Challenge on Ultrasound Beamforming with Deep Learning (CUBDL) Datasets
The purpose of this challenge is to provide standardization of methods for assessing and benchmarking deep learning approaches to ultrasound image formation from ultrasound channel data that will live beyond the challenge.
- Participants had the freedom to create their own training data to build networks that accomplish specified tasks; this option is still available now that the challenge is closed.
- Specified tasks and evaluation methods are described on the challenge website: https://cubdl.jhu.edu/
- Participant submissions were facilitated by IEEE DataPort while the challenge was open
- Data sharing is facilitated by IEEE DataPort
Although the challenge is now closed, evaluation code remains available (more details on the challenge website https://cubdl.jhu.edu/), and datasets are available for release by submitting a signed user agreement (be sure to include all pages).
The following journal paper describes dataset details, top challenge submissions, and the evaluation process implemented by the challenge organizers:
D. Hyun, A. Wiacek, S. Goudarzi, S. Rothlübbers, A. Asif, K. Eickel, Y. C. Eldar, J. Huang, M. Mischi, H. Rivaz, D. Sinden, R.J.G. van Sloun, H. Strohm, M. A. L. Bell, Deep Learning for Ultrasound Image Formation: CUBDL Evaluation Framework & Open Datasets, IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control (accepted July 1, 2021) [pdf]