Robotic versus Human Teleoperation for Remote Ultrasound

TL;DR

The paper addresses limited ultrasound access in remote communities and compares human versus robotic teleoperation for remote ultrasound. It implements two parallel systems with matched interfaces to isolate the follower modality and conducts randomized experiments with 15 volunteers performing a vessel-phantom scan, measuring task completion, image-space tracking, force consistency, and setup effort. Results indicate that human and robotic teleoperation yield similar overall performance, with humans delivering more consistent force and requiring substantially less setup time and cost; direct ultrasound remains fastest but does not provide remote access. These findings support using human teleoperation as a practical, lower-cost alternative for remote ultrasound in small communities, while highlighting areas for future work and consideration of real-patient imaging scenarios.

Abstract

Diagnostic medical ultrasound is widely used, safe, and relatively low cost but requires a high degree of expertise to acquire and interpret the images. Personnel with this expertise are often not available outside of larger cities, leading to difficult, costly travel and long wait times for rural populations. To address this issue, tele-ultrasound techniques are being developed, including robotic teleoperation and recently human teleoperation, in which a novice user is remotely guided in a hand-over-hand manner through mixed reality to perform an ultrasound exam. These methods have not been compared, and their relative strengths are unknown. Human teleoperation may be more practical than robotics for small communities due to its lower cost and complexity, but this is only relevant if the performance is comparable. This paper therefore evaluates the differences between human and robotic teleoperation, examining practical aspects such as setup time and flexibility and experimentally comparing performance metrics such as completion time, position tracking, and force consistency. It is found that human teleoperation does not lead to statistically significant differences in completion time or position accuracy, with mean differences of 1.8% and 0.5%, respectively, and provides more consistent force application despite being substantially more practical and accessible.

Figures (5)

• Figure 1: General diagram of teleoperation for ultrasound. The follower can be a robot or a person with a mixed reality headset. The expert interacts with a haptic device to input their desired motion and receive force feedback, while the follower interacts with the patient. Force and position are interchanged between the two pairs.
• Figure 2: Overview diagram of the robotic teleoperation system, showing the expert (left) and follower (right) sides, with haptic device and visualization for the expert, and a Franka Panda robot with BK ultrasound probe and tissue phantom on the follower side.
• Figure 3: Human teleoperation system, in which the follower aligns the real ultrasound probe with a virtual one (blue in the image) in a mixed reality headset. The expert side interface and communication system are the same as in Fig. \ref{['fig:roboDiagram']} for the robotic teleoperation.
• Figure 4: Diagram of the ultrasound scanning task on the branching vessel phantom. The typical ultrasound images from steps 1-5 are shown, and the steps are described in the text. The red lines in the ultrasound images show the subjects where to keep the vessels centered during the sweeps.
• Figure 5: Vessel eccentricity distribution, showing the magnitude and consistency of the applied force on the large and branch vessel sweeps. Robotic teleoperation shows a larger tail towards an eccentricity of 1, indicating higher and less consistent forces.