Bone Layer Proportion Recognition and Cutting Depth Control Based on Intraoperative Signal Processing in Robotic Laminectomy

Motivation: The easily displaced structure and the depth interference caused by the mixed bone layer make the depth-of-cut control of complex bone a key task for the automatic removal of hard tissue in orthopedics. Method: First, mappings for tool current and depth of cut are pre-established individually on cancellous bone and cortical bone, thereby obtaining two calibrated mapping functions. When the cutting area is with mixed bone tissues, the depth of cut is further predicted according to these two mappings and bone tissue proportion. Next, to obtain accurate bone tissue proportion, a vibration harmonic feature extraction method was introduced to construct a robust vector, and mainstream regression models were trained and tested. The effectiveness of the proposed feature extraction method is proved by comparative experiments, and the best regression model for the bone tissue proportion recognition task is selected. Finally, to avoid excessive instantaneous cutting energy during the robotic bone removal process, the controller parameter is optimized in real-time according to the bone tissue proportion. The accuracy and robustness of depth control when cutting complex bone tissue were tested by automatically cutting porcine scapula samples. Results: Experiment results show that when desired depths change within [0.5 1] mm, the depth of cut can be maintained at the set value, the maximum average depth deviation is less than 0.05 mm, and the maximum absolute depth deviation is less than 0.15 mm. Conclusion: this article enables surgical robots to achieve automatic cutting on complex bones based on sensing signal feedback.