Identification of Astrocytoma Grade Using Intensity, Texture, and Shape Based Features

Abstract

Astrocytoma is a common type of brain tumor that develops in the glial cells in cerebrum or astrocytes. In a malignant form, it is associated with high mortality. Identifying its grade helps the physicians to think about effective treatment. However, the irregular structure of this tumor type creates difficulty in the identification of its grade. Due to this, medical practitioners suggest additional examinations such as Magnetic Resonance Spectroscopy (MRS) and biopsy for accurate grade identification. In this work, we propose a method to identify astrocytoma grade from brain Magnetic Resonance Imaging (MRI). The proposed method can classify the tumor into low grade and high grade. The segmentation of the brain MRI is performed using spatial fuzzy clustering. We have used intensity, texture, and shape based features for classification. Five classifiers are used for the classification purpose. Our experiment results show that we can achieve an accuracy rate of 92.3% by integrating all three types of features together and applying a suitable classifier.