Segmentation, Feature Extraction and Classification of Astrocytoma in MR Images

The purpose of this research paper is to segment tumorous tissue from brain magnetic resonance images using image processing. For this the astrocytoma, a type of brain tumor is considered in which cells grow abnormally in the brain. Astrocytoma is classified as low grade or high grade using k-nn classifier and then analyzed its performance on the basis of three parameters i.e., accuracy, severity and specificity. The Magnetic Resonance Images (MRI) of astrocytoma has been taken from BRATS database so as to explore different algorithms for segmentation of brain tumour. The enhancement of MRI images is done using Contrast Limited Adaptive Histogram Equalization (CLAHE). After that basic global thresholding method is used for the automatic segmentation of tumorous tissue. In feature extraction step, the shape based features and texture based features are extracted. On the basis of features extracted from the segmented image, K-nn classifier is used to classify the images in two grades i.e., low grade or high grade. The performance of the system is evaluated by parameters accuracy, severity and specificity. The accuracy is coming out to be 93%.