Abstract

Brain imaging reveals fine-grained tissue details across the four cerebral lobes. Accurate partitioning of brain images into distinct, homogeneous regions is a prerequisite for early and precise diagnosis, since MRI serves as a fundamental imaging modality for radiologists to visualize internal brain structures. Despite advances in imaging and analysis, brain tumors continue to exhibit one of the highest mortality rates worldwide. Improved accuracy in tumor region identification enables more reliable interpretation of radiological findings and supports effective clinical decision-making. In this study, a comprehensive and systematic comparison is presented by explaining the operational principles of classical and AI-driven segmentation architectures using common evaluation criteria, including procedural workflows, performance metrics, dataset employed, methodological novelties, and inherent limitations. The comprehensive performance analysis indicates that Cellular Automata (CA)-based methods are computationally efficient with satisfactory segmentation performance. In contrast, AI-based methods generally involve higher computational complexity and longer processing time. Furthermore, the integration of explainable AI (XAI) enhances segmentation accuracy and robustness, enabling reliable detection of diverse tumor types across heterogeneous magnetic resonance imaging (MRI) data. The exploration of XAI is essential for bridging scientific advancements with real-world healthcare requirements, particularly in the development of reliable and clinically acceptable segmentation architectures. The proposed work is expected to be highly beneficial for researchers in MRI brain image segmentation, as it presents a comprehensive overview of existing technologies and their implementations, thereby supporting the design and development of novel, improved, and computationally efficient segmentation methods. Furthermore, there is a critical need to develop lightweight, real-time segmentation approaches to enable seamless clinical deployment, improving efficiency in patient diagnosis and treatment planning.

Authors

Imran Qadir, V. Devendran
Lovely Professional University, India

Keywords

Brain MRI Image, XAI, Machine Learning, Deep Learning, Image Segmentation, Cellular Automata