Magnetic robotics obviate the physical connections between the actuators and end effectors resulting in ultra-minimally invasive surgeries. Even though such a wireless actuation method is highly advantageous in medical applications, the trade-off between the applied force and miniature magnetic end effector dimensions has been one of the main challenges in practical applications in clinically relevant conditions. This trade-off is crucial for applications where in-tissue penetration is required (e.g., needle access, biopsy, and suturing). To increase the forces of such magnetic miniature end effectors to practically useful levels, we propose an impact-force-based suturing needle that is capable of penetrating into in-vitro and ex-vivo samples with 3-DoF planar freedom (planar positioning and in-plane orienting). The proposed optimized design is a custom-built 12 G needle that can generate 1.16 N penetration force which is 56 times stronger than its magnetic counterparts with the same size without such an impact force. By containing the fast-moving permanent magnet within the needle in a confined tubular structure, the movement of the overall needle remains slow and easily controllable. The achieved force is in the range of tissue penetration limits allowing the needle to be able to penetrate through tissues to follow a suturing method in a teleoperated fashion. We demonstrated in-vitro needle penetration into a bacon strip and successful suturing of a gauze mesh onto an agar gel mimicking a hernia repair procedure.
Model selection and real-time skill assessment for suturing in robotic surgery
Department of Mechanical Engineering, Imperial College London（帝国理工学院机械工程系）；Department of Computing, Imperial College London（帝国理工学院计算机系）；Department of Mechanical Engineering, Stanford University（斯坦福大学机械工程系）；Intuitive Surgical, Inc.（Intuitive Surgical公司）；Department of Bioengineering, Imperial College London（帝国理工学院生物工程系）
Safe Uncertainty-Aware Learning of Robotic Suturing
SDU Robotics, The Maersk Mc-Kinney Moller Institute, University of Southern Denmark（SDU机器人研究所、马士基麦金尼莫勒研究所、南部丹麦大学）；Department of Mechanical Engineering, Myongji University（机械工程系、明州大学）
Relational Graph Learning on Visual and Kinematics Embeddings for Accurate Gesture Recognition in Robotic Surgery
The Chinese University of Hong Kong（香港中文大学）；Johns Hopkins University（约翰斯·霍普金斯大学）
自动手术手势识别对于机器人手术中的智能认知辅助至关重要。随着近年来机器人辅助微创手术的发展，可以记录包括手术视频和机器人运动学在内的丰富信息，这些信息为理解手术手势提供了补充知识。然而，现有的方法要么单独采用单模态数据或直接连接多模态表示，这不能充分利用视觉和运动学数据中固有的信息相关性来提高手势识别的准确性。在这方面，我们提出了一种新的多模态关系图网络的在线方法（即，MRG-Net）通过在潜在特征空间中的交互式消息传播来动态地整合视觉和运动学信息。具体来说，我们首先使用时间卷积网络和LSTM单元从视频和运动序列中提取嵌入。接下来，我们识别这些多模态嵌入中的多关系，并通过层次关系图学习模块来利用它们。我们的方法的有效性证明了国家的最先进的结果在公共JIGSAWS数据集，优于目前的单模态和多模态的方法在两个节点和结类型的任务。此外，我们在两个中心的da Vinci Research Kit（dVRK）平台收集的内部视觉运动学数据集上验证了我们的方法，并实现了一致的有希望的性能。
