This study explores the application of Magnetic Resonance Imaging (MRI) for brain tumor classification, addressing the challenges posed by the heterogeneous nature of tumors and associated low survival rates. MRI serves as a preferred non-invasive imaging modality due to its ability to produce high-contrast anatomical images without exposing patients to ionizing radiation. The research evaluates the performance of deep convolutional neural networks (CNNs) on a curated dataset comprising over 3,000 MRI scans, encompassing both healthy cases and various tumor types, including pituitary tumors, meningiomas, and gliomas. State-of-the-art architectures such as AlexNet, ResNet, and VGG16 are employed alongside a custom baseline model incorporating convolutional layers, max-pooling, and batch normalization. To further enhance classification performance, the Salp Swarm Algorithm (SSA) is utilized for optimal feature selection, enabling the extraction of the most discriminative features. The proposed framework achieves a classification accuracy of 95.28%, demonstrating its efficacy in improving diagnostic accuracy and computational efficiency. These findings suggest a promising direction for computer-aided diagnosis and effective treatment planning in neuro-oncology.