The analysis of biomedical images, particularly brain MRI scans, is critical in healthcare and medical research. However, conventional approaches such as convolutional neural networks (CNNs) often struggle to capture complex spatial and contextual relationships within medical imaging data, limiting their generalization capabilities. To address these limitations, we propose a novel framework for brain MRI analysis using Vision Transformers (ViTs) augmented with a Shifted Patching Technique (S.P.T.). This approach combines conventional non-overlapping patching with shifted patching to better model intricate spatial dependencies and contextual associations. The proposed method is evaluated on the BRAIN MRI dataset, with standardized image resolution achieved through Lanczos5 interpolation. Further validation is performed on the ChestX-ray14 and Camelyon17 datasets to assess its generalizability across medical imaging modalities. Performance metrics such as accuracy, F1-score, and precision demonstrate that the proposed ViT model consistently outperforms traditional CNN-based methods. Ablation studies further confirm the significant contribution of the S.P.T. and other architectural components to overall performance. These findings highlight the potential of ViT-based architectures with innovative patching techniques as powerful tools for biomedical image analysis. This work offers a robust and scalable solution for enhancing diagnostic accuracy and advancing medical research applications.