IJARCCE ISSN (Online) 2278-1021 ISSN (Print) 2319 5940 International Journal of Advanced Research in Computer and Communication Engineering ISO 3297:2007 Certified Vol. 5, Issue 8, August 2016 Copyright to IJARCCE DOI 10.17148/IJARCCE.2016.58116 553 Brain Tumor Classification Based on Singular Value Decomposition Nidahl K. El Abbadi 1 , Neamah E. Kadhim 2 Computer Science Department, Education College, University of Kufa, Najaf, Iraq 1 Computer Science Department, Science College, University of Babylon, Babylon, Iraq 2 Abstract: Every day over 100 adults will be diagnosed with a primary brain tumor and many more will be diagnosed with a cancer. Diagnosing a specific type of brain tumor can be a complicated affair, making confirmation of its diagnosis essential. In this paper we suggested new method for detection of brain tumor based on singular value decomposition (SVD). The algorithm first trained/learned with normal brain MR images, then in the second step the algorithm become capable to classify the brain MR images into healthy and non-healthy image (that have a tumor). The algorithm is trained with 20 of normal brain MR images and tested with 50 brain MR images. The accuracy of this method was up to 97%. Keywords: MRI, brain tumor, classification, image processing, SVD. I. INTRODUCTION The medical image processing is a method of using computer algorithm in processing the medical image. The brain magnetic resonance image (MRI) classification is one of the most important field in medical image processing. The manual classification of brain MR image is a time consuming and not always gives an accurate results. Therefore there is an important for automatic classification using various technique. In these techniques the classifier first trained or learned with known data (images) that belonging to number of classes, then the classifier became capable of classifying unseen imagesaccurately into a specificclass [1]. Currently, brain tumor has become one of the main cause for increasing mortality among children and adults. It's found that the total number of people that hardship and dying from brain tumors has been increased to 300 for each year during past few decades. The primary diagnosis is frequently stated as a brain lesion. A lesion is a region in an organ or tissue that has suffered damage through disease or shock; mainly there is a different think about the brain and a part of it does not appear to like normal tissue [2]. Extra testing is commonly ordered to get a better notion of the location, size and effect of the tumor, as well as detecting any cancers in other portions of the body. It takes knowledge to be able to see specific subtle differences in MRIs. An additional opinion from knowledgeable doctor or team that regularly deals with brain tumors may alteration the preliminary diagnosis, in either tumor class or grade, and therefore alteration in the treatments. MRI only can be unsatisfied (it may not be a tumor at all), making an exhaustive examination of all of our indications (symptoms), and as soon as possible, a biopsy, vital to our diagnosis [3]. II. RELATED WORKS Said Charfi [4] suggested a hybrid intelligent machine learning procedure for computer-aided detection for automatic recognition of brain tumor through MR images. The proposed method is based on the following computational procedures; the histogram dependent thresholding for image segmentation, the discrete wavelet transform for features extraction, then using PCA for reducing the dimensionality of the extracted features, and the feed forward back-propagation neural network to classify inputs into normal or abnormal. The accuracy of the classification on both training and test images is 90% which was considerably good. J. Umamaheswari and Dr. G. Radhamani [5] proposed a hybrid approach for medical image classification. This approach consists of feature extraction and classification. The classification consists of Multi Linear Discriminate Analysis (MLDA) and Support Vector Machine (SVM). Classification is done on the base of parameter extracted by Gray Level Co-occurrence Matrix (GLCM) and histogram texture feature extraction method.The efficiency of the method has been assessed using the following measures:Accuracy, Sensitivity and Specificity, the proposed method get 94%, 87% and 92% respectively. Rosy Kumari [6] suggested method based on Grey-Level Co-occurrence matrix (GLCM) and Support Vector Machine (SVM). The suggested method involves of two steps: feature extraction and classification. In first step,the features are extracted from images using GLCM. In the next step, the features that are extracted fed as input to SVM classifier. Theimages are classified into normal and abnormal along with variety of disease depending upon features.