Advances towards autonomous robotic suturing: Integration of finite element force analysis and instantaneous wound detection through deep learning

摘要

The automation of suturing is essential in robotic surgery to enhance precision and safety. This study addresses two challenges: analysis of needle-tissue interaction forces using a FEM model and real-time detection of puncture wounds using U-Net2D, SegNet, and DeepLabV3＋. DeepLabV3＋ demonstrated higher accuracy in wound detection, surpassing previous results. The system, with real-time processing and validated on a UR3 robot with an EndoWrist clamp, successfully executed an autonomous suture guided by wound detection. • Integration of FEM and CNN for precise autonomous robotic suturing. • DeepLabV3＋ outperforms previous models in real-time wound detection. • System tested with UR3 robot performs autonomous vision-guided suturing. • Significant advances towards full automation of surgical suturing. • Detailed analysis of needle-tissue forces enhances safety in robotic surgery.