Recent Advancements in Grad-CAM and variants: Enhancing Brain Tumor Detection, Segmentation, and Classification

Brain tumors, caused by abnormal tissue growth in the brain, can significantly impair brain functions and pose significant health risks. As the tumor progresses to higher stages, the patient's prognosis and survival decline, resulting in a high mortality rate. With advances in medical imaging, particularly the use of MRI, AI approaches have emerged as powerful tools for detecting, segmenting, and classifying brain tumors. CNN and hybrid models such as Vision Transformers (ViTs) have provided promising insights in this area. Despite their high accuracy, but the reasoning behind the prediction often remained unanswered. AI models lack transparency and interpretability, paving the way for the development of explainable AI methods in diagnosing brain diseases. In recent years, gradient weighted class activation mapping (Grad-CAM) and its variants have emerged as powerful techniques for visualizing and interpreting deep learning models in medical imaging tasks, including brain tumor detection, segmentation, and classification. This paper provides a comprehensive overview of Grad-CAM and its improvements, with particular emphasis on their applications in brain tumor analysis. In this article, we reviewed the 31 research papers based on data from various research databases. This review highlights the importance of interpretability in deep learning for clinical applications and explores how Grad-CAM can complement traditional metrics to provide deeper insights into model decisions. The results provide valuable guidance for researchers developing explainable AI frameworks for accelerating brain tumor analysis, with the objective of developing trust, efficiency and clinical utility in resource-limited settings.