Brain Tumor Classification using Deep Learning: A State-of-the-Art Review
Mohammed Rasool, Abdulfatah Noorwali, Hamza Ghandorh, Nor Azman Ismail, Wael M. S. Yafooz
Abstract
Given that the number of available brain tumor images has grown, Deep Learning (DL) plays a critical role in brain tumor classification in terms of accurately diagnosing and predicting such tumors. Regarding the classification of several large-scale images, DL-driven techniques, such as convolutional neural networks, have not only shown significant results, but have also demonstrated that they can progressively learn features from data at multiple levels. As the use of medical imaging for analysis and education grows in popularity and the same occurs with the unstructured multi-faceted nature of the data, a state-of-the-art review of brain tumor classification is important. This study provides a systematic review of the state-of-the-art techniques and approaches utilized to classify massive Magnetic Resonance Imaging (MRI) data, especially for cancerous brain tissues. Thorough research was conducted on the subject of DL utilization in brain tumor classification based on studies between 2020 and 2023 derived from a variety of scholarly databases. Of the 142 studies retrieved, 20 were included to investigate the proposed or applied DL techniques for the recognition and categorization of brain tumors using MRI. A meta-analysis of current DL classification techniques, algorithms, and their validation was introduced. Overall, DL techniques should receive more attention due to their automatic and accurate feature extraction capacity.