A Study of Brain Tumor Segmentation and Classification using Machine and Deep Learning Techniques

Abstract

One of the most common methodologies in medical study is to detect a brain tumor and its development from a magnetic resonance imaging (MRI) of the brain. As a result, manually segmenting brain tumor from MRI images is a time-consuming and difficult process. Furthermore, an automatic brain tumor categorization based on an MRI scan is non-invasive, eliminating the need for a sample and making the diagnosing procedure safer. The scientific community has been working tirelessly since the turn of the millennium and the late 1990s to develop an automated brain tumor segmentation and classification approach. As a result, there is a lot of literature on segmentation employing region growth, classical machine learning, and deep learning methods. Similarly, several tasks in the domain of brain tumor categorization into their various histological types have been completed, with outstanding performance outcomes. The goal of this study is to present a complete assessment of three newly suggested important segmentation and classification methods for the brain tumor, namely, region growth, shallow machine learning, and deep learning, taking into explanation state-of-the-art techniques and their performance. Technical issues such as the strengths and drawbacks of alternative methods, pre-and post-processing methodologies, feature extraction, datasets, and model performance of the evaluation metrics are also covered in the conventional works involved in this study.