calculate the approximation of the derivatives. It calculates one for vertical changes and other of horizontal changes. Let's consider A as the source image, and then Gx and Gy are the two images say the horizontal and vertical derivative approximations. It is denoted as a 2-dimensional convolution operation. Now the horizontal edge is calculated using the kernel Gy and the vertical edge is calculated using the kernel Gx and the gradient magnitude is given by: Outline Of Brain Tumor Segmentation Methods: Segmentation Methods Merits Demerits Intuitionistic Fuzzy Approach Fuzzy C Means Threshold based Artificial Neural Networks Region Based Watershed segmentation Hybrid Segmentation Approach It uses minimum amount of defuzzification that helps in the prevention of loss of information to a great extent. It belongs to the unsupervised category and it converge the tumor boundaries. It is fast and simple in computation and involves less complexity. It has the capability to model non trivial distributions and nonlinear dependence [7]. The main advantage is that it correctly segments regions that have similar properties and produces connected region [21]. This approach is easy, simple, instinctive knowledge, and can be parallelized.[28] It delivers more efficiency, better quality and accuracy of image. It is highly sensitive to noise and outliers It takes long period for computation and it is sensitive to noise [24]. Its application to the tumor is limited [23]. It does not consider the spatial domain and hence the regions may not be connected. It is a tedious task to gather the training samples and the learning process is slow [25]. It is expensive and involves high computation time and memory. Partial Volume effect [22]. This methods lands up in over segmentation due to the presence of many local minima.[28] It uses demographics as a basis for segmentation and this is the main drawback of this approach.

5 301 S. Shanmuga Priya and Dr.A.Valarmathi., 2017/Advances in Natural and Applied Sciences. 11(9) July 2017, Pages: 297- Outline Of Edge Based Brain Tumor Segmentation Methods: Edge segmentation Methods Merits Demerits Cellular Automata Based Edge Detection This method outperforms the other methods and it considers only linear rules of CA for extraction of edges under null boundary condition.[26] Sobel edge detection This method is simple and provides a approximation to the gradient magnitude. It is capable of detecting edges and their orientations canny edge detection Major features of this method are the smoothing effect used for removing noise and improve the signal to noise ratio with the help of non-maximal suppression. Prewitt edge detection It helps in finding the directions of gradient magnitude as the approximation of gradient magnitude is easy[29] It involves high computational complexity[27] It is highly sensitive to noise. It is time consuming. This method is unreliable, noise susceptibility in edge detection and in direction of gradient magnitude [29] Conclusions: This paper has demonstrated an overview of the various state-of-the-art techniques used in MRI-based brain tumor segmentation methods. MRI is preferred for brain tumor detection due to the fact that it is non-invasive and provides good soft tissue contrast. Another major advantage of using MRI for tumor scan is that it is fast in detecting the tumor area, accurate and reproducible. Brain tumor segmentation technique has proved to be best in detecting and analyzing tumor in medical images and further research works will continue in future. REFERENCES 1. Mohammed, Y. Kamil, Morphological Gradient in Brain Magnetic Resonance Imaging Based on Intuitionistic Fuzzy Approach, in Al-Sadeq International Conference on Multidisciplinary in IT and Communication Science and Applications (AIC-MITCSA) IRAQ. 2. Ms. Neha Mathur, Mr. Pankaj Dadheech, Mr. Mukesh Kumar Gupta, THE K-MEANS CLUSTERING BASED FUZZY EDGE DETECTION TECHNIQUE ON MRI IMAGES, in Fifth International Conference on Advances in Computing and Communications. 3. Charutha, S., M.J. Jayashree, An Efficient Brain Tumor Detection By Integrating Modified Texture Based Region Growing And Cellular Automata Edge Detection in International Conference on Control, Instrumentation, Communication and Computational Technologies (ICCICCT). 4. Shang-Ling Jui, Shichen Zhang, Weilun Xiong, Fangxiaoqi Yu, Mingjian Fu, Dongmei Wang, Aboul Ella Hassanien and Kai Xiao, Brain MR Image Tumor Segmentation with 3-Dimensional Intracranial Structure Deformation Features, IEEE INTELLIGENT SYSTEMS. 5. Kimmi Verma, Aru Mehrotra, Vijayeta Pandey, Shardendu Singh, IMAGE PROCESSING TECHNIQUES FOR THE ENHANCEMENT OF BRAIN TUMOR PATTERNS, in International Journal of Advanced Research in Electrical, Electronics and Instrumentation Engineering, 2: Anithadevi, D. and K. Perumal, A HYBRID APPROACH BASED SEGMENTATION TECHNIQUE FOR BRAIN TUMOR IN MRI IMAGES. 7. Tao Wang, Irene Cheng and Anup Basu, Fluid Vector Flow and Applications in Brain Tumor Segmentation, in IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, 56: Andac Hamamci, Nadir Kucuk, Kutlay Karaman, Kayihan Engin, and Gozde Unal, Tumor-Cut: Segmentation of Brain Tumors on Contrast Enhanced MR Images for Radiosurgery Applications, in IEEE TRANSACTIONS ON MEDICAL IMAGING, 31: S ergio Pereira, Adriano Pinto, Victor Alves and Carlos A. Silva, Brain Tumor Segmentation using Convolutional Neural Networks in MRI Images, in 2016, IEEE Transactions on Medical Imaging. 10. Anamika Ahirwar, Study of Techniques used for Medical Image Segmentation and Computation of Statistical Test for Region Classification of Brain MRI, in I.J. Information Technology and Computer Science, 05: Rohini Paul Joseph1, C. Senthil Singh2, M. Manikandan3, BRAIN TUMOR MRI IMAGE SEGMENTATION AND DETECTION IN IMAGE PROCESSING, in IJRET: International Journal of Research in Engineering and Technology, 03: Shubhangi, S. Veer (Handore)1, Dr. P.M. Patil, An efficient method for Segmentation and Detection of Brain Tumor in MRI images, in International Research Journal of Engineering and Technology (IRJET) 02: 09.

6 S. Shanmuga Priya and Dr.A.Valarmathi., 2017/Advances in Natural and Applied Sciences. 11(9) July 2017, Pages: Atiq Islam, Syed M.S. Reza and Khan M. Iftekharuddin, Multi-fractal Texture Estimation for Detection and Segmentation of Brain Tumors. 14. Kadkhodaei, M., S. Samavi, N. Karimi, H. Mohaghegh, S.M.R. Soroushmehr, K. Ward, A. All, K. Najarian, Automatic Segmentation of Multimodal Brain Tumor Images Based on Classification of Super-Voxels. 15. Gopal Gupta and Sandeep K Tiwari, An Improved Biomedical Edge Detection Based on Dynamic Subtraction Method, in International Conference on Electrical, Electronics, and Optimization Techniques (ICEEOT). 16. Aleksandar Stojak, Eva Tuba, Milan Tuba, Framework for Abnormality Detection in Magnetic Resonance Brain Images. 17. Parveen, Amritpal singh, Detection of Brain Tumor in MRI Images, using Combination of Fuzzy C- Means and SVM, 2nd International Conference on Signal Processing and Integrated Networks (SPIN). 18. Jin Liu, Min Li, Jianxin Wang_, Fangxiang Wu, Tianming Liu, and Yi Pan, A Survey of MRI-Based Brain Tumor Segmentation Methods, in Tsinghua Science and Technology, 19: Xiaoli Zhang1, Xiongfei Li1, Hongpeng Li2, and Yuncong Feng, A SEMI-AUTOMATIC BRAIN TUMOR SEGMENTATION ALGORITHM. 20. Ashima Anand, Harpreet Kaur, Survey on Segmentation of Brain Tumor: A Review of Literature, in International Journal of Advanced Research in Computer and Communication Engineering, 5: Salman, Y., A. Badawi, M. Assal, S. Alian, New automatic technique for tracking brain tumor response.int Conf Biol Med Phys, p: Sato, M., S. Lakare, M. Wan, A. Kaufman, A gradient magnitude based region growing algorithm for accurate segmentation. Int Conf Image Process, 3: Gibbs, P., D. Buckley, S. Blackb, M.R. Horsman, images by morphological segmentation. Phys Med Biol A. Tumour determination from, 41: Kannan, S., A new segmentation system for brain MR images based on fuzzy techniques. Appl Soft Comput, 8: Iftekharuddin, K., J. Zheng, M. Islam, Ogg R. Fractal-based brain tumor detection in multimodal MRI. Appl Math Comput, 207: Deepak Ranjan Nayak, A Cellular Automata based Optimal Edge Detection Technique using Twenty-Five Neighborhood Model in International Journal of Computer Applications. 27. Xiaoli Zhang, A SEMI-AUTOMATIC BRAIN TUMOR SEGMENTATION ALGORITHM, IEEE Transactions on Image Processing. 28. Amanpreet kaur, The Marker-Based Watershed Segmentation- A Review, International Journal of Engineering and Innovative Technology (IJEIT), 3: Monica Avlash, PERFORMANCES ANALYSIS OF DIFFERENT EDGE DETECTION METHODS ON ROAD IMAGES, international Journal of Advanced Research in Engineering and Applied Sciences, 2: 6.