HCT Survival Prediction

A robust machine learning-based system for predicting post-transplant survival outcomes in patients undergoing Hematopoietic Cell Transplantation (HCT). This project combines a FastAPI backend with a Next.js frontend to provide clinicians with an interactive, real-time decision-support tool that outputs personalized survival risk scores.

---

🚀 Features

• 🔍 Personalized survival risk prediction using clinical features
• 📊 Dashboard with variable definitions, data distribution, and model performance graphs
• 🧠 Ensemble of 10 ML models using LightGBM and CatBoost across five survival targets
• 📈 HLA-based feature engineering and fairness-aware training
• 🌐 FastAPI backend and Next.js frontend for real-time access and prediction

---

🎥 Watch Demo Video

🔧 Backend (FastAPI)

📋 Requirements

Make sure you have Python 3.10 or later installed.

Install the required Python libraries by running:

cd backend
pip install -r requirements.txt

▶️ Running the Backend

Navigate to the backend directory:

cd backend/

Start the FastAPI server using:

uvicorn main:app --reload

The API will be available at: http://localhost:8000

Note: Some model files are large (>100MB) and are tracked using Git Large File Storage (LFS). If you're cloning this repo, ensure Git LFS is installed and run:

git lfs install
git lfs pull

---

💻 Frontend (Next.js)

📋 Requirements

Ensure Node.js (v18+) and npm are installed.

▶️ Running the Frontend

Navigate to the frontend directory:

cd frontend/hct

Install the required packages:

npm install

Start the development server:

npm run dev

The app will be available at: http://localhost:3000

---

✅ Accessing the Application

• Go to http://localhost:3000
• Use the Home page to explore variable definitions and model insights.
• Use the Predict page to input patient data and generate a survival prediction score.