Engine Spare Parts,Large Engine Spare Parts,Car Engine Spare Parts,Diesel Engine Spare Parts NINGBO CITY YINZHOU RUICAN MACHINERY CO.,LTD , https://www.ruicanmach.com
New method for visual tracking of unmarked surgical instruments for mirror-holding robots
[ Instrument network instrument research and development ] Minimally invasive surgery is the inevitable trend and pursuit of surgical development. Laparoscopic surgery is a representative of minimally invasive surgery. It has been widely used in clinical practice due to its obvious advantages such as less trauma, faster recovery of patients, and fewer postoperative complications. It covers almost all operations in general surgery and has become the surgical treatment of many benign and functional diseases. "Gold Standard", the global annual number of cases exceeds 7.5 million. In addition to the surgeon, during laparoscopic surgery, a mirror holder must also be equipped to provide the surgeon with corresponding lighting and visual feedback. Prolonged operation makes the hand with the mirror prone to fatigue, hand tremor, and distraction, which affects the efficiency of the operation.
The robot has the advantages of high positioning accuracy, fast response speed, and strong movement stability, and is not affected by factors such as emotion and fatigue. It provides an effective solution to the problems faced by the hand holding the mirror during laparoscopic surgery. Active guidance based on endoscopic images is the interactive mode trend of mirror-holding robots. The core is the visual tracking of surgical instruments. Most of the existing methods use manual marking to track surgical instruments, but this method is not suitable for general surgical instruments. Additional disinfection and sterilization processes are needed in the clinic, and the scope of application is limited. Therefore, visual tracking of unmarked surgical instruments is more clinically universal. The environment of the abdominal cavity during the operation is complex and changeable. It is difficult to accurately locate the unmarked surgical instruments in the laparoscopic field of view, and the tracking speed cannot meet the real-time requirements of the surgeon for the robot with the scope, which greatly restricts the promotion of the robot And application.
Recently, Gao Xin's team from the Suzhou Institute of Biomedical Engineering and Technology of the Chinese Academy of Sciences proposed a lightweight target extraction neural network architecture that integrates region segmentation ideas and target search strategies to build a fully automatic label-free surgical instrument visual tracking method. The research builds a multi-instance segmentation model based on the LinkNet network, quickly extracts the different parts of the surgical instrument in the laparoscopic field of view, uses the structural characteristics of the surgical instrument, uses the joints of the surgical instrument as the tracking target, and combines the movement characteristics of the surgical instrument in the laparoscopic field of vision to introduce vision Contextual information of the tracking process to achieve precise tracking of surgical instruments.
The results of the study show that the proposed method achieves 100% tracking accuracy, 15 frames/second tracking speed, and low tracking accuracy on the publicly available laparoscopic surgery video data set m2cai16-tool (accumulated time is 9 hours 56 minutes, totaling 894,000 frames) Compared with the existing method, the positioning error of less than 6 pixels increases the tracking speed by 50% and reduces the positioning error by 30%, realizing high-precision and fast visual tracking of unmarked surgical instruments. This research provides a new research and development idea for the vision tracking module of the mirror robot.
The related results were published in International Journal of Computer Assisted Radiology and Surgery, and the research was funded by Jiangsu Provincial Department of Science and Technology and other institutions.