A new hybrid approach for brain tumor classification using BWT-KSVM

Brain tumor is one kind of malignancy and a leading cause of death worldwide. Brain tumor detection is dependent on the radiologist's interception and experience. Complex characteristics of brain tumor and noises in the MR images, make it a difficult task for the radiologists. An automated system can provide assistance to the radiologists by reducing workload and by improving diagnostic accuracy. In this study, a tumor detection and classification method are proposed using K-Means, Gray Level Co-occurrence Matrix (GLCM), Berkeley Wavelet Transform (BWT), Principal Component Analysis (PCA) and Kernel Support Vector Machine (KSVM). This proposed method utilizes the advantages of both GLCM and BWT in case of feature extraction. To segment and detect the tumor region from MRI images k-means clustering is used. Results of the extracted features are used for classifying normal and abnormal (low-grade, high-grade glioma) using SVM classifier. From the experimental result, it is illustrated that the proposed method earned 95.2% accuracy and can be an effective method for real time application.