Automated MRI Brain Tumour Segmentation and Classification Based on Deep Learning Techniques

Abstract

A brain tumour is a significant death problem among other cancer types because the brain is a susceptible, complicated, and significant part of the human. The precise and appropriate examination can control the lifespan of an individual to a remarkable period. The image-segmentation of MRI (magnetic resonance images) is significant for envisioning and analyzing irregular tissues, notably during a medical examination. Intricacy and modifications of the tumour formation intensify difficulties in computerized brain tumour detection and segmentation in MRIs. This proposed system performs an automated brain tumour segmentation process in the MRI brain image accompanied by classification. Since, in this method, an effective brain tumour detection and classification scheme is intended using fusing GLCM features and CNN. The proposed method consists of four steps: pre-processing, image segmentation, extraction of features, and optimization and classification. First, noise elimination is done as the pre-processing step at the brain MR images. Following the classification method, irregular brain MR images are provided to the segmentation part to detect tumours and segments using the fuzzy c means (FCM) technique. Following that, GLCM and Ant colony optimization (ACO) which features are obtained from these noiseless MRI images of the brain. A tremendous numeral of features is decreased founded on Ant colony optimization (ACO). Finally, chosen features of brain images are provided to the CNN classifier to categorise MRI brain images as abnormal or normal. The proposed method performance is examined in various metrics, and testing outcomes are comparable to present systems.