AURA-CVC: Autonomous Ultrasound-guided Robotic Assistance for Central Venous Catheterization

TL;DR

This work addresses autonomous central venous catheterization (CVC) by developing an end-to-end robotic-ultrasound pipeline that initializes with depth-based anatomical landmark prediction, autonomously scans, localizes veins and arteries, and performs ultrasound-guided needle insertion with operator feedback. A Dense-UNet landmark predictor from depth images guides a triangular scanning region, followed by vessel segmentation, 3D reconstruction, and vein–artery differentiation via compression; a closed-loop needle controller finalizes puncture in the sagittal plane. Validated on a high-fidelity phantom across 10 scenarios, the system achieved 10/10 first-pass needle placements with mean vessel-centerline reconstruction errors around 2.15 mm and needle-tip targeting errors near or below 1 mm, indicating strong potential for clinical translation. Limitations include vein deformability, the use of a convex probe, and manual needle-tip annotation, with future work aiming at automated tip tracking and validation on more realistic phantoms and human subjects.

Abstract

Purpose: Central venous catheterization (CVC) is a critical medical procedure for vascular access, hemodynamic monitoring, and life-saving interventions. Its success remains challenging due to the need for continuous ultrasound-guided visualization of a target vessel and approaching needle, which is further complicated by anatomical variability and operator dependency. Errors in needle placement can lead to life-threatening complications. While robotic systems offer a potential solution, achieving full autonomy remains challenging. In this work, we propose an end-to-end robotic-ultrasound-guided CVC pipeline, from scan initialization to needle insertion. Methods: We introduce a deep-learning model to identify clinically relevant anatomical landmarks from a depth image of the patient's neck, obtained using RGB-D camera, to autonomously define the scanning region and paths. Then, a robot motion planning framework is proposed to scan, segment, reconstruct, and localize vessels (veins and arteries), followed by the identification of the optimal insertion zone. Finally, a needle guidance module plans the insertion under ultrasound guidance with operator's feedback. This pipeline was validated on a high-fidelity commercial phantom across 10 simulated clinical scenarios. Results: The proposed pipeline achieved 10 out of 10 successful needle placements on the first attempt. Vessels were reconstructed with a mean error of 2.15 \textit{mm}, and autonomous needle insertion was performed with an error less than or close to 1 \textit{mm}. Conclusion: To our knowledge, this is the first robotic CVC system demonstrated on a high-fidelity phantom with integrated planning, scanning, and insertion. Experimental results show its potential for clinical translation.

Figures (10)

• Figure 1: Overview of the experimental testbed showing the robotic system setup, a close-up view of the needle insertion mechanism, and the defined reference frames. The corresponding homogeneous transformation matrices, denoted as $\boldsymbol{T}_1^2$, represent the pose of frame 1 relative to frame 2.
• Figure 2: Our pipeline has four phases: the initialization phase , intra-operative planning, robot motion planning, and the needle guidance
• Figure 3: (a) Depth image of the phantom's neck and torso. (b) Landmark prediction network for anatomical localization. (c) Scanning region (yellow patch) on the neck is defined using three predicted anatomical landmarks. (d) Planned scanning paths are shown as maroon lines. (e) Probe poses along the scanning paths, with surface normals at each pose illustrated by light-blue lines.
• Figure 4: Robot motion planning worlflow to identify optimal insertion point, localize vein-artery, and move to sagittal (insertion) plane while avoiding collision.
• Figure 5: Criteria for selecting optimal needle insertion plane. $a_{j,i}$ is the area of the $j^{th}$ vessel in image $I_i$ from dataset $\boldsymbol{D}$ (Eq. \ref{['eq:rosbag_data']}), with $j$ limited to 2 to exclude noise, as only two vascular structures are expected to be visible.