CervixVisionAI is an Android-based application designed to assist frontline healthcare workers in low resource-settings perform highly accurate and accessible cervical cancer screening.
# CervixVisionAI
An AI-powered, smartphone-based screening tool that puts specialist-level
cervical cancer detection into the hands of frontline health workers in
low-resource settings.
Cervical cancer is preventable, yet kills over 350,000 women a year — 90% of
them in low- and middle-income countries, where Visual Inspection with Acetic
Acid (VIA) remains the dominant screening method despite its low sensitivity
and high inter-observer variability. CervixVisionAI closes that gap with an
ensemble deep-learning model that analyzes VIA images on-device and returns a
risk score, lesion classification, and explainable AI (Grad-CAM) output a
health worker can act on immediately.
## How it fits together
```
Healthcare Worker (mobile app)
│ capture image, quality check
▼
Backend API Gateway (Node.js/Express + Socket.io)
│ proxies analysis requests, persists results, pushes live updates
▼
AI Engine (FastAPI + ONNX ensemble: Swin Transformer + EfficientNet-B3/ConvNeXt)
│ risk_score, lesion_class, xai_output (Grad-CAM)
▼
Backend → MongoDB (results) + Socket.io (real-time notification to the app)
▼
Clinical decision support, treatment recommendation, EMR/HIS hand-off
```
The mobile app never talks to the AI engine directly — the backend acts as the
API gateway, routing analysis requests, persisting results in MongoDB, and
emitting Socket.io events back to the app in real time.
## Repository layout
| Path | Service | Stack |
|---|---|---|
| `frontend/` | Healthcare worker mobile app | React Native (Expo) — see frontend/README.md |
| `backend/` | API gateway, auth, MongoDB persistence, Socket.io | Node.js / Express / Mongoose — see backend/README.md |
| `ai_engine/` | Ensemble inference service (ONNX), training pipeline | Python / FastAPI / PyTorch — see ai_engine/README.md and ai_engine/TRAINING.md |
| `nginx/` | Reverse proxy / TLS termination configs (dev + prod) | nginx |
| `docker-compose.yml` / `docker-compose.prod.yml` | Local and production orchestration | Doc …