Mathematical Modeling And Implementation Of Magnetic Laser Endoscopic Scanner

The laser endoscopic scanner is a new medical device for minimally invasive endoscopic surgery,which delivers laser through an optical fiber and employs a corresponding driving mechanism to complete the scanning of the laser to achieve cauterization and excision of the lesion.The laser endoscopic scanners overcome the disadvantages of traditional free-beam systems that require a direct line of sight for surgery and are more flexible,easier for the surgeon to operate,as well as more suitable for complex body structures.The magnetic laser endoscopic scanner is a laser endoscopic scanner that adopts the magnetic field as the driving mechanism.Due to the simplicity of mechanical structure,easy production and low cost,it is expected to be made into a disposable medical device,which can save the cost of sterilization and avoid cross infection.Moreover,its driving voltage and driving current are low,resulting in higher safety.However,the existing closed-loop control method of the magnetic laser endoscopic scanner is proportional-integral(PI)control,which has a relatively poor precision with a maximum error close to 0.2 mm and there is a large gap compared with its counterpart and advanced free-beam systems.To address the weaknesses of the magnetic laser endoscopic scanner,this paper designs a novel magnetic laser endoscopic scanner and its control platform.Based on this platform,the magnetic laser endoscopic scanner is controlled by an augmented model predictive controller to achieve higher precision trajectory scanning.For the implementation of the control of the magnetic laser endoscopic scanner the following work was done in this paper: According to the electromagnetic coil structure,the triple integral of the Biot-Savart Law is derived.And the final magnetic field model of the electromagnetic coil of the magnetic laser endoscopic scanner is obtained by using the Gauss-Legendre quadrature.On the basis of the final magnetic field model,a magnetic field excitation-current formula is developed that enables the calculation of the current in the four coils of the scanner from the desired magnetic field.Driving the magnetic laser endoscopic scanner with the magnetic field excitation-current formula and collecting excitation and response data.These data are used to train the amplitude function-LSTM(Long Short-Term Memory)neural network system model.After training,the amplitude function-LSTM neural network model is to be applied as a prediction model for the augmented model predictive controller.The augmented model predictive controller in this paper is to introduce the error integral compensation of the desired trajectory and the actual trajectory into the output of the model predictive controller to improve the control precision and robustness.By comparing the proposed augmented model predictive controller,PID controller,and conventional model predictive controller for scanning trajectory control,it is demonstrated that the proposed augmented model predictive controller is capable of achieving better control results for the magnetic laser endoscopic scanner,which realizes control precision with error less than 0.1 mm in circle and rectangle trajectories,and error less than 0.153 mm in eight-shaped and infinite trajectories.