GANs for ultrasound denoising and simulation
Authors/Creators
Contributors
Project leader:
Project manager:
Project members:
- 1. Wageningen University & Research
-
2.
Instituto Superior Técnico
Description
In this work, we developed a pipeline for generating pixel-aligned ground truth image pairs to train a denoising model for b-mode ultrasound images.
Two pix2pix GAN models are provided for:
(1) denoising a bmode ultrasound image
(2) simulating a b-mode ultrasound image from a speckle free modality image.
The pytorch version of pix2pix was used, and an usage demo is provided in the linked github repository.
All development and testing was done on Google Colab.
If you find these models useful, please cite our article:
D. F. Vieira, A. Raposo, A. Azeitona, M. V. Afonso, L. M. Pedro and J. Sanches, "Ultrasound Despeckling With GANs and Cross Modality Transfer Learning," in IEEE Access, vol. 12, pp. 45811-45823, 2024, doi: 10.1109/ACCESS.2024.3381630.
Files
Files
(616.9 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:072201adb862c96d32f75753e97bc28d
|
616.9 MB | Download |
Additional details
Additional titles
- Alternative title
- Ultrasound denoising
- Alternative title
- Ultrasound simulation
Funding
Software
- Repository URL
- https://github.com/manyaafonso/ultrasound_denoising_GAN
- Programming language
- Python
- Development Status
- Active