Research On State Sensing Method Of Robot-assisted Spine Surgery Based On Multi-source Information

Drilling pedicle is one of the key operations in the spinal surgery,which requires the surgeon to drill a hole in the pedicle with a bone drilling to implant the screw.During the process of drilling,the doctor needs to rely on his own experience to judge the state of the bone drilling in the pedicle.When the angle of the bone drilling is deflected or drilled through the pedicle,the surrounding blood vessels,nerves and spinal cord will be damaged,causing paralysis and even death to the patient.Thus the risk during the operation is great.In recent years,the application of orthopedic surgery robots has become more and more extensive and some complicated and difficult operations in surgery have been completed by robots.The perception of the pedicle drilling state can be further realized by CT images and force perception.The chance of an error during surgery reduces and the need for surgeon's surgical skills reduces.Aiming at the operation for the drilling of the spinal surgery robot,this paper proposes a robot state sensing method based on multi-source information.Aiming at the vibration of the vertebrae and the difference of the acoustic frequency during the operation of the drilling,a method of sensing the state of the bone layer based on vibration and acoustic signals is proposed respectively.In the process of processing the vibration signal,the vertebral segments are modeled and the vibration model is simplified.After the vibration signal is collected by the non-contact laser displacement sensor,the time-frequency transformation of the vibration signal is performed by using fast Fourier transform and wavelet transform.Harmonic quantities are processed using harmonic energy ratios to obtain perceptual results.In the process of processing the acoustic signal,after introducing the mechanism of sound signal generation,the energy characteristics of the acoustic signal are analyzed from the time domain and the frequency domain,and the frequency distribution based method is proposed based on the frequency domain.The processing of the acoustic signal is realized,and the effectiveness of the proposed method is verified by a plurality of different bone samples.Foe the reason that the acoustic signal changes sensitively and the force signal changes slowly,a method of fusion analysis of acoustic and force signals can be conducted.Research on information fusion is studied,and according to different fusion levels and methods,the appropriate processing methods are seleted.Data on acoustic and force signals are performed on the feature layer for convergence and comparison by the following methods,including simple convex combination algorithm,generalized pseudo-Bayesian fusion method and interactive multi-model fusion method.It is found that the fusion characteristic curve has better recognition effect than the single curve.The Support Vector Machine is used in the decision-making layer and the combination of Support Vector Machine and the D-S evidence theory is used to train and identify the feature quantities of the acoustic and force signals,and the recognition rate reaches 88%.The effectiveness of the proposed fusion method was verified and compared by using different bones and spine bones under multiple parameters.