Diagnosis and segmentation of brain tumor from MR image

Abstract

Abnormal growth of tissues inside the brain leads to the formation of brain tumor. In order to decrease the mortality rate due to brain tumor, efficient techniques for the earlier detection of the tumors are required. The present-day technological advancements include the magnetic resonance imaging (MRI) scans, computed tomography (CT) scans, and advanced X-rays which help us in tumor detection. The MRI scans are widely used nowadays because of its noninvasive radiation and the accuracy of the images. This paper proposes a strategy for efficient detection of a brain tumor in MRI brain images. The system proposed in this paper is a handy tool for accurate prediction and segmentation of brain tumor. The general properties of the images called the gray-level co-occurrence matrix (GLCM) features and the spatial feature (mean) are combined with the transform domain-based discrete cosine transform (DCT) features that are extracted from the MR images are used as feature sets. Support vector machine (SVM) is used as the classifier which classifies the images as tumorous or non-tumorous. Once the classification is done using SVM classifier, the tumorous images are alone subjected to watershed transform-based segmentation to exactly extract the tumor region alone. The entire experiment is conducted on the images of various patients acquired from MEDALL DIAGNOSTICS, Tiruchirappalli, India. The method gives accuracy of over 99%, hence improving the chances for the patients’ survival by proper detection of tumors at an early stage.