Hybrid Approach to Differentiate Healthy and Pathological Tissues from MR Brain Images

Abstract

Medical image processing is one of the areas in which researchers have an affinity to working. By the advancements in the computer technology, improved techniques of data acquisition newline, analysis, processing, and visualization have a great impact on medical image processing. Magnetic resonance images (MRI) provides information about potential abnormal tissues necessary for medical newline follow up. Brain MRI gets additional importance in medical science as it is the only preliminary method of diagnosing a brain tumor. MRI helps the radiologist to acquire and visualize images of the brain tumor for anatomical judgment in a non-invasive way. A Brain tumor is one of the destructive and devastating types of disease. It can be cure by its detection in early stage followed by the treatment. Therefore, it is necessary to propose the method which can efficiently identify whether the patient is suffering from a brain tumor or not. This thesis addresses the newline problem of detection and classification of brain tumors. The important points in this research have been to develop and implement a robust algorithm for newline the classification of the Brain MRI images as normal or abnormal in this thesis. a hybrid approach has been proposed for detecting and classifying brain tumor using MR Images. The proposed approach comprises of four phases. In the phase-1, pre processing is done after that in phase-2 thresholding technique is applied. In next phase, feature selection is done using Horlick features and finally in the fourth phase, classification is done using different classifier. The dataset comprises of three MRI scan images that are T1-weighted scan, Proton Density scan and T2-weighted scan. The experimental results of proposed method have been evaluated and validated for performance and quality analysis on brain MRI images based on specificity, accuracy and specificity.