A Machine Learning Approach in Medical Image Analysis for Brain Tumor Detection

Abstract

Brain tumor is a serious medical condition if not detected early will reduce the life span of humans. Magnetic resonance imaging (MRI) is a common method nowadays for abnormality detection and classification. But the manual detection is less accurate also a large amount data to be processed. Thus, manual detection leads to error in tumor segmentation. The data obtained from MRI images also inherent to noise produced by the MRI machine parts. The detection and removal of this noise play a vital part in the detection accuracy of tumor. So, here, we propose a Total Variation (TV) homomorphic filter to reduce the noise and enhance the edges of MRI data. SVM classifier is employed for learning and classification. The method produces better results than conventional methods like median filtering, anisotropic filtering, etc.