Comparative Analysis of Glioblastoma Brain Tumour Classification

Abstract

A brain tumour is a disease in which cells develop and grow abnormally. The glioblastoma brain tumour is an aggressive type of cancer with a high diagnosed rate and low survival rates. Despite the development of high-precision imaging equipment for tumour detection, the classification of brain tumour remains a challenge. In addressing this issue, numerous research which combined image analysis with Artificial Intelligence (AI) have been proposed to aid in decision-making. The present study intends to compare the performance of glioblastoma brain tumour classification utilizing common feature extraction approaches and a combination of two separate characteristic classifiers. The process began with a skull stripping process using High-Definition Brain Extraction Tools (HD-BETS), followed by image enhancement using Contrast Limited Adaptive Histogram Equalization (CLAHE), and texture feature extraction using Gray Level Co-Occurrence Matrix (GLCM), Local Binary Pattern (LBP), and Intensity-based Analysis (IBA) methods. The glioblastoma brain tumour classification was then performed using two classifiers which are K-Nearest Neighbor (KNN) and Support Vector Machine (SVM). The classification performances of KNN and SVM with the combination and non-combination texture features of GLCM, LBP, and IBA were then measured. The results indicated that the combination of GLCM, LBP, and IBA features was capable to achieve 100% of glioblastoma brain tumour classification accuracy. It was also observed that KNN performs slightly better than the SVM in all features' combinations. Yet, in the future, the classification performances could also be compared with more advanced classifiers or machine learning techniques such as Deep Learning and Convolutional Neural Networks.