FusGAN: GAN-Based Ultrasound Simulation from CT Slices

Overview

This project utilizes Generative Adversarial Networks (GANs) to generate ultrasound simulations from CT slices and a transducer mask. By leveraging GANs, the system can produce realistic ultrasound intensity maps given a CT scan and a transducer mask input.

Features

• Ultrasound Simulation Generation: Convert CT slices into simulated ultrasound images.
• Mask Input: Utilize masks to define the transducer placement and orientation guide the simulation process and focus on specific regions.
• Customizable Settings: Adjust parameters to fit different use cases and requirements.

Installation

1. Clone the Repository:

   git clone https://github.com/aconesac/fusGAN.git
   cd fusGAN
   

2. Install Dependencies:

   It is recommended to use a virtual environment. Install the required Python packages with:

   pip install -r requirements.txt
   

   Make sure you have the necessary libraries for GANs and image processing, such as TensorFlow, NumPy, scikit-learn.

Usage

1. Prepare Your Data:

   • Place your CT slices and corresponding mask images in the data/ct_slices and data/tr_masks directories, respectively. Place the output simulations for training in data/pi_maps_.

2. Train the GAN:

   python train_gan.py --ct_data_path=data/ct_slices --mask_data_path=data/masks --sim_path=data/pii
   

   This command trains the GAN model using your CT and mask data and saves the trained model in the models/ directory.

3. Generate Ultrasound Simulations:

   python generateSimulation.py --ct_image_path=data/ct_slices/example_ct_slice.png --mask_path=data/masks/example_mask.png --model_path=models/trained_gan_model.h5 --output_path=results/
   

   This command generates an ultrasound simulation for a given CT slice and mask, saving the result in the results/ directory.

Examples

• Example Input: data/ct_slices/example_ct_slice.png, data/masks/example_mask.png, data/pii/sim_out.png
• Example Output: results/simulated_ultrasound.png

Notes

• Ensure your input CT slices and masks are properly aligned and preprocessed for optimal results.
• The performance and quality of the generated ultrasound images depend on the quality and quantity of the training data.

Contributing

If you'd like to contribute to this project, please fork the repository and submit a pull request with your changes.

License

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

Contact

For any questions or issues, please contact Agustin Conesa.