Accurate milling force estimation and surgical state recognition in robot-assisted laminectomy

Abstract

In robot-assisted laminectomy, it is difficult for the robot to judge the bone milling status. During the bone milling process, once the milling cutter penetrates the inner cortical bone, it is easy to damage the spinal cord, increasing the risk of surgery. This paper proposed a method to identify the bone milling state using the relative amplitude of the milling force signal and the vibration signal. A six-axis force/torque sensor was employed to estimate the milling force between the surgical tool and the bone, while a laser displacement sensor measured the vibration signals on the bone surface. The amplitudes of both signals were calculated separately. Considering that the vibration signal is affected by both bone density and milling depth, this paper divided the milling force amplitude by the vibration signal amplitude to obtain the relative amplitude that was only related to bone density. The vibration signal amplitude and relative amplitude were subsequently used to construct feature vectors, trained and classified using neural networks for identification of bone milling status. The experimental results show that the recognition rate of milling state is higher when the feature vector containing relative amplitude. When only the amplitude of vibration signal was input, the recognition rate of bone milling state reached 86.1%. After the relative amplitude was added, the recognition rate was increased to 96.86%. The proposed method is conducive to improving the safety of robot-assisted surgery.