Abstract Robotic ultrasound systems represent a potential solution for the automation of medical ultrasound examinations. By addressing the demographic change in our society and the increasing shortage of staff in clinics, they can provide reliable support in future patient care. In fully automated examination routines, the initial placement of the ultrasound probe on the patient’s body must also be automatically performed by the robotic system before the acquisition of ultrasound images can start. This paper proposes a method to overcome this challenge using soley RGB-D data, allowing for markerless operation in clinical settings. The approach builds upon an algorithm that maps 2D image pixels onto a reference body atlas. This allows for the calculation of a 6D transducer pose tailored to any patient’s anatomy. The method has been experimentally evaluated to achieve an optimized robustness to accuracy ratio, demonstrating its suitability for real-world applications.
Kriebisch et al. (Mon,) studied this question.