Survey on Different Brain Tumor Detection Methods or Algorithms

Transcription

1 Volume-5, Issue-5, October-2015 International Journal of Engineering and Management Research Page Number: Survey on Different Brain Tumor Detection Methods or s Shivani Goyal 1, Nancy Johari 2, Natthan Singh 3, Anita Pal 4 1,2 Student, Bansthali University, Jaipur, Rajasthan, INDIA 3 Assistant Professor, Institute of Engineering and Technology, Lucknow, Uttar Pradesh, INDIA 4 Research Scholar, Institute of Engineering and Technology, Lucknow, Uttar Pradesh, INDIA ABSTRACT Image processing is one of the developing research areas today. Medical image processing is the most vindicate and highly needed field in that. Brain tumor is an unsuppressed expension (growth) of tissues in human brain. This tumor, when turns in to cancer become critical. So, medical imaging, it is necessary to detect the exact location of tumor and its type. For locating tumor in magnetic resonance image (MRI) segmentation of MRI plays an crucial role. The method of brain tumor segmentation is nothing but the differentiation of different tumor area from MR images. Segmentation of images is one of the most troublesome tasks thus holds an important position in image processing which determines the quality of final result. Image segmentation is a process of dividing an image into various homogeneous domain (regions). In this paper, we constitute survey on different segmentation techniques applied to MR images for locating tumor. In this paper we also discuss the advantage and disadvantage of different methods of segmentation. Keywords Brain tumor, Magnetic Resonance Imaging (MRI), Segmentation, Thresholding, Tumor detection I. INTRODUCTION The Brain is the most important part of the central nervous system. For images of human brain different techniques are used to capture image. These techniques involve X-Ray, Computer Tomography (CT) and MRI. For diagnosis, MRI has more advantage over other techniques such as MRI provides rich information about anatomical structure, enabling quantitative pathological or clinical studies. The structure and function of the brain is need to studied non invisible by doctors and researchers using MRI imaging techniques. The body is made up of various type of cells. Each type of cell has special function. When cells loose their capability to control their growth, they devise regularly and without any order. The extra cells form a mass of tissue called a tumor. MRI acts as an assistant diagnostic tool for the doctor during disease diagnosis and treatment. This imaging modality produces images on soft tissues. The acquired medical images show the internal structure, but the doctors want to know more than coequal images, such as accentuate the peculiar tissue, quantifying its size, characterizing its shape, and so on. If such tasks are covered by the doctors themselves, it may be unreliable, time consuming and load them heavily. So computer aided system can be designed for accurate brain tumor detection from MRI images. Brain tumor can be broadly classified as primary brain tumor i.e. the tumor originates in brain and secondary brain tumor i.e. spread to brain from somewhere else in the body through metastasis. Primary brain tumors do not spread to other body parts and can be malignant or benign and secondary brain tumors are always malignant. Malignant tumor is more risky and serious than benign tumor. Detection is also difficult of benign tumor than malignant tumor. For the accurate detection of the malignant tumor that needs a 3-D representation of brain and 3-D analyser tool. II. LITERATURE SURVEY AND RELATED RESEARCH A number of research paper related to medical image segmentation methods were studied. A report of literature survey is present here. Yongyue Zhang et al (2001) proposed segmentation of brain MR images through a hidden markov random field model and the expectation-maximization. This developed shows ability to encode both spatial and statistical properties of an image [1]. Ahmed et al (2002) introduced a modified fuzzy c-means for bias field estimation and segmentation of MRI data. This developed shows how fast the results we can generate [2]. Tolba et al (2003) proposed MR- brain image segmentation using gaussiuan multi resolution analysis and the EM. This proposed method shows how the results are less sensitive to noise [3]. Sing et al(2005) presented segmentation of MR images of human brain using fuzzy adaptive radial basis function neural network[4]. This introduced method shows how we can preserve the sharpness of an image. Sasi kala et al 94 Copyright Vandana Publications. All Rights Reserved.

2 (2006) presented an automatic segmentation of malignant tumor in MR images of brain using optimal texture features. Texture features are extracted from normal and tumor regions in the brain images under study using spatial grey level dependence method and wavelet transform. Eman abdel-maksud (2015) proposed brain tumor segmentation based on a hybrid clustering technique. This proposed method shows minimization of time to calculating the results [10]. Kailash sinha et al (2014) introduced efficient segmentation methods for tumor detection in MRI images. It shows convergence and computing time reduced through this proposed work [11]. Gauri P. anandgaonkar et al detection and identification of brain tumor in brain MR images using fuzzy C-means segmentation. This proposed work shows wheather the tumor is benign or maliganant [12]. III. SEGMENTATION TECHNIQUES The main objective of this section is to describe about the segmentation techniques which are useful and helpful in detection and identification of brain tumor in MR images. A. K-Means Clustering K-Means Clustering partition n observation into k clusters in which each pixel belongs to the clusters by minimizing an objective function in a way that the within cluster sum of squares is get minimized. It starts within initial K cluster centres and it reassigns the observations to clusters based on the similarity between the observations and cluster centre. K-Means Clustering is a two step process. In the first step, skull stripping is performed by generating a skull mask from the MRI image and in the second step, an advanced K-means improvised by two level granularity oriented grid based localization process based on standard local deviation is used to segment the image into grey matter, white matter and tumor region and then length and breadth of the tumor is assessed [16]. B. Region growing In this segmentation technique the images are partitioned by organizing the nearest pixel of similar kind. It starts with a pixel that having similar properties. Accordingly the neighbouring pixels based on homogeneity criteria are appended progressively to the seed. In splitting process, region get divided into subregions that do not satisfy a given homogeneity criteria. Splitting and merging can be used together and its performance mostly depends on the selected homogeneity criterion. Without tuning homogeneity parameters, the seeded region growing technique is controlled by a number of initial seeds. If the number of regions was approximately known and used it to estimate the corresponding parameters of edge detection [16] C. Soft Computing A self-organizing map(som) or self organizing feature map is a type of artificial neural network for unsupervised learning. Self arranging map arrange in training and mapping mode. Training process builds map using vector quantization process and mapping automatically classifies a new input vector. SOM map includes neurons and nodes. Self organizing maps each of which are neurons associated with a weight vector map data input vectors and position in the map space. The self organizing maps a higher dimensional input space to a lower dimensional map space. Energy, entropy, contrast, mean, median, variance, correlation, maximum and minimum intensity values used to provide clear description of tumor [16]. D. Fuzzy C-Means Clustering Fuzzy C-means clustering is a data clustering method in which each data point belongs to a cluster to a degree specified by a membership value. Fuzzy C- means divides a collection of n vectors into c fuzzy groups and finds a cluster center in each group such that a cost function of dissimilarity measure is minimized [16]. E. Image/ Symmetry analysis: This analysis is an interactive segmentation method that in addition to area of the region and edge information uses prior information, also its symmetry analysis which is more consistent in pathological cases. A conceptually easy supervised block based, shape texture; content based technique has been analyse MRI brain images with relatively lower computational requirements. Classifying regions by means of their multi parameter values does the study of the regions of physiological and pathological interest easier and more definable. F. Thresholding: Thresholding is one of the most common and oldest method of image segmentation. In this method, image is supposed to be composed of regions and these regions belong to different ranges of grey scale. In the histogram of any image there is various peaks and valleys, in which each peak consider as one region and the velly between the peaks shows a threshold value. Histogram thresholding method is generally based on concept that divides the image into two equal halves and histograms are compared to detect the tumor and cropping method is used to find a proper physical dimension of brain tumor. The threshold technique makes decision based on the local raw pixel information. It helps in extracting the basic shape of an image, overlooking the title unnecessary details [16]. Figure 1: Histogram of brain MRI image 95 Copyright Vandana Publications. All Rights Reserved.

3 This idea can be fruitfully cashed in the digital images. The image gradient can be viewed as terrain. The homogeneous regions Figure 2: Result of thresholding of brain MRI image G. Optimized C-Means Clustering MRI images of brain are segmented and tumor can be efficiently extracted and detected is c-means clustering is optimized. The c-means clustering method has been implemented and its performance can be improved by using optimization with the use of genetic. The combined method results an improvement in segmentation efficiency and higher area of affected region extraction and detection. The aim of optimized clustering and its analysis is to divide a given set of data into a number of clusters which should follow the properties given below: i. Homogeneity inside clusters: The information or the data of a cluster is a similar as possible. ii. Heterogeneity between the clusters: Here the data belongs to different clusters are different. H. Genetic The term Genetic is derived from Greek word genesis which means to grow or to become, and therefore the makes a function grow. This was introduced by John Holland on the basis of a heuristic method. The method grows in search of survival of the fittest. Since fittest is searched by the and hence used in optimization tasks, The implementation of genetic begins with an initial population of chromosomes which are randomly selected. A chromosome is a long thread of DNA (deoxyribonucleic acid). Particulars traits determine the hereditary of an individual where each trait is coded by some combination of DNA bases. The four main bases of DNA are A (Adenine), C (Cytosine), T (Thymine) and G (Guanine). Just like English alphabet, the combinations of various letters give some meaningful information; GA also follows the same concept. I. Meta Heuristic A meta heuristic is a set of ic concepts that can be used to define heuristic methods applicable to a wide set of different problems [13]. In other words, a meta - heuristic is a general ic framework, which can be applied to different optimization problems with relatively few modifications to make them, adapted to a specific problem. The use of meta heuristics has significantly increased the ability of finding very high-quality solutions to hard, practically relevant combinatorial optimization problems in a reasonable time [13]. J. Watershed Watershed method comes under the edge-based method. The term watershed is a geographical one. In geography, a watershed line is defined as the line separating two catchment s basins. The rain that falls on either side of the watershed line will flow into the same lake of water. Figure 3: Example of watershed in the image usually have low gradient values. Thus, they represent valleys while the edges represent the peaks having high gradient values. The watershed transform is often preferred to separate the touching objects in an image. The watershed transform finds the catchment basins and watershed ridge lines in an image by treating it as a surface. The basic watershed is well recognized as an watershed transformation is that it produces a large number of segmented regions in the image around each local minima embedded in that image. A solution to sort out this problem is to introduce markers and flood the gradient image starting from these markers instead of regional minima efficient morphological segmentation tool which has been used in a variety of gray scale image processes and video processing applications. However, a major problem with the watershed transformation is that it produces a large number of segmented regions in the image around each local minima embedded in that image. A solution to sort out this problem is to introduce markers and flood the gradient image starting from these markers instead of regional minima. K. Ant Colony Optimization Ant Colony Optimization (ACO) is a recent population based approach is inspired by the observation of real ant colony and based upon their collective foraging behaviour Real ants are capable of finding the shortest path from a food source to the nest without using visual cues[12]. Ants are moving on a straight line that connects a food source to their nest is a pheromone trail. Pheromone is a volatile chemical substance lay down by ants while walking, and each ant probabilistically prefers to follow a direction rich in pheromone. This elementary behaviour of real ants can be used to obtain optimum value from a population. In ACO, solutions of the problem are constructed within a stochastic iterative process, by adding solution components to partial solutions. Each individual ant constructs a part of the solution using an artificial pheromone, which reflects its experience accumulated while solving the problem, and heuristic information dependent on the problem [12]. L. Particle Swarm Optimization Particle swarm optimization (PSO) is one of the modern heuristic s that can be applied to non linear and non continuous optimization problems. It is a population-based stochastic optimization technique for 96 Copyright Vandana Publications. All Rights Reserved.

4 continuous nonlinear functions. PSO learned from the scenario and used it to solve the optimization problems. Particle Swarm Optimization is an optimization technique which provides an evolutionary based search. This search was introduced by Dr Russ Eberhart and Dr James Kennedy in The term PSO refers to a relatively new family of s that may be used to find optimal or near to optimal solutions to numerical and qualitative problems. M. KNN(nearest neighbourhood) segmentation K-Nearest Neighbour (KNN) classification technique is the simplest technique that provides good classification accuracy. The KNN is based on a distance function and a voting function in K-Nearest Neighbour s, the metric employed is the Euclidean distance, Hamming distance, Minkowski distance. The KNN has higher accuracy and stability for MRI data than other common statistical classifiers, but has a slow running time [15]. Segmentation of brain tumor by KNN: I. Euclidian distance: This is used as distance function due to its simplicity. II. Hamming distance: This method detects edges in the image. The Hamming distance is a metric on the vector space of the words of length n, as it fulfils the conditions of non-negativity, identity of indiscernible and symmetry.it can be shown by complete induction that it satisfies the triangle inequality. III. Minkowski distance: This is a metric on Euclidean space which can be considered as a generalization of both the Euclidean distance and the Manhattan distance. IV. Cosine similarity: This detects intense parts of the tumour [15]. N. Edge Based Methods Edge detection is a well-developed field on its own within image processing. Region boundaries and edges are closely related, since there is often a sharp adjustment in intensity at the region boundaries. Edge detection techniques have therefore been used as the base of another segmentation technique [14]. The types of edge detection techniques are: I. Gradient Based Edge Detection: This method detects the edges by looking for the maximum and minimum in the first derivative of the image. The assumption is edges are the pixels with a high gradient. These are implemented with convolution mask and based on discrete approximations to differential operators. The quality of the edge image depends on the threshold. These are sensitive to noise and inaccurate [14]. II. Laplacian: The laplacian method searches for zero crossings in the second derivative of the image to find edges. An edge has the one-dimensional shape of a ramp and calculating the derivative of the image can highlight its location. The laplacian may be used when we are interested only in edge magnitudes without regard to their orientations. The Laplacian has the same properties in all directions and is therefore invariant to rotation in the image. The disadvantage is malfunctioning at the corners, curves and where the grey level intensity function varies. Not finding the orientation of edge because of using the Laplacian filter [14]. O. Spatial Clustering Different In image segmentation, the grouping of the image is carried out in spatial domain. In image clustering, grouping is performed in the measurement space. Overlapping regions can be the result of clustering. It is not possible to produce overlapping regions from segmentation. Clustering and spatial segmentation can be combined to form spatial clustering, which combine histogram techniques with spatial linkage techniques for better results. P. Split and Merge Segmentation The split method begins with the entire image, and repeatedly splits each segment into quarters if the homogeneity criterion is not satisfied. These splits can sometimes divide portions of one object. The merge method joins adjacent segments of the same object. It is important to distinguish the separate regions for intensity based segmentation so that over segmentation and undersegmentation of regions can be differentiated. Task of this kind can be performed using split segmentation or merge segmentation. If a region is not segmented fully, correction can be made by adding boundaries to, or splitting, certain regions that contain parts of different objects. If a region is segmented more that is necessary, correction can be made by eliminating false boundaries and merging adjacent regions if they belong to the same object or feature. Q. Hierarchical Clustering Method: There are many times when clusters have sub classes within them, and which in turn have sub-classes of thrir qwn examples of such occurrences can be found in biological taxonomy when we have a species in the animal kingdom, with a chordate phylum, a sub-phylum of vertebrata, and so on. Such classifications are hierarchical and proper partitions can be found by the following classification methods. More formally, a sequence is said to be a hierarchical clustering if there exists 2 samples, c1 and c2, which belong in the same cluster at some level k and remain clustered together at all higher levels > k. One example of hierarchical clustering is a correspondence tree, or a dendrogram [9]. 97 Copyright Vandana Publications. All Rights Reserved.

5 TABLE I. COMPARISION TABLE AMONG DIFFERENT SEGMENTATION TECHNIQUES ACCORDING TO DIFFERENT AUTHORS AUTHOR TITLE ALGORITHM USED PROPOSED TECHNIQUE ADVANTAGES PROBLEM Yongyue Zhang et al(2001) Mohamed N. Ahmed et al (2002) Mohamed F. Tolba et al (2003) Jamuna Kanta Sing et al (2005) Hai-yan yu et al (2008) Mukesh Kumar et al (2011) Sudipta Roy et al (2012) Hakeem Azaz Aslam et al (2013) Rajeshwari et al (2014) Segmentation of brain MRI images through a hidden markov random field model and the expectation-minimization [1] A modified fuzzy c-means for bias field estimation and segmentation of MRI data [2] MR-Brain Image Segmentation Using Gaussian Multi resolution Analysis and the EM [3] Segmentation of MR Images of the Human brain using Fuzzy Adaptive Radial Basis function Neural Network [4] Three-level Image Segmentation Based on Maximum Fuzzy Partition Entropy of 2-D Histogram and Quantum Genetic [5] A Texture based Tumor detection and automatic Segmentation using Seeded Region Growing Method [6] Detection and Quantification of Brain Tumor from MRI of Brain and it s Symmetric Analysis [7] A new approach to image segmentation for brain tumor detection using pillar K-means [8] Tumor detection and segmentation using Watershed and Hierarchical [9] Expectation Minimization Modified Fuzzy C-Means Gaussian Multiresolution Expectation Minimization Fuzzy Adaptive Radial Base Function Quantum Genetic Seeded Region Growing Symmetry Analysis K-means Watershed Segmentation Bias Field Estimation Gaussian Multi Resolution Analysis Neural Network Fuzzy partition entropy of 2D histogram and genetic A Texture based Tumor detection and automatic Segmentation using Seeded Region Growing Method Modular approach to solve MRI segmentation Pillar K-means Combine approach of Watershed and Hierarchical clustering Technique which shows ability to encode both spatial and statistical properties of an image Faster to generate results Less Sensitive to noise It preserves sharpness of an image Quantum genetic is selected for optimal combinations of parameters By this it is possible to determine abnormality is present in the image or not. The proposed approach can be able to find the status of increase in the disease using quantitative analysis Improve the accuracy and enhance the quality of image segmentation in all colour spaces Provides optimal solution The method requires quantify threshold and does not produce accurate results most of the time. Technique is bound to single point of input Rarely Preserved Noise It has Ability to do only one task related to fusion This can not implemented Time consuming It is very time taking A series of experiments with four different colour spaces with restricted variance and execution conducted It works in the bottom up manner 98 Copyright Vandana Publications. All Rights Reserved.

6 Eman-Abdel Maksud et al (2015) Brain Tumor segmentation Based on a hybrid clustering techniques [10] Fuzzy C-Means K-means Clustering technique integrated with fuzzy C-means Minimize the execution time Intensity adjustment IV. CONCLUSIONS In this paper we have studied the different type of brain tumor segmentation techniques and how it works. We also discussed about the comparative analysis of different techniques which are proposed by different authors. These different techniques shows different aspects of medical imaging especially for the application of detection of brain tumor using MRI. We have presented the review of different brain image segmentation methods presented so far, as well their pros and cons are discussed as comparative analysis. For the future work we suggest to present more accurate, efficient as well as faster method for early detection and classification of brain tumors and work upon 3-D MR images of brain. REFERENCES [1] Y. Zhang, M. Brady and S. Smith, Segmentation of Brain MR Images through a Hidden Markov Random Field Model and the Expectation-Maximization, Proceedings of the IEEE transaction on Medical Images, January2001 [2] M.N. Ahmed, S.M. Yamany, N. Mohamed and T. Moriarty, A modified fuzzy c-means for bias field estimation and segmentation of MRI data, Proceedings of the IEEE transaction on Medical Images, KY, USA, March [3] M.F.Tolba, M.G. Mostafa, T.F. Gharib and M.A Salem, MR-Brain Image Segmentation Using Gaussian Multi resolution Analysis and the EM, ICEIS, [4] J.K.Sing, D.K. Basu, M. Nasipuri and M. Kundu, Segmentation of MR Images of the Human brain using Fuzzy Adaptive Radial Basis function Neural Network.Pattern Recognition and Machine Intelligence, LNCS, Berlin, Heidelberg, [5] H. Yu and J.L. Fan, Three-level Image Segmentation Based on Maximum Fuzzy Partition Entropy of 2-D Histogram and Quantum Genetic,Advanced Intelligent Computing Theories and Applications.With Aspects of Artificial Intelligence. Lecture Notes in Computer Science, Berlin, Heidelberg [6] M.Kumar and K.K. Mehta, A Texture based Tumor detection and automatic Segmentation using Seeded Region Growing Method, International Journal of Computer Technology and Applications, August [7] S. Roy and S.K. Bandyopadhyay, Detection and Quantification of Brain Tumor from MRI of Brain and its Symmetric Analysis, International Journal of Information and Communication Technology Research, KY, USA, June [8] Hakeem aezaz aslam et al, A new approach to image segmentation for brain tumor detection using pillar k-means algoritham, international journal of advance research in computer and communiacation engineering, march [9] R. Rajeshwari et al, Tumor detection and segmentation using Water- shed and hierarchical clustering algoritham, International journal of innovative research of computer and communication engineering, October [10] Eman abdal maksood et al, brain tumor segmentation based on a hybrid clustering technique, Egyptian infomatics journal, cairo university [11] Kailash sinha et al, Efficient segmentation methods for tumor detection in MRI images, conference on electrical, [12] Gauri P. Anandgaonkar et al, Detection and identification of brain tumor in brain MR images using Fuzzy C-means segmentation, international journal of advanced research in computer and communication engineering, October [13] K.selvanayki et al, Intelligent brain tumor tissue segmentation from MRI using metaheuristic algorithams, journal of global research in computer science, [14] Mr. Shital S. Agarwal, Detection of brain tumor using different edge detection, International journal of emerging research in management and technology, april [15] D. Manju, Comparision study of segmentation techniques for brain tumor detection, International journal of computer science and mobile computing, November [16] Ruchi D. Deshmukh, Study of different brain tumor MRI image segmentation techniques, International journal of computer science and technology, april [17] Roopali l. ladha et al, A review on brain tumor detection using segmentation and threshold operation, international journal of computer science and technology, Copyright Vandana Publications. All Rights Reserved.