Challenge
A clinical web platform that helps neonatal teams screen premature infants for Retinopathy of Prematurity (ROP), combining a hybrid fuzzy deep-learning model with patient timelines, medical history and PDF reporting.
Healthcare
A clinical web platform that helps neonatal teams screen premature infants for Retinopathy of Prematurity (ROP), combining a hybrid fuzzy deep-learning model with patient timelines, medical history and PDF reporting.
Challenge
A clinical web platform that helps neonatal teams screen premature infants for Retinopathy of Prematurity (ROP), combining a hybrid fuzzy deep-learning model with patient timelines, medical history and PDF reporting.
Solution
Retinopathy of Prematurity (ROP) is a leading cause of preventable childhood blindness, and screening depends on specialists who are scarce in many neonatal units. NeoVuo gives clinicians a secure workspace to register premature infants, upload right- and left-eye fundus images, and receive an AI-suggested ROP stage alongside their own clinical assessment.
What the platform does:
- Patient registry with gestational age, birth weight, and postnatal and post-menstrual age.
- Upload of JPG, PNG and DICOM fundus images per eye, with an AI-suggested stage, per-stage probabilities and confidence.
- Longitudinal timeline with a progression chart, side-by-side comparison, magnifier and freehand annotation.
- Medical history for each infant: conditions, treatments, perinatal risk factors, eye history and medications, with severity and status.
- Single-scan and full-timeline PDF reports for the medical record.
- Hospital portal for inviting and managing staff, plus an admin console that monitors the AI model and can wake it.
The model is a hybrid architecture: a CNN backbone and a transformer encoder extract image features, a fuzzy layer turns them into graded evidence, and gestational age and birth weight are combined at the classifier. It runs on a private Hugging Face Space and is called only by the server.
The platform was built with security first. A shared security core enforces sessions, CSRF protection, role checks, per-doctor data isolation, rate limits and validated uploads. Every view and change is written to an audit log.
Results
- A full security review found and fixed 17 issues, including unauthenticated uploads, remote code execution through image uploads, SQL injection, cross-doctor data access and stored XSS.
- Every doctor sees only their own patients. This was verified with cross-account tests on the web pages and on the JSON API.
- The interface was redesigned into a clean clinical theme built around status colours that clinicians recognise: green for no ROP, amber for stages 1–2, red for stage 3 and above.
- Four roles: doctor, hospital admin, admin and super admin.
- Live at app.neovuo.com, with a public product site at neovuo.com.
- Written up as an academic research paper (graduation project, MTI University, 2026).
Next project
Tolba – Local Ordering Marketplace
Next step
Tell us what you need. We call you back in the window you choose to plan it together.