Loading...
Final Year Design Project · NED University
Sharper medical images and draft reports, in one workflow.
SuperRay combines generative super-resolution, image classification and LLM-assisted reporting to help radiologists review X-rays and CT scans. It is a research prototype and is not intended for clinical use.
Chest radiographSample image
How it works
From upload to reviewed report
- 01
Upload
A PNG or JPEG X-ray or CT image is uploaded through the browser and validated by the Django REST API.
- 02
Super-resolution
An SR-GAN model upscales the image to recover detail lost to low resolution or poor acquisition.
- 03
Classification
A DenseNet classifier identifies the imaging domain, such as chest, and screens out non-medical images.
- 04
Report drafting
A language model drafts a diagnostic report from the model outputs for a radiologist to review.
- 05
AI radiologist chat
Signed-in users can save reports and ask follow-up questions about a report in a chat assistant.
Project walkthrough
SuperRay in a nutshell
A recorded walkthrough of the original system, including image upload, report generation and the AI radiologist chat, captured while the backend and models were running.
Project status
What is running today
The original backend and GPU inference service are offline, so this deployment runs in demonstration mode. Pages that need them show clearly labelled sample data instead of live results.
View the frontend and backend source code| Component | Runs on | Status |
|---|---|---|
| Web interface | Next.js frontend on Vercel | Online |
| Accounts and saved reports | Django REST API with PostgreSQL | Offline · sample data |
| SR-GAN, DenseNet and LLM inference | External GPU inference service | Offline · sample data |
| AI radiologist chat | Depends on the inference service | Offline · sample data |