Remote Teleoperation of Endovascular Robots: A Network Emulation Framework for Systematic Performance Evaluation

Xingyu Chen, Nikola Fischer, Harry Robertshaw, Weiyuan Deng, Christos Bergeles, Thomas BootharXiv 2026 · arXiv

TL;DR

Researchers have developed a modular benchmarking framework that pairs physical endovascular robots with high-fidelity network emulation to systematically evaluate how network lag and packet loss affect surgery. Their findings reveal a critical "acceptability gap": clinicians can often successfully complete a procedure under poor network conditions, but they find the experience so untrustworthy that they deem it clinically unacceptable.

What Problem Does This Paper Address?

Mechanical thrombectomy (MT)—the physical removal of a blood pool clot—is the gold standard for treating acute ischemic stroke. However, "time is brain," and access to specialized interventional neuroradiologists is highly unequal, often requiring patient transfers that waste precious hours. Remote teleoperation of endovascular robots could bring the specialist to the patient virtually.

The challenge is that endovascular navigation is a delicate "human-in-the-loop" process. A surgeon manipulates a guidewire proximally based on distal visual feedback from fluoroscopy. When you introduce a telecommunication network into this loop, you introduce latency, jitter, and packet loss. While previous studies have shown remote surgery is feasible, we lack a systematic understanding of how specific network degradations impact not just the success of the task, but the clinician's confidence and willingness to use the system.

Remote teleoperated endovascular intervention setup showing the geographical separation between the operator and the patient-side robot.

Methods

The authors built a modular benchmarking framework that connects a physical endovascular robot to a control console through a configurable network emulation layer.

  • The Stack: They used the ns-3 network simulator to inject real-time impairments (bandwidth limits, delay, jitter, and loss) into the communication path. Commands were sent via the Zenoh protocol, while visual feedback used a dedicated GStreamer/WebRTC pipeline with H.264 encoding.
  • The User Study: 16 participants (8 experienced interventional clinicians and 8 engineering students) performed 480 navigation attempts in a 3D-printed vascular phantom.
  • Conditions: They tested 15 distinct network scenarios, ranging from "regional terrestrial" (15ms delay) to "geostationary satellite" (250ms delay), along with targeted sweeps of individual parameters like packet loss (up to 5%).
Fig. 2. Framework components and acquisition paths.

Datasets

The study generated a primary dataset of 480 navigation attempts. Evaluation was performed across:

  1. Technical Metrics: One-way latency, RTP packet loss, and video fidelity (SSIM, VMAF).
  2. Navigation Outcomes: Success rate, completion time, and error counts (e.g., entering the wrong vessel branch).
  3. Subjective Assessment: NASA-TLX workload scores, perceived delay, operator confidence, and clinician-rated acceptability.

Key Insights and Ideas

  • The "Success vs. Acceptability" Paradox: This is the paper's most striking finding. Under 5% packet loss, clinicians still succeeded in 87.5% of attempts. However, 100% of the clinicians rated this condition as unacceptable. This suggests that "task success" is a dangerous metric for certifying medical robots; we must measure operator trust.
  • Video Quality Trumps Latency: Within the tested ranges, perceived video quality had a much stronger correlation with clinician confidence (0.72) and acceptability (0.78) than perceived delay (0.52 and 0.61, respectively). If the image is clear, clinicians can adapt to a bit of lag; if the image stutters or artifacts appear, they lose trust immediately.
  • Bottlenecks in the Arch: Most of the time lost and errors made occurred during the initial navigation of the aortic arch. This phase requires high-frequency, precise corrections. When feedback is degraded, clinicians tend to "over-correct" or pause, leading to significantly higher procedural times.
Clinician acceptability ratings across different network conditions, highlighting the sharp drop in agreement as impairments increase.

Key Results

  • Performance Degradation: In the most impaired condition (C4: 250ms delay, 20ms jitter, 1% loss), the success rate dropped to 78.1% (from 93.8% in the reference condition), and the mean attempt time jumped from 63.5s to 94.3s.
  • Error Rates: The most impaired condition saw a 2.5x to 3.4x increase in errors per attempt compared to the reference.
  • Bandwidth Sensitivity: Reducing bandwidth from 40 Mbps to 30 Mbps caused a significant spike in packet loss (from 0.01% to 4.47%) and video latency, proving that even minor congestion can push a system past its "cliff" of usability.

Limitations and Future Work

The authors acknowledge several limitations that provide a roadmap for future research:

  • Visual Fidelity: The study used a standard webcam and a transparent phantom. Real clinical fluoroscopy has different contrast, noise, and texture characteristics that might interact differently with video compression.
  • Auditory Cues: Interestingly, some participants noted they used the sound of the robot motors in the room to help interpret lag. In a truly remote setup, this cue would be missing unless bidirectional audio is provided.
  • Dynamic Networks: The current emulation uses static parameters. Future work should explore "bursty" loss and dynamic congestion typical of real-world internet routing.

By providing a reusable, modular framework, this work moves the field beyond "it works" toward a rigorous, engineering-led definition of "it is safe and acceptable for clinical use."

Summary updated

Abstract

Remote endovascular teleoperation requires network performance to be evaluated in relation to physical navigation and operator experience, while the effects of network degradation on the usability of remote endovascular systems remain insufficiently characterized from the perspective of clinical users. We developed a network emulation and benchmarking framework that combines an endovascular robot with data streaming. We conducted a user study involving eight clinicians from interventional neuroradiology and general interventional radiology and eight engineering students, who performed 480 navigation attempts across 15 network conditions. Communication performance and received video quality were assessed alongside navigation outcomes, confidence, workload and clinician rated acceptability. Under the most impaired combined condition, success was 78.1\%, compared with 93.8\% under the reference condition, and mean attempt time increased from 63.5 to 94.3\,s. At 5\% configured packet loss, 87.5\% of attempts succeeded, although all eight clinicians rated the condition unacceptable. Perceived video quality was associated with clinician confidence and acceptability, with mean within participant Spearman correlations of 0.72 and 0.78, respectively. Framework integration on a second robotic platform additionally supported successful remote navigation over a real network. Our framework provides a reusable basis that can help identify communication bottlenecks and assess their effects on navigation and clinician acceptance for endovascular intervention telesurgery tasks.
BibTeX
@misc{chen2026remote,
  author = {Xingyu Chen and Nikola Fischer and Harry Robertshaw and Weiyuan Deng and Christos Bergeles and Thomas Booth},
  title = {{Remote Teleoperation of Endovascular Robots: A Network Emulation Framework for Systematic Performance Evaluation}},
  year = {2026},
  eprint = {2611.24280},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.24280},
}
Figure