This repository is free to use, copy, modify, and redistribute under the terms described in LICENSE.
If you use any part of this repository in another project, publication, or derivative work, please cite this repository as "PritamTheCoder/neurosight-ai" and include a link when reasonably possible.
NeuroSight AI is a 2D MRI slice-based tumor segmentation and classification assistant.
The system performs:
- Tumor segmentation (U-Net++)
- Tumor classification (EfficientNetV2-S)
- Grad-CAM visualization
- Confidence calibration
- Aggregated prediction (multi-slice support)
- Case history storage
This tool is designed for research and assistive clinical use only.
Frontend: React Backend: FastAPI Database: SQLite Models: PyTorch
Upload → Inference → Aggregation → Visualization → Storage
- Backend
cd backend
pip install -r ../requirements.txt
uvicorn main:app --reload
Frontend
cd frontend
npm install
npm start
POST /infer Upload MRI slice(s) and return:
- Tumor type
- Confidence
- Segmentation mask
- Grad-CAM
- Tumor area
This system is for research purposes only and not intended for autonomous medical diagnosis.
The license requires attribution to this repository when the code or derived work is reused.
Segmentation:
Classification:
