Enhancing Transparency and Trust in Brain Tumor Diagnosis: An In-Depth Analysis of Deep Learning and Explainable AI Techniques

Brain tumors pose significant health risks due to their high mortality rates and challenges in early diagnosis. Advances in medical imaging, particularly MRI, combined with artificial intelligence (AI), have revolutionized tumor detection, segmentation, and classification. Despite the high accuracy of models such as Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs), their clinical adoption is hampered by a lack of interpretability. This study provides a comprehensive analysis of machine learning, deep learning, and explainable AI (XAI) techniques in brain tumor diagnosis, emphasizing their strengths, limitations, and potential to improve transparency and clinical trust. By reviewing 53 peer-reviewed articles published between 2017 and 2024, we assess the current state of research, identify gaps, and provide practical recommendations for clinicians, regulators, and AI developers. The findings reveal that while XAI techniques, such as Grad-CAM, SHAP, and LIME, significantly enhance model interpretability, challenges remain in terms of generalizability, computational complexity, and dataset quality. Future research should focus on addressing these limitations to fully realize the potential of AI in brain tumor diagnostics.