Magnetic resonance imaging-based brain tumor image classification performance enhancement
Belayneh Sisay Alemu, Sultan Feisso, E. Mohammed, Ayodeji Olalekan Salau
Abstract
Brain cancer is one of the most fatal types of disease, which is caused by an abnormally growing mass of defective brain tissue. Generally, brain cancer can be divided into benign and malignant, however, based on the World Health Organization, it can also be divided into grade I, II, III, and IV tumors. Magnetic Resonance Imaging (MRI) has become a crucial tool in the diagnosis and treatment of brain tumors. However, accurately classifying brain tumor images from MRI scans remains a challenging task due to the complexity and heterogeneity of tumor characteristics. This paper presents a Support Vector Machine (SVM) based classification method for brain tumor classification. The proposed method comprises steps such as noise reduction, segmentation, and feature extraction which were employed using the median filter or wavelet transform, Otsu's thresholding, and Gray-Level Co-occurrence Matrix respectively. Finally, classifications were performed using a Support Vector Machine which achieved a 99.9% accuracy using a dataset of 24 MRI images comprising 13 Malignant and 11 Benign brain tumors for training and 16 images for testing. When compared to previous approaches, the study's findings show a considerable improvement in the ability to classify images of brain tumors.