A survey of deep learning for MRI brain tumor segmentation methods: Trends, challenges, and future directions

PurposeStructural Magnetic Resonance Imaging (MRI) of the brain is an effective way to study its internal structure. Identifying and classifying brain malignancies is a difficult and onerous task commonly handled by radiologists. Digital image processing processes, such as preprocessing, segmentation, and classification, can help clinical specialists diagnose certain types of brain cancers in addition to detecting the precise location of tumors and studying minute alterations. The state of brain tumor identification using MRI scans is discussed in this survey using numerous state-of-the-art machine learning and deep learning approaches.MethodsThis review highlights brain tumor image segmentation techniques, publicly available datasets, deep learning techniques, and deep learning architectures used by various researchers in the process of brain tumor detection. Additionally, the study presents a comprehensive review of the performance of existing deep learning algorithms, challenges, and future research directions.ResultsVarious methods proposed so far have been compared based on their accuracy. Many studies have attained an accuracy of more than 98% using SVM in the segmentation and analysis of brain tumor detection using MR images. The ANN outperformed all the other algorithms with 99% accuracy in brain tumor detection using deep learning.ConclusionThis review's objective is to increase scholars' interest in this difficult field and familiarize them with current advancements in it. To create CAD systems aimed at brain tumor identification using MR images, digital image processing approaches, such as preprocessing, segmentation, and classification, are applied. The classic machine learning and deep learning approaches for brain tumor identification are deliberated in this work. This paper provides a summary of commonly used MR image datasets. For classification, various machine learning and deep learning algorithms have been used. This survey examines current methodologies and can be used in the future to develop effective diagnostic plans for other brain disorders such as dementia, Alzheimer's disease, stroke, and Parkinson's disease using various Magnetic Resonance imaging modalities.