Brain Tumor Analysis using SVM and Score function

Brain Tumor Analysis Using SVM and Score

Function

Ms. Jisney Thomas Ms. Midhu Yesodh

PG Student (M. Tech) PG Student (M. Tech)

Department of Applied Electronics & Communication Systems Department of Applied Electronics & Communication Systems Thejus Engg. College Thejus Engg. College

Ms. Princy P

PG Student (M. Tech)

Department of Applied Electronics & Communication Systems Thejus Engg College

Abstract

Medical field is very much depended on image processing nowadays. Brain tumor is very dangerous and harmful type of cancer. But diagnosis and treatment of brain tumour cost is very high and it lasts for longer period. The number of neuro patients is increasing, which in turn increases burden on small group of radiologists. So we need more efficient Tumour diagnosis system that help the Radiologists. In this project, a new CBIR method is introduced to detect tumour in tumorous image. Every CBIR system has feature extraction and classification. Here feature is extracted using Discrete Wavelet Transform and images are classified using Support Vector Machine .So when we give a Brain MRI image, the image is classified as normal or tumorous image. If the image is tumorous, then change detection method that searches for the most dissimilar region (axis-parallel bounding boxes) between the left and the right halves of a brain in an axial view MR slice. This change detection process uses a novel score function based on Bhattacharya coefficient computed with gray level intensity histograms.

Keywords: Content Based Image Retrieval (CBIR); Discrete Wavelet Transform (DWT); Euclidean Distance; Support Vector Machine (SVM)

________________________________________________________________________________________________________

I.

I

UNWANTED growth of cells inside the skull is termed as brain tumor. They classified into two - benign (non-cancerous) and malignant (cancerous) tumors. The first one is slow growing tumors which causes potentially damaging pressure but does not spread into surrounding brain tissue. But the second one is rapid growing tumor and able to spread into surrounding brain.Magnetic resonance imaging (MRI) is the medical imaging method used for diagnosis of brain tumor. The rich information that MR images provide about the soft tissue anatomy has dramatically improved the quality of brain pathology diagnosis and treatment. It produces high quality images of the anatomical structures of the human body, especially in the brain, and provides rich information for clinical diagnosis and biomedical research. However, the amount of data is far too much for manual interpretation and hence there is a great need for automated image analysis tools.

Text-based image retrieval, the first method available, is the typical and traditional method for retrieving images. In this method, images are annotated by keywords and retrieving is performed through keywords as indices of images. This method however, has many significant disadvantages including manual image annotation which is a labor intensive and time consuming process. Again, it causes errors because each word may have several meanings depending on the context. Therefore various methods and algorithms have been presented for automatic image annotation. However, since those methods describe images just with keywords, they also have some problems noted earlier. Content-based image retrieval (CBIR) [15] systems for medical images are important to deliver a stable platform to catalog, search, and retrieve images based on their content. So CBIR is the process of finding relevant image from large collection of image database using visual queries. In this paper, a new CBIR method is introduced that gives the similar images as well as detect tumor in tumorous image. The rest of the paper is organized as follows. Section II consist of the related works regarding the CBIR. The proposed approach and methods for filtering, feature extraction, classification and tumor detection is discussed in Section III. Experimental results are shown in Section IV, and Section V concludes the paper.

II.

R

classification was proposed by R. Guruvasuki and A. Josephine[3]. Their result gave good result in multiqueries but for single query result is not satisfactory. Hatice Akakin and Metin N. Gurcan [2] have proposed CBIR system which uses a multi-tiered approach for retrieving images. In first tier they classified images using SVM in two main types and K- nearest neighbors are used in second tier for classified the subtypes of the two main types. Here the robustness of the method is increased due to these two classifiers. The method is used only for multi-image query. Here they have not discussed the performance evolution of the retrieval process. Hashem K. et. al. [8] have proposed the MRI image classification method and compare the classification results using two different classifiers, KNN and SVM. They classified the images in eight different classes and concluded that SVM gives better classification result than KNN. Mohanpriya S. and Vadivel M. [1] have proposed a new CBIR system for classified and retrieves the images from the database. They also used two tiered approach for classified the MRI images. Here also the robustness is increased due to the two classifiers. The method is also for multi-image query and cannot be used for single query. Yudong Z. et. al. [6] have proposed a hybrid method for image MRI image classification. They extracted texture features using wavelet transform and classified MRI images using Back Propagation Neural Network. The result of their method is good. Chaplot S. et. al. [7] have proposed the MRI image classification method for classified the images in normal and abnormal class and compare the classification result using different classifiers. They also concluded that SVM classifier gives better result than ANN. Baidya Nath Saha et.al.[18] proposed a novel automated , fast, and approximate segmentation technique. The input is a patient study consisting of a set of MR slices, and its output is a subset of the slices that include axis-parallel boxes that circumscribe the tumors.

III.

P

The block diagram of proposed method is shown in Fig.1. Before classification there is training phase and testing phase. At training phase,both tumorous and normal images are given to the preprocessing step. Preprocessing step is noise removing step. Most of the MRI images are passed through many sources. Therefore noise should be removed before extracting the features. After preprocessed the image is projected onto the feature space by extracting the texture features using wavelet transform.

Fig. 1: Block diagram of proposed method

When input data is too large to process, it is transformed in reduce set of features (feature vector). Thus process of transforming an input data in set of feature is called feature extraction. Here feature is extracted using wavelet transform. The main advantage of using wavelet transform is that it provides localized frequency information of an image which is useful for classification. Wavelet transform decomposed the signal using mother wavelet signal. In this method two levels 2 D Discrete Wavelet Transform (DWT) is used for feature extraction (wavelet coefficients). Haar basis filters are used for decomposition.After feature extraction, using SVM classifier a model data library is created.. Support Vector Machine (SVM) is a binary classifier based on supervised learning. It classifies the images by creating an optimal hyper plane between the data points of two different classes. At testing phase, categorization of query image is done by means of this data library.In this phase test image pass through preprocessing step for removing noise.Then features of test image is extracted using DWT.By using model data library created by SVM classifier, the image is classified into tumorous or normal. After that if any query image is detected as tumorous, the tumor is detected using change detection method that searches for the most dissimilar region (axis-parallel bounding boxes) between the left and the right halves of a brain in an axial view MR slice.

Filtering: A.

Feature Extraction: B.

The feature extraction was done using DWT(Discrete wavelet transform).

Discrete Wavelet Transform:

1)

The discrete wavelet transform (DWT)[7] is a powerful implementation of the wavelet transform using the dyadic scales and positions. The family of wavelet functions is represented in eq.(1)

( ) ⁄ ( ) (1)

where m, n ϵ Z, the set of all integers. The wavelet transform decomposes a signal x(t) into a family of synthesis wavelets as given below in Eqs. (2) and (3),

( ) ∑ ∑ ( ) (2) Here ⟨ ( ) ( )⟩ (3)

2D DWT:

2)

In case of images, the DWT is applied to each dimension separately. As a result, there are four sub-band (LL, LH, HH, and HL) images at each scale. The sub-band LL is used for next 2D DWT. The LL sub band can be regarded as the approximation component of the image, while the LH, HL, HH sub bands can be regarded as the detailed components of the image. As the level of decomposition increased, compacter but coarser approximation component was obtained. Thus, wavelets provide a simple hierarchical framework for interpreting the image information.

Classification using Support Vector Machine: C.

SVMs (Support Vector Machines) are a useful technique for data classification. A classification task usually involves separating data into training and testing sets. Each instance in the training set contains one “target value” (i.e. the class labels) and several “attributes” (i.e. the features or observed variables). The goal of SVM is to produce a model (based on the training data) which predicts the target values of the test data given only the test data attributes. SVM is a binary classification method that takes as input labeled data from two classes and outputs a model file for classifying new unlabeled/labeled data into one of two classes. The SVM originated from the idea of the structural risk minimization that was developed by Vapnik[19]. Support vector machines are primarily two class classifiers that have been shown to be attractive and more systematic to learning linear or non-linear class boundaries. The use of SVM, like any other machine learning technique, involves two basic steps namely training and testing. Training an SVM involves feeding known data to the SVM along with previously known decision values, thus forming a finite training set. It is from the training set that an SVM gets its intelligence to classify unknown data.

Tumor Detection: D.

Tumor detection is done using fast bounding box (FBB) algorithm. In FBB, after finding the axis of symmetry on an axial MR slice, the left (or the right) half serves as the test image I, and the right (or the left) half supplies as the reference image R. The region of change D here is restricted to be an axis-parallel rectangle, which essentially aims to circumscribe the abnormality. We utilize a novel score function that can identify the region of change D with two very quick searches– one along the vertical direction of the image and the other along the horizontal direction. The novel score function is based on Bhattacharya coefficient computed with gray level intensity histograms. After both vertical and horizontal scanning a bounding box is obtained over tumor effected portion .Finally area of tumorous region is find out by finding area of bounding box. The block diagram for Tumor detection is shown in figure 2.

IV.

E

The proposed technique was executed using matlab R2014a. The proposed CBIR method has been implemented on real Human Brain MRI dataset. All the input dataset consist two types of images: Normal and Tumorous in axial plane. The images were collected from Hospitals and from HarvardMedical,School,http://www.med.harvard.edu/AANLIB/home.html) [17] website. Feature extraction has been implemented using DWT and classification was done using support vector machine. The result from feature extraction will be used as input for classification. After classification If the image is tumorous, dissimilar region (axis-parallel bounding boxes) between the left and the right halves of a brain is found. This change detection process uses a nove l score function based on Bhattacharya coefficient computed with gray level intensity histograms. Also we found out the area of tumorous portion..The results of proposed method is given in figure 3 and 4.

Fig. 3:.Results upto feature extraction.(a)Input image,(b)Filtered image,(c)Approximate coefficient (d)Horizontal component (e)Vertical component and(f) Diagonal component

V.

C

With the help of wavelet and machine learning approach, we classified whether a brain image is normal or Tumorous. Here we extracted feature of MRI brain image using DWT (Discrete Wavelet Transform) and the images are classified using SVM (support vector machine). The main advantage of using wavelet transform is that it provides localized frequency information of an image which is useful for classification. After classification if the image is tumorous, dissimilar region (axis-parallel bounding boxes) between the left and the right halves of a brain is found. This change detection process uses a novel score function based on Bhattacharya coefficient computed with gray level intensity histograms. Also we found out the area of tumorous portion.

R

[1] Mohanpriya S., Vadivel M, “Automatic Retrieval of MRI Brain Image using Multiqueries System”, IEEE Conference, pp 1099-1103, 2013. [2] Hatice Cinar Akakin and Metin N. Gurcan, “Content-Based Microscopic Image Retrieval System for Multi-Image Queries” ,IEEE Transaction on

Information Technology in Biomedicine, Vol.16, No. 4, pp 758-768, 2012.

[3] R.Guruvasuki, A. Josephine Pushpa Arasi, “MRI Brain Image retrieval using Multi Support Vector Machine Classifier”, International Journal of Advanced Information Science and Technology, Vol. 10, No 10, pp 29-36, 2013.

[4] Lidiya Xavier, Thusnavis B. , Newton D.R. , “Content Based Image Retrieval Using Texture Features Based On Pyramid-Structure Wavelet Transform” , IEEE Conference, pp 79-83, 2011.

[5] B.Ramasubramanian, G. Praphakar, S. Murugeswari, “A Novel Approach for Content Based Microscopic Image Retrieval system Using Decision Tee Algorithm”, International journal of scientific& engineering research, Vol. 4, No 6, pp 584-588, 2013.

[6] Yudong Zhang, Zhengchao Dong, LenanWua, ShuihuaWanga, “A hybrid method for MRI brain image classification”, Elsevier journal Expert system and Application, Vol. 20, No 2, pp 10049-10053 ,2011.

[7] Sandeep Chaplot , L.M. Patnaik , N.R. Jagannathan, “Classification of magnetic resonance brain images using wavelets as input to support vector machine and neural network”, Elsevier journal on Biomedical Signal Processing and Control, Vo.1,No 1,pp 86 -92,2006.

[8] Hashem Kalbkhania, Mahrokh G. Shayesteha, BehroozZali-Vargahan , “Robust algorithm for brain magnetic resonance image(MRI)classification based on GARCH variances series”, Elsevier journal on Biomedical Signal Processing and Control, Vol. 8, No 6, pp 909-919, 2013.

[9] Z. Iscan, Z. DokurandT. Olmez, “Tumor detection by using Zernike moments on segmented magnetic resonance brain images”,Elsevier Journal of Expert system and Application, Vol. 37, No 3,pp 2540-2549, 2010.

[10] M. M. Rahman, P. Bhattacharya, and B. C. Desai, “A framework for medical image retrieval using machine learning and statistical similarity matching techniques with relevance feedback”, IEEE Transaction on Information Technology in Biomedicine, Vol. 11, No. 1, pp. 58–69, 2007. [11] Mehdi Lofti, Ali Solimani, Aras Dargazany, Hooman Afzal, Mojtaba Bandarabadi, “Combinig Wavelet Transform and Neural Networks for Image

Classification”, IEEE, 41st Southeastern Symposium on System Theory, pp 15-17, 2009.

[12] ShenFurao, Tomotaka Ogura, Osamu Hasegawa, “An Enhanced Self Organizing Incremental Neural network For Online Unsupervised learning”, Elsevier Journal on Neural Network, Vol. 20, No 8, pp 893-903, 2007.

[13] M.Kanimozhi, C.H. HimaBindu, “Brain MR Image Segmentation Using Self Organizing map”, International Journal of Advanced Research in Computer and Communication Engineering, Vol. 2,No 10,pp 3968-3973, 2013.

[14] El- Shayed A El Dahshan, Tamel Hosny, Abdel- badech M. Salem,“Hybrid intelligent techniques for MRI brain images classification”, Elsevier Journal of Digital Signal Processing, Vol. 20 , No 2 ,pp 433-441,2010.

[15] Amit Kumar Rohit, N. G. Chitaliya “A Novel Approach for Content based MRI Brain Image Retrieval” International Journal of Soft Computing and Engineering (IJSCE), ISSN: 2231-2307, Volume-4, Issue-3 July 2014.

[16] Aditi P. Killedar ,Veena P. Patil .Megha S. Borse “Content Based Image Retrieval Approach to Tumor Detection in Human Brain Using Magnetic Resonance Image” 1st International Conference on Recent Trends in Engineering & Technology, Special Issue of International Journal of electronics, Communication & Soft Computing Science & Engineering, ISSN: 2277-9477, Mar-2012.

[17] Harvard Medical School, Web: data available at:http://med.harvard.edu/AANLIB

[18] Baidya Nath Saha, Nilanjan Ray, Russell Greiner, Albert Murtha, Hong Zhang, “Quick Detection of Brain Tumors and Edemas: A Bounding Box Method Using Symmetry”, Computerized Medical Imaging and Graphics, ISSN 08956111. 2012, pp. 95-107