A comprehensive framework for classification of brain tumour images using SVM and curvelet transform

TLDR: This work introduces an efficient approach for brain tumour detection using curvelet transform–based statistical features combined with GLCM (Grey Level Cooccurrence Matrix) texture features together with support vector machine.

Abstract: This work introduces an efficient approach for brain tumour detection using curvelet transform–based statistical features combined with GLCM (Grey Level Cooccurrence Matrix) texture features. The detection of the brain tumour is considered as a challenging problem, due to the irregularity of the highly varying structure of the tumour cells. The major contribution of the proposed work resides in the selection of significant features from both spatial and frequency domains for training the system. It combines the curvelet transform–based statistical features in the frequency domain with the GLCM texture features in the spatial domain. The proposed method applies skull–stripping as the pre–processing step to extract the brain portion from the MRI slice. This pre–processed image is subjected to watershed transform–based segmentation process to extract the necessary region of interest. From the extracted region of interest, frequency and spatial domain–based features are extracted. Finally, the classification model is developed using support vector machine. Experiments reveal that the proposed classifier is good in terms of accuracy.