RUSOpt: Robotic UltraSound Probe Normalization with Bayesian Optimization for In-plane and Out-plane Scanning

TL;DR

This work tackles the challenge of achieving high-quality ultrasound imaging with autonomous robotic systems across diverse patients by focusing on probe orientation normal to the contact surface. It introduces a sample-efficient Bayesian Optimization framework that uses wrist-wrist force sensing to identify the surface normal, augmented with a regularized objective function and EI-based acquisition. The approach is validated on urinary bladder phantoms with planar, tilted, and rough surfaces, as well as on simulated 3D human mesh models, achieving average angular errors around 2–3 degrees and demonstrating robustness across probe types and surface geometries. The results suggest a practical route toward autonomous, robust RUS deployment, with future work aimed at integrating ultrasound image quality into the optimization and extending to continuous, anatomy-aware scanning.

Abstract

The one of the significant challenges faced by autonomous robotic ultrasound systems is acquiring high-quality images across different patients. The proper orientation of the robotized probe plays a crucial role in governing the quality of ultrasound images. To address this challenge, we propose a sample-efficient method to automatically adjust the orientation of the ultrasound probe normal to the point of contact on the scanning surface, thereby improving the acoustic coupling of the probe and resulting image quality. Our method utilizes Bayesian Optimization (BO) based search on the scanning surface to efficiently search for the normalized probe orientation. We formulate a novel objective function for BO that leverages the contact force measurements and underlying mechanics to identify the normal. We further incorporate a regularization scheme in BO to handle the noisy objective function. The performance of the proposed strategy has been assessed through experiments on urinary bladder phantoms. These phantoms included planar, tilted, and rough surfaces, and were examined using both linear and convex probes with varying search space limits. Further, simulation-based studies have been carried out using 3D human mesh models. The results demonstrate that the mean ($\pm$SD) absolute angular error averaged over all phantoms and 3D models is $\boldsymbol{2.4\pm0.7^\circ}$ and $\boldsymbol{2.1\pm1.3^\circ}$, respectively.

Figures (6)

• Figure 1: The two rotations carried out to acquire urinary bladder ultrasound images. (a) describe the ideal probe orientation normal to the point of contact on phantom; (b) and (c) describe the non-ideal probe orientation during out-plane an in-plane rotation of probe. The ultrasound images acquired in these cases result in artifacts like edge or structure shadowing due to inappropriate acoustic coupling.
• Figure 2: Overview of the BO framework for identifying the normal of the robotized ultrasound probe to scanning region.
• Figure 3: Ultrasound probe generates reactive forces $[f_x, f_y, f_z]$ when a desired force $f_d$ is applied along its axis at a specific orientation to the scanning surface
• Figure 4: Experimental setup of RUS with urinary bladder phantom having three different scanning surfaces.
• Figure 5: Simulation environment of robotic ultrasound system with 3D human mesh models denoted as H0, H1 and H2.