Deep Learning Meets Explainability in MRI Brain Tumor Analysis: A Review and Future Research Framework

Abstract Brain tumors (BT) constitute significant diseases that require an early diagnosis with accurate assessment to plan appropriate treatment strategies. MRI scans are utilized extensively to diagnose BTs;however, manual examination is both tedious and prone to inconsistencies. In the past few years, significant advances have been made in BTC using DL models, which have shown exceptional efficacy.Unfortunately, the black box problem associated with DL models prevents their wider implementation in healthcare institutions.The current review aims to explore recent developments in MRI-based BTC through the use of transfer learning and hybrid deep learning algorithms combined with XAI technologies.This review also explores recent developments in this field, identifies existing gaps in current research,and provides a conceptual model to advance BT identification systems.