USB Gym: A Gymnasium Testbed for Closed-Loop Ultrasound View Matching
USB Gym: A Gymnasium Testbed for Closed-Loop Ultrasound View Matching
This research introduces USB Gym, a simulation environment designed to evaluate closed-loop ultrasound guidance methods. The testbed simulates ultrasound images from public patient CT scans and integrates three replaceable components—organ segmenter, controller, and reward function—to benchmark 20 published methods. Results show that combining these methods did not significantly improve probe positioning compared to the original guide, with the best combination achieving only 13.7 mm accuracy versus a pose oracle's 1.5 mm. The authors emphasize that no method was tested on physical hardware, and the task remains under-explored.
The study presents USB Gym, a simulation testbed for closed-loop ultrasound view matching that combines published methods into a benchmark baseline; however, the authors note that no guide was tested on hardware and results remain exploratory.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- Environment
- SimulationOpen source S2
Reported numbers
probe distance from target view
16
View original evidence
nchmark baseline. A pose oracle, reading the exact probe pose available only in simulation, set the reference. The results below are exploratory. Results: On these CTs, the probe started on average 16.0 mm from the target view; the pose oracle brought it to 1.5 mm. Of the results known when the combination was fixed, only an image-similarity reward (SSIM) was slightly but detectably better, so the combination was our
Open source S4probe distance after oracle correction
1.5
View original evidence
nchmark baseline. A pose oracle, reading the exact probe pose available only in simulation, set the reference. The results below are exploratory. Results: On these CTs, the probe started on average 16.0 mm from the target view; the pose oracle brought it to 1.5 mm. Of the results known when the combination was fixed, only an image-similarity reward (SSIM) was slightly but detectably better, so the combination was our
Open source S4number of episodes per patient
6
View original evidence
nt CT, with a network drawing organ textures from CT labels. Its guide has three replaceable parts (slots), an organ segmenter, a controller and a reward, which received 20 published methods, each run on the same 60 episodes of six held-out patients. In each slot, a published method replaced ours only if detectably better on these patients. The better of this combination and our original guide on 20 new CTs is the be
Open source S3number of published methods tested
20
View original evidence
nt CT, with a network drawing organ textures from CT labels. Its guide has three replaceable parts (slots), an organ segmenter, a controller and a reward, which received 20 published methods, each run on the same 60 episodes of six held-out patients. In each slot, a published method replaced ours only if detectably better on these patients. The better of this combination and our original guide on 20 new CTs is the be
Open source S3number of new CTs for benchmark
20
View original evidence
nt CT, with a network drawing organ textures from CT labels. Its guide has three replaceable parts (slots), an organ segmenter, a controller and a reward, which received 20 published methods, each run on the same 60 episodes of six held-out patients. In each slot, a published method replaced ours only if detectably better on these patients. The better of this combination and our original guide on 20 new CTs is the be
Open source S3
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.08352v1 Announce Type: cross Abstract: Background and Objective: Our main goal is ultrasound guidance: moving a probe until its image matches a target view. Published methods for each part of a guide are evaluated in separate simulators. We asked whether these methods, combined into one guide, are good enough. Methods: We built USB Gym, a testbed whose standard engine simulates ultrasound from public patie
Open source S2
nchmark baseline. A pose oracle, reading the exact probe pose available only in simulation, set the reference. The results below are exploratory. Results: On these CTs, the probe started on average 16.0 mm from the target view; the pose oracle brought it to 1.5 mm. Of the results known when the combination was fixed, only an image-similarity reward (SSIM) was slightly but detectably better, so the combination was our
Open source S4
nt CT, with a network drawing organ textures from CT labels. Its guide has three replaceable parts (slots), an organ segmenter, a controller and a reward, which received 20 published methods, each run on the same 60 episodes of six held-out patients. In each slot, a published method replaced ours only if detectably better on these patients. The better of this combination and our original guide on 20 new CTs is the be
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org