🐉 WYVERN

This project provides a GUI for chest X-ray (CXR) inspiration and rotation inspection. 🩻✨

🛠️ How It Works

The WYVERN application uses deep learning models to analyze chest X-ray images. Through its graphical user interface, you can:

• Click "Select Img" to choose images (.dcm or .png or .jpg) for inspection
• Click "Analyze" button to start the process
• We detect inspiration by segmenting the lungs and 9th posterior rib and calculate overlap ratio called 'Rib over Lung (ROL)' to determine if the inspiration is sufficient. If it above a certain threshold, we classify it as 'Adequate Inspiration', otherwise 'Inadequate Inspiration'.
• For rotation, we analyze the x position medial end of clavicles and the spinous process to determine if the patient is rotated. We calculate 'Alpha' which is quantify metric for access symmetry. If the alpha is above a certain threshold, we classify it as 'Rotated', otherwise 'Not Rotated'.

The models in src/models are used to perform the image analysis, ensuring accurate and fast inspection. The GUI makes it easy for users to operate without needing to write code.

🚀 How to Run

1. 📦 Install the required Python packages:

   pip install -r requirements.txt

2. 🧠 Place the required model files in the src/models directory.

3. 🎉 Run the main application:

   python main.py

📂 Directory Structure

├── README.md
├── requirements.txt
├── main.py                                 # main application file for the GUI (runs the application)
├── src/
│   ├── logo/
|   |   ├── Logo.png
|   |   ├── icon.png
|   |   ├── icon.ico
|   │   └── demo.png
│   └── models/
│       ├── lung_segmentation_model.pth     # pre-trained model for lung segmentation
│       ├── rib_segmentation_model.pth      # pre-trained model for rib segmentation
│       └── rotation_model.pth              # pre-trained model for rotation detection
├── result/
|   └── result.png                          # result image will be saved here
└── example_images/                         # example images for testing
    ├── EX166/                              # DICOM images
    ├── artifact/                           # images with artifacts
    ├── not full/                           # images with inadequate inspiration
    └── full/                               # images with adequate inspiration