Medical Image Classifier

A deep learning application demonstrating pneumonia detection in chest X-ray images using TensorFlow and Streamlit.

Project Overview

This computer vision project showcases a complete CNN implementation for medical image classification. It includes a training pipeline, model evaluation, and an interactive web interface for image upload and prediction.

Features

• Deep Learning Model: CNN architecture for binary image classification
• Training Pipeline: Complete model training with evaluation metrics
• Web Interface: Streamlit app for image upload and prediction
• Model Persistence: Save and load trained models
• Error Handling: Robust error handling throughout the pipeline

Tech Stack

• Deep Learning: TensorFlow, Keras
• Computer Vision: OpenCV, Pillow
• Web Framework: Streamlit
• Data Science: NumPy, Pandas, Matplotlib
• Model Evaluation: scikit-learn

Project Structure

medical-image-classifier/
├── data/
│   ├── train/              # Training images folder
│   ├── test/               # Test images folder
│   └── val/                # Validation images folder
├── models/                 # Model storage
├── src/
│   └── model_training.py   # CNN training script
├── streamlit_app.py        # Web interface
├── requirements.txt        # Python dependencies
└── README.md              # Project documentation

Installation

1. Clone the repository:

git clone https://github.com/sunnynguyen-ai/medical-image-classifier.git
cd medical-image-classifier

2. Create virtual environment:

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

3. Install dependencies:

pip install -r requirements.txt

Usage

Training the Model

python src/model_training.py

Running the Web Application

streamlit run streamlit_app.py

Visit http://localhost:8501 to access the interface.

Model Implementation

• Architecture: CNN with convolutional blocks and global average pooling
• Training: Uses sample data for demonstration purposes
• Evaluation: Provides accuracy, precision, recall, and F1-score metrics
• Web Interface: Allows image upload with real-time predictions

Technical Details

Model Architecture

• Convolutional layers with ReLU activation
• Batch normalization for stable training
• Dropout layers for regularization
• Binary classification output with sigmoid activation

Training Features

• Early stopping to prevent overfitting
• Model checkpointing for best weights
• Comprehensive evaluation metrics
• Training history visualization

Development Notes

This project currently uses synthetic sample data to demonstrate the complete ML pipeline. The architecture and training process are designed to work with real medical imaging datasets when integrated.

Medical Disclaimer

This tool is for educational and demonstration purposes only. It should not be used for actual medical diagnosis. Always consult qualified healthcare professionals for medical decisions.

Future Enhancements

• Integration with real medical imaging datasets
• Docker containerization for deployment
• Additional preprocessing techniques
• Model performance optimization
• Extended evaluation metrics

Contributing

1. Fork the repository
2. Create feature branch (git checkout -b feature/improvement)
3. Commit changes (git commit -am 'Add new feature')
4. Push to branch (git push origin feature/improvement)
5. Create Pull Request

Contact

Sunny Nguyen

• GitHub: @sunnynguyen-ai
• Email: sunny.nguyen@onimail.com
• Website: sunnyinspires.com