Deep Learning Techniques for the Classification of Brain Tumor: A Comprehensive Survey

Abstract

Researchers have given immense consideration to unsupervised approaches because of their tendency for automatic feature generation and excellent performance with a reduced error margin. Deep learning (DL) models are emerging as vital methods for image analysis in medical fields, such as classification, segmentation, and reconstruction. Deep learning relies on learning hierarchical features and data representation, making it superior to its antecedent. Deep learning models efficiently discover descriptive information about the optimal representation of various brain tumors when applied for brain tumor classification from MRI. Despite various efforts, there remains a gap in the current literature for inclusive representation of recently developed deep learning-based classification methods. The current study attempts to fill this gap by briefly reviewing the current state of the art on brain tumor segmentation and classification methods while focusing on deep learning methods. the proposed survey dedicates itself to reviewed the current state of the art on automated classification techniques for brain tumor MRI to produce an inclusive picture of the most recent and worthy of adoption models proposed in this area. Despite various attempts to conduct surveys on brain tumor segmentation and classification techniques, no such study could be found in the current literature that has dedicated its focus to the most effective approach towards classification. This research begins by identifying major classes of brain tumor segmentation and classification while presenting its focused area and reviewing the most recent state-of-the-art classification approach, the deep learning-based classification method. The powerful learning ability of deep learning mechanisms has been reviewed for their performance, and a comparison between them is presented to encourage its applications. Future recommendations and directions are also drawn up to establish a pursuable course for welcoming widespread adoption of potential applications in the area.

Topics & Concepts