Artificial intelligence identifies milling status based on vibration signals in robot-assisted cervical laminectomy

Background: With the development of the science and technology, the application of artificial intelligence (AI) in the field of medicine has made great progress. The purpose of this study is to explore whether the machine learning k-nearest neighbors (KNN) can identify three milling states based on vibration signals, namely the cancellous bone (CCB), the ventral cortical bone (VCB) and the penetration (PT) in robot-assisted cervical laminectomy. Methods Cervical laminectomy was performed on the cervical segments of eight pigs by the robot. Firstly, bilateral dorsal cortical bone and part of the CCB were milled with the 5 mm blade and then the bilateral laminas were milled to penetration with 2 mm blade. During milling process of 2 mm blade, the vibration signals were collected by the acceleration sensor, and the harmonic components were extracted by the fast Fourier Transform (FFT). The feature vectors was constructed with the vibration signals amplitudes of 0.5 kHz, 1.0 kHz, 1.5 kHz and then the KNN was trained by the FV to predict milling states. Results The amplitudes of vibration signals between VCB and PT were statistically different at 0.5 kHz 1.0 kHz, and 1.5 kHz (P < 0.05), and the amplitudes of vibration signals between CCB and VCB was significantly different at 0.5 kHz and 1.5 kHz (P < 0.05). KNN recognition success rates of the CCB, VCB, and PT were 92%, 98%, and 100% respectively. 6% and 2% of CCB were identified as VCB and PT respectively, and 2% of VCB was identified as PT. Conclusions KNN can distinguish different milling states of the high-speed bur in robot-assisted cervical laminectomy based on the vibration signals. This method provides a feasible method to improve the safety of the posterior cervical decompression surgery.