Updated July 5, 2021: Our journal paper describing challenge outcomes was accepted for publication: 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] If any of the resources shared on this website are useful to you, required citations include the above paper + the associated conference paper, which contains complementary information regarding challenge organization: MAL Bell, J Huang, D Hyun, YC Eldar, R van Sloun, M Mischi, “Challenge on Ultrasound Beamforming with Deep Learning (CUBDL)”, Proceedings of the 2020 IEEE International Ultrasonics Symposium, 2020 [pdf] + any reference to our associated datasets requires citation to the following database, which includes all contributors to the database as co-authors: Muyinatu A. Lediju Bell, Jiaqi Huang, Alycen Wiacek, Ping Gong, Shigao Chen, Alessandro Ramalli, Piero Tortoli, Ben Luijten, Massimo Mischi, Ole Marius Hoel Rindal, Vincent Perrot , Hervé Liebgott, Xi Zhang, Jianwen Luo, Eniola Oluyemi, Emily Ambinder, “Challenge on Ultrasound Beamforming with Deep Learning (CUBDL) Datasets”, IEEE DataPort, 2019 [Online]. Available: http://dx.doi.org/10.21227/f0hn-8f92 Visit our Resources & Citation page for more details.
Overview
The Challenge on Ultrasound Beamforming with Deep Learning (CUBDL) was offered as a component of the 2020 IEEE International Ultrasonics Symposium.
Recent developments in deep learning have created immense potential for ultrasound imaging research. CUBDL is designed to explore the benefits of using deep learning for both focused and plane wave transmissions. We challenge participants to obtain the best image quality under the fastest possible frame rates. To achieve this objective, we aim to provide answers to the following three questions:
- Can deep learning enable single plane wave imaging to match or significantly improve the performance of multi-plane wave imaging?
- Given a fixed number of transmission angles, what is the best performance that can be achieved when using deep learning to create images from multi-plane wave transmissions?
- Can deep learning provide the equivalent of dynamic transmit focusing using input data acquired with a single transmission focus?
We explore these three questions by providing participants with the following tasks, which are discussed in more detail here.
| Task | Objective | |
|---|---|---|
| Task 1a | Beamforming with deep learning after a single plane wave transmission | Task 1a is explicitly focused on creating a high-quality image from a single plane wave to match a higher quality image created from multiple plane waves. |
| Task 1b | Task 1b gives more freedom to create an image that will be benchmarked against the highest contrast, SNR, gCNR, etc. These values can be better than those obtained from an image formed by multiple plane waves. | |
| Task 2 | Beamforming with deep learning after a few plane wave transmissions | Task 2 imposes a maximum of 10 plane waves but lets participants choose from provided angles to create the best image quality possible. |
| Task 3 | Beamforming with deep learning to achieve dynamic transmit focusing | Task 3 enables participants to compare the results of a deep learning dynamic transmit focusing implementation that will be useful with current transmit beamforming techniques implemented on most clinical systems today. |
Organizers
Muyinatu Bell, Johns Hopkins University
Jiaqi (Justina) Huang, Johns Hopkins University
Dongwoon Hyun, Stanford University
Yonina Eldar, Weizmann Institute of Science
Ruud van Sloun, Eindhoven University of Technology
Massimo Mischi, Eindhoven University of Technology
