Breakthrough Prediction in Robotic Laminectomy Based on Electrical Impedance Monitoring with LSTM-FCN

In robotic laminectomy, precise identification of cutting state and prediction of breakthrough are key guarantee for safe operation. Ultrasonic osteotome has been gradually adopted in laminectomy due to its characteristics of safely cutting bone tissue without hurting surrounding soft tissue. Utilizing the electrical impedance exported from the ultrasonic osteotome directly, this paper designed a LSTM-FCN-based breakthrough prediction model for robotic laminectomy. The model is used to identify whether the ultrasonic osteotome tip broke through the inner cortical bone. First, rationality of the electrical impedance for identifying cutting state is compared with the feedback force by calculating the Pearson correlation coefficient between them. Then, a breakthrough prediction model based on LSTM-FCN is proposed. Time delay and maximum breakthrough distance are used to characterize the prediction performance of the model. The results showed that there was a strong correlation between impedance and force with a Pearson correlation coefficient of 0.83; the prediction accuracy of breakthrough in the LSTM-FCN network is 90.78%, the maximum time delay of prediction is 22.2 ms, and the corresponding maximum breakthrough distance between the ultrasonic osteotome tip and the inner cortical bone is 0.09 mm, which satisfied the clinical requirements in robotic laminectomy. This model can be used to improve the operation safety of robotic laminectomy without introducing additional sensors.