Web-Based Framework for Brain Tumor Segmentation and Clinical Data Management

In neuro-oncological practice, early diagnosis and accurate volumetric assessment of brain tumors are of high importance, but manual MRI segmentation is labor intensive, expensive and subject to inter-observer variability. This paper describes the design, implementation, and rigorous empirical validation of an end-to-end medical imaging platform as a Django-powered medical imaging tool tailored for the needs of the Indian health care domain. The platform is a combination of 2D attention-based U-Net for accurate tumor segmentation, large language model (LLM) for generating intelligent reports in both English and Hindi, and secure standards-based longi- tudinal patient record management solution. By taking these key components into account, the platform aims to revolutionize the way neuroimaging workflows are handled, making them more efficient and accessible. Comprehensive evaluations on the BraTS 2020 dataset and 150 real-world MRI scans from Indian hospitals reveal exceptional segmentation accuracy with Dice scores exceeding 0.90, alongside superior report fidelity and adherence to clinical standards. Furthermore, the inclusion of detailed architectural blueprints and regulatory compliance measures underscores the platform’s readiness for practical, real- world deployment in resource-constrained environments.