OrthoScrew ID

AI-powered orthopedic screw identification from x-ray images.

OrthoScrew ID was developed through a collaboration between medical and computer science researchers at Saint Louis University. The app addresses a real challenge in orthopedic revision surgeries: quickly and accurately identifying the manufacturer of previously implanted hardware. Upload AP and Lateral x-ray views, crop to the screw of interest, and the model returns ranked manufacturer predictions with confidence scores and similar reference cases.

Running the App

OrthoScrew ID is a Progressive Web App built with Expo, designed to provide a near-native mobile experience in the browser. It auto-deploys and is hosted on GitHub Pages.

Live app: https://austin-carnahan.github.io/iris-orthopedic/

Local Development

# Install dependencies
npm install

# Start for web
npm run web

The app will open at http://localhost:8081.

Authentication Setup

The app uses Google sign-in via Firebase Authentication. For login to work, you must configure whitelisting in both places:

1. Firebase Console → Authentication → Settings → Authorized domains — add your domain (e.g., localhost, austin-carnahan.github.io)
2. Google Cloud Platform → APIs & Services → Credentials → OAuth 2.0 Client → Authorized JavaScript origins and redirect URIs — add the same domains

Both must be configured or authentication will fail silently.

Firebase config is in firebase.js.

Machine Learning Backend

This app communicates directly with Hugging Face as its backend — there is no separate server. The ML code lives in the ml/ directory.

Resource	URL
Inference Space	austin-carnahan/orthopedic-screw-identification
Model Repo	austin-carnahan/orthopedic-screws-model
Dataset	austin-carnahan/orthopedic-screw-images

The app sends cropped x-ray images to the HF Space via the Gradio client library and receives manufacturer predictions with similar reference cases. See ml/space/README.md for API details.

Native Builds (iOS / Android)

Basic EAS Build configuration is provided in eas.json. To create native app builds:

eas build --platform ios     # or android

See the EAS Build introduction for setup and prerequisites.

Note: Native builds require additional setup before they will succeed:

• Add ios.bundleIdentifier and android.package to app.json
• Camera and photo library permissions may need to be configured in app.json plugins
• An Apple Developer account (iOS) or Google Play Console (Android) is needed for distribution