🪁 PneumoScan AI

An advanced deep learning-powered web application for pneumonia detection from chest X-rays using Streamlit and PyTorch.

🌟 Features

• Real-time pneumonia detection from chest X-rays
• Interactive visualization with feature heatmaps
• Confidence meter for prediction reliability
• Detailed analysis reports
• Historical tracking of detections
• Responsive design with modern UI
• Comprehensive statistics dashboard

🚀 Live Demo

pneumoscan-ai-fyycvjlm6hh7xqhzasepey.streamlit.app/

🛠️ Tech Stack

• Python 3.8+
• Streamlit
• PyTorch & TorchVision
• OpenCV
• Plotly
• Pandas & NumPy
• PIL (Python Imaging Library)

👌 Prerequisites

• Python 3.8 or higher
• pip (Python package manager)
• Git

⚙️ Installation

1. Clone the repository:

git clone https://github.com/arpanksasmal/PneumoScan-AI.git
cd PneumoScan-AI

2. Create a virtual environment (recommended):

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

3. Install required packages:

pip install -r requirements.txt

💽 Project Structure

pneumoscan-ai/
├── app.py                  # Main Streamlit application
├── requirements.txt        # Python dependencies
├── style.css              # Custom CSS styles
├── model/
│   └── best_pneumonia_model_densenet121.pt  # Trained model weights
├── detection_history/     # Folder for storing detection logs
├── test_data/             # Sample chest X-ray images for testing
│   ├── sample_1.jpg
│   ├── sample_2.jpg
│   ├── sample_3.jpg
|   ├── sample_4.jpg
└── README.md             # Project documentation

🚀 Running the Application

1. Ensure you're in the project directory and virtual environment is activated
2. Run the Streamlit app:

streamlit run app.py

3. Open your browser and navigate to http://localhost:8501

📊 Model Architecture

The application uses a modified DenseNet121 architecture with:

• Custom classifier layers for binary classification
• Batch normalization and dropout for regularization
• Pre-trained weights for feature extraction

📊 Performance Metrics

Our model achieves the following performance on the test set:

• Accuracy: 87.82%
• Precision: 90.00%
• Recall: 88.00%
• F1 Score: 87.00%
• AUC-ROC: 96.87%

Cross-validation results show consistent performance across different data splits, with a standard deviation of less than 1.5% for all metrics.

🔐 Security Considerations

• The application is designed for testing and educational purposes
• Not HIPAA compliant by default
• Medical data should be handled according to relevant privacy regulations
• Use appropriate security measures when deploying in production

🌐 Deployment

Local Deployment

Follow the installation instructions above.

Cloud Deployment (Recommended Platforms)

• Streamlit Cloud (Easiest)

  1. Push code to GitHub
  2. Connect repository to Streamlit Cloud
  3. Configure requirements.txt
  4. Deploy

• Heroku

  1. Create Procfile:

     web: streamlit run app.py
     

  2. Configure runtime.txt with Python version
  3. Deploy using Heroku CLI or GitHub integration

• AWS/GCP/Azure

  • Use container services (ECS, GKE, AKS)
  • Configure environment variables
  • Set up load balancers if needed

📕 Test Data Instructions

To help users verify model predictions, we provide sample chest X-ray images in the test_data/ folder.

1. Navigate to the test_data/ folder in your project directory.
2. Choose an image (e.g., sample_1.jpg).
3. Upload the image using the Streamlit web app.
4. Click "Predict Pneumonia" to see the results.

This ensures a consistent and easy testing experience for users.

⚠️ Important Notes

• This tool is for screening purposes only
• Not a replacement for professional medical diagnosis
• Consult healthcare providers for medical decisions
• Regular model updates recommended for optimal performance

🤝 Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit changes (git commit -m 'Add AmazingFeature')
4. Push to branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

👍 License

This project is licensed under the MIT License – see the LICENSE file for details.

👏 Acknowledgments

• Special thanks to OpenAI and Deep Learning Research Community
• Dataset sourced from Chest X-ray Pneumonia Dataset

📄 Conclusion

This project serves as an advanced AI-driven tool for pneumonia detection. While it provides high accuracy and real-time predictions, it is intended for educational and research purposes only. Always consult medical professionals for official diagnoses.

📧 Contact

• Arpan Kumar Sasmal
• LinkedIn: www.linkedin.com/in/arpan-kumar-sasmal-2b4421240
• Email: arpankumarsasmal@gmail.com

---

Created with ❤️ by Arpan Kumar Sasmal 😎