AI-powered lung cancer malignancy risk assessment tool combining deep learning with clinical risk factors

UCSC BioHacks Project 1st Place in AI Applications in Biomedical Research

LungOC is a hybrid decision support system that combines:

• Deep Learning: CNN-based image classification of CT scans
• Clinical Risk Assessment: Patient demographics (age, smoking history, family history)
• Risk Stratification: Provides actionable risk levels (Low/Moderate/High) with recommendations

• Backend: FastAPI + PyTorch (ResNet18 model)
• Frontend: React + TypeScript + Vite + Tailwind CSS
• Model: ResNet18 trained on lung CT scan dataset

• Python 3.12+
• Node.js 18+
• PyTorch, FastAPI, Streamlit (see requirements)

1. Backend (FastAPI)

cd src/backend
pip install -r requirements.txt
uvicorn main:app --reload

Backend runs at: http://127.0.0.1:8000

2. Frontend (React)

cd frontend
npm install
npm run dev

Frontend runs at: http://localhost:5173

Alternative: Streamlit Frontend

cd src
streamlit run app.py

See DEPLOYMENT.md for detailed deployment instructions to:

• Backend: Render.com (Free tier)
• Frontend: Vercel or Netlify (Free tier)

Health check and API info

Service health status

Predict lung cancer risk from CT scan

Request:

• file: CT scan image (JPG/PNG)
• age: Patient age (integer)
• smoking: Pack-years of smoking (integer)
• family_history: Family history (boolean)

Response:

{
  "prediction": "Malignant cases",
  "image_probability": 0.873,
  "final_risk": 0.742,
  "risk_level": "High"
}

• Architecture: ResNet18 (modified final layer for 3 classes)
• Classes: Benign, Malignant, Normal
• Input: 224x224 RGB images
• Framework: PyTorch

final_risk = 0.7 × image_malignancy_prob + 0.3 × clinical_score

clinical_score = 0.01 × age + 0.02 × smoking_pack_years + (0.1 if family_history)

This tool is a research prototype and not intended for clinical diagnosis. Always consult qualified healthcare professionals for medical decisions.

See LICENSE file for details.

Heli :)

• Deployment Guide
• API Documentation (when running locally)