Asian Journal of Engineering and Technology Innovation

An Ensemble approach for Classification of Glioma using MR

Ashwini V D, Divya R, Chaitra A, Dheeraj D

Dept Of CSE, REVAITM, Bangalore, Karnataka

Abstract

Today, detection of tumors in human brain in the field of image processing has become a challenging task. Abnormal cells in the brain are considered as tumor. Brain tumors can be cancerous or noncancerous. Glioma is the most occurring type of brain tumor in humans which makes up about 30% of central nervous system and brain and 80% of all malignant brain tumors. The most widely used modality for brain tumor analysis is Magnetic resonance imaging (MRI). In current practice, using an MRI and sometimes with the help of a biopsy, a doctor would diagnose a tumor to be benign or malignant. In our work around eighty MRI images including T1 and T2 weighted samples, FLAIR and ROI samples are collected from radiologists and global dataset BRAININX for experimentation. In this paper we propose an ensemble approach like Global thresholding, morphological feature extraction techniques combined with non parametric local transforms for detection of tumor.

Keywords:

Glioma, non parametric local transforms, segmentation, global thresholding