title	Burn Assessment
emoji	🔥
colorFrom	red
colorTo	yellow
sdk	gradio
sdk_version	6.14.0
app_file	app.py
pinned	false

🔥 Burn Wound Assessment System (Web)

Web-deployable AI-powered burn injury assessment.

• 📸 Upload burn photo → AI vision analysis (Gemini)
• 📊 TBSA estimation via Lund-Browder chart (age-adjusted)
• 💧 Parkland fluid resuscitation calculation
• 📚 Clinical explanation via RAG (ChromaDB knowledge base + Gemini)

This is a single-file Gradio app deployed on Hugging Face Spaces.

---

🚀 Run Locally

1. Prerequisites

• Python 3.10 or newer
• Gemini API key (free): https://aistudio.google.com/apikey

2. Setup

git clone <this-repo-url>
cd burn_assessment_web

python3 -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

pip install -r requirements.txt

3. Configure API key

cp .env.example .env
# Edit .env, replace `your_key_here` with your actual Gemini key

4. Run

python app.py

Open http://localhost:7860

First run downloads the embedding model (~90 MB) and builds the RAG index (~15 sec total).

---

☁️ Deploy to Hugging Face Spaces

One-time setup

1. Create account at https://huggingface.co (free)
2. Get a write-enabled access token: Settings → Access Tokens → New token → Role: Write → save the value
3. Create new Space:
   • https://huggingface.co/new-space
   • Owner: your username
   • Space name: burn-assessment (or any slug)
   • License: MIT
   • SDK: Gradio
   • Hardware: CPU basic (free)
4. Add the API key as a Secret:
   • In the Space → Settings → Variables and secrets → New secret
   • Name: GEMINI_API_KEY
   • Value: your Gemini API key from aistudio.google.com

Deploy

The Space provides a Git URL. Add it as a remote and push:

cd burn_assessment_web

# Add the HF Space as a remote (use your username and space name)
git remote add hf https://huggingface.co/spaces/YOUR_USERNAME/burn-assessment

# Push (you'll be prompted for username and the access token from step 2)
git push hf master:main

HF Spaces will install dependencies and start the app (2–5 minutes). When done, the app is live at:

https://YOUR_USERNAME-burn-assessment.hf.space

Updating the deployment

After making changes:

git add .
git commit -m "your change"
git push hf master:main

The Space auto-rebuilds on each push.

---

📁 Project Structure

burn_assessment_web/
├── app.py                 # Gradio UI + Gemini vision + analyze() orchestrator
├── calculator.py          # Lund-Browder chart + Parkland formula (pure logic)
├── rag_engine.py          # ChromaDB retrieval + Gemini explanation
├── knowledge_base/        # 7 medical reference .txt files
├── tests/                 # pytest unit tests
├── requirements.txt
├── .env.example
└── README.md (this file)

---

🛠️ Tech Stack

• UI: Gradio 6.x
• AI Vision + Text: Google Gemini API (free tier)
• Vector store: ChromaDB (persistent)
• Embeddings: sentence-transformers/all-MiniLM-L6-v2
• Hosting: Hugging Face Spaces (free CPU)

---

⚠️ Medical Disclaimer

This tool is a clinical aid, not a substitute for professional medical judgment.

• AI TBSA estimates are approximations and must be clinically verified.
• Always confirm with a physical Lund-Browder chart.
• Parkland formula gives an initial estimate — adjust based on actual urine output.
• Final clinical decisions belong to the responsible clinician.

---

🧪 Run tests

source venv/bin/activate
pytest tests/ -v

Smoke tests that hit Gemini are skipped automatically if GEMINI_API_KEY is unset or still the placeholder.

MedAssistant