Sepsis Detection Application

🩺 Overview

This web application provides a two-stage analysis system to detect sepsis risk in patients. The system first detects pneumonia anomalies in chest X-ray images, then combines these results with patient lab results and vital signs to assess the overall sepsis risk.

📋 Features

• Chest X-ray Analysis: Upload and analyze chest X-rays for pneumonia detection
• Clinical Data Integration: Input patient lab results and vital signs
• Two-Stage Risk Assessment: Combined analysis of imaging and clinical data
• Interactive Interface: User-friendly visualization of analysis progress and results
• Random Data Generation: Quick testing with auto-generated random clinical values
• Example Datasets: Built-in example X-rays for demonstration purposes

🔧 Technologies Used

• React.js
• TypeScript
• Tailwind CSS
• ShadCN UI Components
• Lucide React Icons

🚀 Getting Started

Prerequisites

• Node.js (v16 or higher)
• npm or yarn

Installation

1. Clone the repository:

   git clone https://github.com/your-username/sepsis-detection-app.git
   cd sepsis-detection-app

2. Install dependencies:

   npm install
   # or 
   yarn install

3. Start the development server:

   npm run dev
   # or
   yarn dev

4. Open your browser and navigate to http://localhost:3000

🏗️ Main Project Structure

sepsis-detection-app/
├── public/
│   └── assets/
│       └── xrays/     # Sample X-ray images
├── src/
│   ├── app/    # CSS and main app
│   └── components/
│       └── ui/        # UI components
│           └── Sepsis_Detection_App.tsx  # Main application component
├── README.md
├── package.json
└── tsconfig.json

📊 API Endpoints

The application connects to two machine learning models:

1. Pneumonia Detection API

   • Endpoint: https://dsc180-resnet.bobbyzhu.com/predict
   • Method: POST
   • Input: Chest X-ray image (JPEG format)
   • Output: Binary classification (pneumonia detected/not detected)

2. Sepsis Risk Assessment API

   • Endpoint: https://dsc180-rf.bobbyzhu.com/predict
   • Method: POST
   • Input: JSON payload with clinical parameters and pneumonia detection result
   • Output: Binary classification (sepsis risk high/low)

📝 Clinical Parameters

The application uses the following clinical parameters for sepsis risk assessment:

Parameter	Normal Range	Unit
Bilirubin	0.3-1.2	mg/dL
Creatinine	0.7-1.3	mg/dL
Heart Rate	60-100	bpm
INR	0.8-1.2	ratio
Mean BP	70-100	mmHg
Platelets	150-450	K/µL
PTT	25-35	seconds
Resp Rate	12-20	/min
Systolic BP	90-120	mmHg
WBC	4.5-11.0	K/µL
Temperature	36.5-37.5	°C
Bands	0-10	%
Lactate	0.5-2.0	mmol/L

⚠️ Important Note

This application is a clinical decision support tool. Results should be interpreted by healthcare professionals in conjunction with other clinical findings and patient history. This tool does not replace clinical judgment.

📸 Screenshots

Initial Upload Screen

Results Display Screen

🧪 Model Details

Pneumonia Detection Model

• Architecture: ResNet-based convolutional neural network
• Training Dataset: Chest X-ray dataset (AP/PA views)
• Performance: 92% accuracy on test set

Sepsis Risk Assessment Model

• Architecture: CatBoost classifier
• Features: 13 clinical parameters + pneumonia detection result
• Performance: 78% accuracy on test set

👥 Team

• Ahmed Mostafa - Project Manager
• Bobby Zhu - Cloud Engineer
• Ojas Vashishtha - ML Engineer
• Raine Hoang - Data Engineer
• Rohan Duvur - ML Engineer
• Tongxun Hu - Data Scientist

🙏 Acknowledgements

• Professor Kyle Shannon for his mentorship and guidance throughout the project
• Halıcıoğlu Data Science Institute for supporting this research

---

© 2025 Sepsis Risk Assessment | UC San Diego