Tele-Operative Low-Cost Robotic Lung Ultrasound Scanning Platform for Triage of COVID-19 Patients

Ryosuke Tsumura, John W. Hardin, Keshav Bimbraw, Anne V. Grossestreuer, Olushola S. Odusanya, Yihao Zheng, Jeffrey C. Hill, Beatrice Hoffmann, Winston O. Soboyejo, Haichong K. ZhangIEEE RA-L 2021 · DOI

TL;DR

This paper introduces a novel, low-cost tele-operative robotic platform designed for lung ultrasound (LUS) scanning, specifically to aid in the triage of COVID-19 patients. By enabling remote operation, the system significantly reduces the risk of transmission between healthcare workers and infected individuals, addressing a critical need in resource-limited settings.

What Problem Does This Paper Address?

The COVID-19 pandemic has highlighted the urgent need for rapid and accessible diagnostic tools, especially in resource-limited areas. Lung ultrasound (LUS) has emerged as a valuable, cost-effective, and radiation-free imaging modality for diagnosing COVID-19 related pulmonary issues. However, its widespread adoption faces several challenges:

  • Healthcare worker exposure: LUS requires close physical contact between the operator and the patient, increasing the risk of SARS-CoV-2 transmission.
  • Limited expertise: There's a shortage of trained sonographers and physicians skilled in performing LUS.
  • Operator dependency: The accuracy of LUS can be highly dependent on the operator's skill and experience.
  • Accessibility in resource-limited settings: The cost and availability of advanced imaging equipment can be prohibitive in many regions.

This paper aims to address these issues by developing a robotic platform that allows for tele-operative LUS, thereby minimizing physical contact and potentially democratizing access to this diagnostic tool.

Methods

The core of the proposed solution is a tele-operative robotic LUS platform with a focus on low cost and high usability. Key methodological aspects include:

  • Gantry-style Positioning Unit: Instead of complex multi-link robotic arms, the system utilizes a simpler gantry-style kinematics. This design offers sufficient reachability and maneuverability to cover all required chest scan regions (based on the modified BLUE protocol) while minimizing the risk of joint collisions with the patient. This approach is also more cost-effective and easier to implement in resource-constrained environments.
  • Passive-Actuated End-Effector: This component is designed to hold the ultrasound probe and apply a controlled contact force to the patient's body. It incorporates a passive mechanism using a constant spring to maintain a consistent, safe contact force (around 10 N) without relying on high-precision sensors or complex control algorithms. This passive approach also helps absorb patient body movements and individual anatomical variations. The end-effector also includes passive rotational axes to maintain the probe perpendicular to the body surface and an active rotational axis for fine-tuning the US image visualization.
  • Tele-operation Interface: The system allows operators to remotely control the robotic arm using a joystick. Visual feedback is provided by three cameras (top and two side views), and real-time audio communication is enabled via a two-way microphone.
  • Safety Features: The passive end-effector's design inherently limits the applied force, preventing excessive pressure on the patient. The system also monitors contact force and uses a Visual Analog Scale (VAS) to assess patient comfort, with a predefined threshold for terminating the procedure if discomfort is reported.

Datasets

The study did not utilize pre-existing datasets for training. Instead, it involved a pilot human subject study for validation.

  • Human Subjects: Three healthy subjects were enrolled for experimental validation. Their demographic and anthropometric data (age, sex, height, weight, chest dimensions) are provided in Table II.
  • Acquired Images: LUS images were acquired from these subjects using both the robotic platform and a conventional manual approach. These images were then used for qualitative assessment by expert physicians and for quantitative analysis of Contrast-to-Noise Ratio (CNR).

Key Insights and Ideas

  • Low-Cost Robotics for Healthcare: The paper demonstrates that a simplified, gantry-style robotic design with passive end-effector mechanisms can achieve the necessary functionality for medical imaging at a significantly lower cost than complex industrial robots.
  • Passive Mechanisms for Safety and Robustness: Relying on passive mechanical elements for force control and probe orientation offers inherent safety advantages, reducing the risk of system failures and making the system more robust to variations in patient anatomy and movement.
  • Tele-operation as a Pandemic Mitigation Tool: The study strongly supports the potential of tele-operated LUS as a critical tool for reducing healthcare worker exposure during infectious disease outbreaks, particularly in settings where resources are scarce.
  • Trade-off Between Safety and Efficiency: While the robotic system offers enhanced safety, it currently results in longer procedure times compared to manual operation. This highlights an important trade-off that needs to be considered and addressed in future development.

Key Results

  • Operational Safety: The mean contact force applied by the robot was comparable to manual operation (9.52 ± 1.02 N for manual vs. 10.48 ± 2.72 N for robot-assisted, with no significant difference, p = 0.003). The contact force remained within a safe range (never exceeding 15 N) and no pain or discomfort was reported by subjects.
  • Image Quality: Expert physicians scored the image quality from the robot-assisted LUS as comparable to manual LUS (mean scores of 7.85 ± 1.54 for robot vs. 8.21 ± 1.04 for manual, suggesting non-inferiority, p = 0.004). The Contrast-to-Noise Ratio (CNR) of the pleural line was also similar between the two methods (4.38 ± 0.95 for robot vs. 4.48 ± 0.70 for manual, also suggesting non-inferiority, p = 0.005).
  • Procedure Time: The robot-assisted LUS procedure took significantly longer than the manual operation (27.5 ± 5.4 min vs. 18.2 ± 3.2 min; p = 1).

Limitations and Future Work

The authors acknowledge several limitations and suggest avenues for future research:

  • Limited Fine Adjustment: The current end-effector orients the probe perpendicular to the surface but lacks active rotation adjustment, which can hinder the visualization of certain features like A-lines. Future work could involve implementing additional actuators for finer probe orientation control.
  • Patient Habitus: The passive force control might require adjustments for patients with significant subcutaneous tissue.
  • Lack of Cross-Validation: The study was conducted on healthy subjects. Further validation with COVID-19 patients and comparison with other imaging modalities (like CT scans) is necessary.
  • Manual Gel Application: The application of ultrasound gel remains a manual step, posing a potential risk of infection. Incorporating an autonomous gel dispenser is suggested for fully tele-operative LUS.
  • Procedure Duration: The longer procedure time is attributed to limited visibility from the current camera configuration and user interface. Optimizing camera placement, user interface design, and potentially incorporating augmented reality (AR) navigation systems are suggested to reduce this duration.
  • Cognitive Workload: The mental and cognitive workload for the remote operator was not explicitly studied and could be a factor in procedure efficiency.

Summary updated

BibTeX
@article{tsumura2021tele,
  author = {Ryosuke Tsumura and John W. Hardin and Keshav Bimbraw and Anne V. Grossestreuer and Olushola S. Odusanya and Yihao Zheng and Jeffrey C. Hill and Beatrice Hoffmann and Winston O. Soboyejo and Haichong K. Zhang},
  title = {{Tele-Operative Low-Cost Robotic Lung Ultrasound Scanning Platform for Triage of COVID-19 Patients}},
  journal = {IEEE RA-L},
  year = {2021},
  doi = {10.1109/lra.2021.3068702},
}
Figure