Path Re-planning for Robotic Ultrasound Imaging of the Lumbar Spine


Ultrasound (US) scanning has become increasingly popular in a wide range of clinical applications as it provides radiation-free, compact and real-time imaging solutions for diagnosis and intraoperative navigation during interventions on, for example, the lumbar spine [1]. However, US image quality and the ability to interpret the images is highly dependent on the sonographer’s proficiency. Robotic approaches could be considered to offer a more standardised procedure and consistent performance. Robotic ultrasound systems could, for example, more easily ensure a constant contact force, target speed and orientations. This can be of benefit when aiming to 3D reconstruct complex anatomic structures with ultrasound for intraoperative guidance [2], [3]. Recent research in this field looks, for example, to deep learning to improve image quality or automatically segment anatomic regions [3]. However, besides following the skin surface profile to ensure contact with the skin, more research is needed on the control part to devise trajectories that optimise the visualisation of underlying targeted anatomies. This work focuses on path planning for US-based visualisation of the lumbar spine. To image bones, US waves travel through the skin and reflect or deflect on the bone depending on the inclination of the surface. Reflection and hence intensity of the image is maximal when the incident US wave is perpendicular to the bone surface. Therefore, optimising the orientation of the US probe with respect to the underlying anatomy could be especially beneficial in visualising complex bony structures, such as the lumbar spine. The sides of the articular and spinous process, beneficial for, for example, pedicle screw path planning, landmark identification and registration to corresponding 3D CT reconstructions, are rarely visualised when scanning at an angle perpendicular to the skin. Ma et al. demonstrate the use of laser distance sensors to automatically optimise the US probe orientation, but only with respect to the skin surface [4]. Li et al. further present a Shadow-aware Dual-Agent Framework to navigate towards standard views of the spine [5], but do not focus on changing the pose to optimise image quality or visualise inclined bony surfaces. This abstract presents a new robotic path planning approach that optimises the orientation of the US probe by relying on a first exploratory scan. The proposed method computes a new motion plan that improves the ultrasound image quality and hence the resulting 3D reconstructions of complex bony structures such as the lumbar spine.

12th Conference on New Technologies for Computer and Robot Assisted Surgery