Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{198703,
        author = {Dinesh Neela and Gunnala Keerthi and Paindla Vamshi Vardhan Reddy and Kothapally Tejashri Reddy and Anusha Dasari},
        title = {A Hybrid U-Net and Vision Transformer Framework for MRI-Based Brain Tumor Segmentation and Classification},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {9904-9913},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=198703},
        abstract = {Brain tumor diagnosis from Magnetic Resonance Imaging (MRI) remains a challenging task due to the complexity of tumor structures and the need for both precise localization and accurate classification. This paper presents MediVision, a unified deep learning framework that integrates segmentation and classification into a single, efficient pipeline. The proposed approach combines the strengths of a U-Net architecture for detailed tumor region extraction with a Vision Transformer (ViT) for robust classification based on global contextual understanding. The system begins with preprocessing steps including normalization, resizing, and augmentation to improve data quality and model generalization. The U-Net model performs pixel-level segmentation to accurately isolate tumor regions, which are subsequently analyzed by the Vision Transformer to classify images into clinically relevant categories such as glioma, meningioma, pituitary tumor, or normal cases. This hybrid design addresses the limitations of conventional CNN-based methods by capturing both fine-grained spatial details and long-range dependencies within MRI scans. Experimental evaluation demonstrates that the proposed model achieves high performance, with accuracy approaching 96%, while maintaining consistency across diverse imaging conditions. In addition to the core model, a web-based application built using FastAPI and React.js enables real-time interaction, allowing users to upload MRI scans, visualize segmentation outputs, and obtain diagnostic predictions. By integrating segmentation and classification within a single framework, MediVision reduces computational redundancy, improves diagnostic efficiency, and minimizes reliance on manual interpretation. The proposed system highlights the potential of hybrid deep learning architectures in advancing reliable, scalable, and accessible solutions for medical image analysis.},
        keywords = {Brain Tumor Detection, MRI Image Analysis, U-Net Segmentation, Vision Transformer (ViT), Deep Learning in Healthcare, Medical Image Classification},
        month = {April},
        }