文章摘要

文章信息/Info

Title:: Intelligent positioning method for puncture points of intravenous indwelling needle puncture based on deep learning

Author(s):: JIA Haijing; GUO Keyou; School of Materials Science and Mechanical Engineering, Beijing Technology and Business University, Beijing 100048, China

Keywords:: Intravenous indwelling needle; Intelligent positioning; Faster RCNN; Non-maximal suppression; Puncture midline

摘要:: 本研究基于深度学习算法设计一套静脉留置针穿刺点智能定位方法，进而实现智能穿刺。基于深度学习算法Faster RCNN训练超声图像中血管自动检测模型，优化检测模型的网络架构与非极大值抑制参数，血管检测器AP=0.896。基于穿刺靶血管位于穿刺中分线上这一先验知识，确定穿刺靶血管，根据靶血管的物理位置以及留置针的穿刺角度，确定留置针在患者皮肤表层上的穿刺点的物理位置以及留置针的进针距离，进而机械手臂驱动留置针自动穿刺。在体膜上验证上述方案有效。

Abstract:: To design a set of intelligent positioning method for intravenous needle puncture point based on deep learning algorithm, and then achieve intelligent puncture. An automatic detection model of blood vessels in the ultrasound image based on deep learning algorithm Faster RCNN was trained, the detection model of the network architecture and non - maximum suppression parameters was optimized, vascular detector’s AP was 0.896. Based on the priori knowledge of the puncturing target vessel located on the puncture line, the puncturing target vessel was determined. Depending on the physical location of the target vessel and the puncture angle of the indwelling needle, the physical location of the puncturing point on the skin surface of the patient and the puncture distance of the indwelling needle were computed, and then the robot arm drived indwelling needle automatic puncture. Above program has been validated in the body membrane.