E. MURALI et al: A HYBRID APPROACH TO THE CLASSIFICATION OF BRAIN TUMOURS FROM MRI IMAGES . DOI 10.5013/IJSSST.a.19.04.03 3.1 ISSN: 1473-804x online, 1473-8031 print A Hybrid Approach to the Classification of Brain Tumours from MRI Images using Fast Bounding Box Algorithm E.Murali 1 , K. Meena 2 1 SIST Puttur, Andhra Pradesh, India 1, 2 Department of Computer Science & Engineering Veltech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology Chennai, India. 1 sai4murali@gmail.com, 2 meen.nandhu@gmail.com Abstract - Recently, image processing in medical applications has been one of the most inspiring emerging fields. In medical image processing MRI techniques are widely used to detect brain tumours. There is a different strategy to detect and extract the brain tumour signal. The process of detecting and extracting the brain tumour signals is based on the MRI scanned images for the cerebrum. This method includes segmentation, morphological operations and some noise removal functions. This hybrid approach includes Bhattacharyya Coefficient for extraction of features and bounding box methodology for brain tumour classification. The work is implemented in MATLAB 2016a. Keywords - Medical Images, Magnetic Resonance Images, Hybrid Approach, Machine Learning, Segmentation, Morphological Operator. I. INTRODUCTION At present, brain tumour has become one of the prominent causes of demise among children as well as in adults [1]. To identification of tumour are using segmentation, patient observing, remedy scheduling, neurosurgery and radiotherapy making plans. The aim of segmentation is to discover the tumour and consult with distinct sub-regions of the tumour, specifically edema, non-enhanced and enhanced regions (Fig. 1) [2]. Different modalities can used to diagnose a brain tumour using MRI images. Fig. 1 Ground truth Tumour Segmentation Properly, tumour (especially glioblastomas, metastases) can show up successfully several positions in the brain. They don't have earlier shape, and every now and again have un-all around categorized edges. Additionally, they at first sighting initiate themselves in dark scales that are existing in solid tissues also. As prominence, still we are doing brain tumour segmentation [10] manually. This practice is consuming lot of time and tedious [3]. The growing of abnormal group of cell inside the skull is called a tumour. It may occur in any person at any stage and display at any district and has huge assortment of sizes and styles. It may be divided through radiotherapy or by means of chemotherapy [4]. II. OUTLINE OF PROPOSED APPROACH The first stage in specifying a course of treatment is identifying the presence of brain tumour. By and large the detail of a mind tumour includes neurological tests, cerebrum sweeps and investigation of the mind tissue. The tumour classification is from least to most aggressive by using diagnostic statistics. Tumour can occur in various parts of the brain i.e. benign and malignant, and it may or may not be primary. The objective of the proposed system is: - Design and implement the GUI for segmentation of Tumour region - The input image is divided into two slices, once containing tumour region (original image) and other containing without tumour (reference image) - Identify the tumour region from an MR image. - Segmentation of tumour region using FBB and Mean shift clustering.