Early and accurate diagnosis of brain tumors is crucial for improving patient outcomes. However, challenges such as healthcare worker shortages and resource limitations in South Korea highlight the pressing need for AI-driven diagnostic solutions. This paper introduces NeuroViT-Net, a hybrid Vision Transformer (ViT)-based deep learning model designed for brain tumor classification across both 2D and 3D medical imaging data. Unlike conventional CNN-based models (e.g., VGG and ResNet), NeuroViT-Net leverages self-attention mechanisms to capture global dependencies in MRI scans, achieving superior classification accuracy while maintaining computational efficiency. Notably, the model demonstrates enhanced performance in detecting Glioma and Meningioma, two of the most challenging tumor types, by effectively integrating spatial and volumetric information. Its hybrid processing capability allows simultaneous learning from 2D slice-based images and 3D volumetric MRI scans, improving robustness and adaptability to diverse clinical settings. Despite these advancements, future research will focus on enhancing detection of rare tumor types and exploring multi-modal extensions for broader neurological disorder diagnosis.