Employing Deep Learning Approaches for the Partitioning of Brain Tumefactions in MRI Images: A Brief Abridgement

Abstract

The segmentation of images is vital as well as a burdensome constituent within the discipline of processing medical images. In the interest of performing anatomical studies on the human body or even put forth a medical treatment plan for a given situation of utmost medical importance, the results meeting satisfactory conditions relies widely on the technologies used on the medical images and the quality of the medical images referred to herein. Magnetic Resonance Imaging is the most preferred approach while referring to medical images having to do with the human brain, mainly to detect the presence of brain tumors. These images are often known to possess with them a lot of noise and ambiguity making the segmentation process difficult if performed manually. This calls for the development of improved approaches capable of dealing with the above-mentioned challenges. As a result, the analysis, as well as the overall treatment of brain tumors, can achieve the best possible results with minimal time consumption and human errors. This paper presents a brief review of a few lately proposed approaches to perform segmentation of MRI images using deep-learning techniques. It is observed that many the new techniques are capable of performing well with reduced error rates when it comes to segmenting and detecting brain tumors with varying modalities. However, it is also important to note that there is no one-for-all approach to date that is known to outperform every other existing method used today.