This thesis concentrates on the development and optimization of a deep learning methodology for precise brain tumor segmentation utilizing multi-modal MRI data and advanced methods such as attention mechanisms and optimization techniques. Brain tumor segmentation is an essential part of diagnostic and treatment planning; yet, it presents difficulties due to the intricate morphology of tumors, fluctuating intensity levels, and the necessity for models that generalize effectively across heterogeneous datasets. This research incorporates attention mechanisms and optimization techniques into a CNN-based architecture to enhance the accuracy and robustness of tumor segmentation in both public (BraTS) and real clinical datasets. The proposed model employs multi-modal MRI sequences (T1, T2, T1ce, FLAIR) to acquire complementary information, hence enhancing segmentation accuracy by emphasizing certain features within each modality. Attention processes are integrated to improve the model’s capacity to discern crucial areas, especially in challenging instances with fuzzy tumor margins. Furthermore, innovative optimization methodologies, including the Multiverse Optimizer (MVO), Red Deer Algorithm (RDA), and Ebola Optimization Search Algorithm (EOSA), are utilized to enhance feature selection, optimize hyperparameters, and guarantee effective training. The findings indicate that integrating attention mechanisms with improved model parameters markedly enhances performance metrics, including the Dice Similarity Coefficient (DSC) and Intersection over Union (IoU), resulting in superior segmentation outcomes relative to contemporary approaches. To ensure practical applicability, the research includes the development of a Brain tumor segmentation using MRI images Docker-based API that enables straightforward deployment of the segmentation model in clinical environments, delivering real-time outcomes and seamless connection with hospital systems. Additionally, explainability techniques like Grad-CAM are utilized to elucidate the model’s decision-making process, hence augmenting trust and reliability for clinical applications. The research findings underscore the possibility of integrating deep learning, attention mechanisms, and explainability tools to develop effective and clinically pertinent solutions for brain tumor segmentation, hence enhancing diagnostic processes and patient outcomes.