Robust and Dense Visual SLAM for Robot-assisted Minimally Invasive Orthopaedic Procedures

Abstract

Orthopaedic surgeons are currently overburdened by physical and mental challenges that significantly increase the risk of injury to more than 2 million patients every year. This thesis proposes a robotic surgical-assistant for minimally-invasive orthopaedic surgeries that can alleviate surgeon workload and reduce the risk of unintended patient injury. A robotic prototype, along with a set of state-of-the-art robotic vision algorithms, was designed and validated in knee arthroscopy, the most common orthopaedic procedure worldwide. The proposed system can reliably inform surgeons of the location of an instrument within a detailed 3D map of the anatomy. Extensive experimentation, including cadaveric trials, demonstrated the system's unparalleled performance in real operative conditions.