Brain Tumor Classification Using Enhanced Statistical Texture Features
Mallikarjun Mudda, R. Manjunath, Narayanan Krishnamurthy
Abstract
Radiology is a vast subject and we require more knowledge and understanding for exact detection of tumor in medical science. Thus a need for tumor detection system overcomes the shortage of skilled radiologists. Biomedical image processing using Magnetic Resonance Imaging (MRI) makes the task of detection and localization of brain tumor. In this article, a brain tumor segmentation and detection approach has been designed using MRI sequence images as input image for defining the tumor region. This process is difficult due to the large diversity in the presences of tumor tissues with respect to different patients and in most of the cases similarity within the normal tissues makes the task difficult. The main goal is to classify the brain into the presence of brain tumor or a healthy brain. The proposed system provides Edge-based Contourlet Transformation for multiple input image registration, fusion and pre-processing, for Region of Interest(ROI) of tumor region the region-growing segmentation algorithm provides accurate boundaries, in feature extraction the Gray Level Run Length Matrix (GLRLM) and Center-Symmetric Local Binary Patterns(CSLBP) texture features are combined for efficient brain tumor detection and for classification Adopting Neural Network(ANN) techniques is carried out. The experimental results of our proposed method are compared with different algorithms in terms of accuracy.