Gliomas, which constitute approximately 33% of brain tumors, present considerable diagnostic and treatment challenges, particularly in low-resource settings such as Sub-Saharan Africa. While Magnetic Resonance Imaging (MRI) remains the gold standard for glioma evaluation, manual segmentation is time-consuming and inconsistent in environments with limited radiological expertise. To address this, we propose a self-supervised deep learning pipeline for automated glioma segmentation in multi-modal brain MRI. The framework leverages a SwinUNETR backbone with masked image modeling (MIM-SSL) pretraining to learn robust representations from unlabeled data, followed by fine-tuning on annotated subsets of the Brain Tumor Segmentation BraTS 2021 (BraTS 2021) and BraTS-Africa Challenge (BraTS-Africa 2024). Our model achieves Dice scores of 0.7987, 0.8311, and 0.8518 for Enhancing Tumor (ET), Tumor Core (TC), and Whole Tumor (WT), respectively, demonstrating improved segmentation accuracy and generalizability. This work underscores the value of self-supervised learning for addressing annotation scarcity and offers a practical solution for advancing neuro-oncological imaging in low-resourced regions.