Publication

Abstract : A brain tumour is a deadly disease, and it is an unwanted cells development in the human brain. In medical technology, brain tumour detection and diagnosis increase the patient's life days. In this manuscript, an effective block-based brain tumour detection method is proposed for MR images. The proposed method has four steps: convert T2W images into 8 × 8 blocks, Feature extraction, Feature selection and Classification. This method uses the Brain Tumour Segmentation (BRATS)-2013 dataset with High-grade glioma (HGG) and Low-grade glioma (LGG) MRI multi-modal on T2-Weighted images. Feature extraction is achieved by GLCM texture features. The Chi-square test method is implemented to rank features in the feature selection process. Finally, the classification process is achieved by SVM, and it has two main phases training and testing. This method uses 90,000 blocks for training and 36,000 blocks for testing in the combination of HGG and LGG images. The blocks are classified into normal or tumour based on their features during the testing phase. The proposed block-based brain tumour detection method achieves 100% sensitivity, 100% specificity and 100% accuracy in the testing phase.