Image-Guided Navigation of a Robotic Ultrasound Probe for Autonomous Spinal Sonography Using a Shadow-aware Dual-Agent Framework

Abstract

Ultrasound (US) imaging is commonly used to assist in the diagnosis and interventions of spine diseases, while the standardized US acquisitions performed by manually operating the probe require substantial experience and training of sonographers. In this work, we propose a novel dual-agent framework that integrates a reinforcement learning (RL) agent and a deep learning (DL) agent to jointly determine the movement of the US probe based on the real-time US images, in order to mimic the decision-making process of an expert sonographer to achieve autonomous standard view acquisitions in spinal sonography. Moreover, inspired by the nature of US propagation and the characteristics of the spinal anatomy, we introduce a view-specific acoustic shadow reward to utilize the shadow information to implicitly guide the navigation of the probe toward different standard views of the spine. Our method is validated in both quantitative and qualitative experiments in a simulation environment built with US data acquired from 17 volunteers. The average navigation accuracy toward different standard views achieves 5.18mm/5.25 • and 12.87mm/17.49 • in the intra-and inter-subject settings, respectively. The results demonstrate that our method can effectively interpret the US images and navigate the probe to acquire multiple standard views of the spine.

Reliability of Robotic Ultrasound Scanning for Scoliosis Assessment in Comparison with Manual Scanning

“…To reduce the effect of the human factor on spinal ultrasound examinations, several research groups proposed using robotic ultrasound systems [12][13][14][15].…”

“…No extensive study has been conducted to compare the performance of robotic ultrasound assessment for spinal assessment with standard manual ultrasound procedures. Li et al, [15] presented a robotic approach for ultrasound imaging-based navigation to locate spinal features using deep learning (DL) and reinforcement learning on dataset of 648 images of spinous process to imitate the human decision-making during US scanning. The design was evaluated in an ultrasound simulation environment, and the resulted average error on the test set was 9.5 ± 10.0mm.…”

Comparison of ultrasound scanning for scoliosis assessment: Robotic versus manual

“…To reduce the effect of the human factor on spinal ultrasound examinations, several research groups proposed using robotic ultrasound systems. [12][13][14][15] The robotic ultrasound system used in this study (Figure 1B) was designed for scoliosis assessment and was initially presented in Victorova et al 12 and extended in Victorova et al 16 It uses a DL-based navigation control to detect the centre of the spine, which is defined as spinous process tip, in real-time at each B-mode ultrasound frame; thus, there is no need for additional external sensors, such as a camera. The proposed robotic system can also adjust the probe orientation to keep the ultrasound probe normal to the surface and maintain good acoustic coupling with the back of the subject.…”