Vertebral Lamina State Estimation in Robotic Bone Milling Process via Vibration Signals Fusion

Abstract

During the robot-assisted laminectomy, vertebral laminas are removed via automatic milling operation, while inadequate estimation of the remaining lamina thickness will result in a dangerous bone milling. In this manuscript, a dynamic milling model is established to analyze the harmonic amplitudes in the milling vibration signals. It is proved that the ratios of harmonic amplitudes in the lamina’s and cutter’s vibration signals are sensitive to the remaining lamina thickness and insensitive to other milling motion parameters. To verify this concept, a robot is applied to mill and thin the porcine spinal lamina layer by layer. The vibration signals of the cutter and lamina are collected simultaneously with an accelerometer and a laser displacement sensor. Fast Fourier transform is applied to extract the amplitude of the harmonic components whose frequency are integer multiples of the cutter spindle frequency. The results show that the classifier can estimate two critical states of the vertebral lamina before the lamina cut off with a success rate of 98.4% based on these harmonic amplitude ratios. The proposed method is beneficial to improve the safety of the automatic bone milling operation of the spinal surgical robot.