查看文章

1 Purpose

Autonomous robotic systems for ultrasound (US) imaging could offer distinct advantages in several medical procedures, especially where great precision and repeatability is required. Among many, this includes needle insertions, quantitative measurements based on standardized views, therapeutic US and radiosurgical procedures, but also more complex soft-tissue surgeries. Robotic US may also reduce costs for examinations that are performed very frequently, such as screening for abdominal aortic aneurysms (AAAs), by automating the entire examination. While clinical use of autonomous robots is challenging due to both the complex and dynamic setting and the large anatomical variations between patients, new developments within machine learning, computer vision and collaborative robots have brought autonomous US examinations closer to reality. Robotic US has been an active field of research for many years [9]. Recently, more or less autonomous systems have been presented, especially for needlebased procedures [3, 5, 7], but also for abdominal artery examinations [13]. In addition to hardware such as robot arm, US scanner and cameras, many software components are needed to conduct research on ultrasound robotics. To our knowledge, there is currently no open-source framework available which supports US and 3D camera streaming, real time (RT) image processing including machine learning, as well as controlling a robot arm. Thus, in this work, our aim is to develop a new and generic open-source framework that integrates core libraries for building autonomous robotic US applications. Hopefully, this framework …