In this article, we introduce a sophisticated deep learning architecture designed to enhance the detection of brain tumors using multi-modal imaging data. This architecture integrates various imaging modalities, such as MRI, CT, and PET scans, to leverage the distinct advantages each has to offer in medical diagnostics. The core of the architecture comprises a series of interconnected layers that process and analyze the imaging data, extracting critical features essential for the identification of brain tumors. Our approach utilizes a unique arrangement of interconnected layers to refine feature extraction and increase the fidelity of tumor detection. This includes the novel application of dynamic routing within a capsule network to preserve the integrity of spatial relationships and hierarchical features, which are crucial for detailed and precise medical analysis. By synthesizing information across different imaging types and computational models, our architecture aims to provide a more comprehensive understanding of tumor characteristics. The efficacy of this architecture was evaluated through a series of tests, demonstrating its capability to effectively identify and classify brain tumors with high precision. This research not only contributes to the advancements in medical imaging analysis but also offers a potential pathway for improving diagnostic procedures and patient outcomes in neuro-oncology