CN112214931B - Electromagnetic force feedback device and method for virtual interventional operation system - Google Patents

Abstract

The invention discloses an electromagnetic force feedback device and method for a virtual interventional operation system. The device includes an electromagnetic coil array and a surgical instrument; the electromagnetic coil array consists of four electromagnetic coils of identical size distributed on the same plane with a specific topology structure Composition; The surgical instrument consists of two long cylindrical permanent magnets and a rigid rod-shaped operating rod. At the same time, the present invention proposes an electromagnetic force feedback method suitable for the device, including the basic operation principle of the device, an optimal current distribution strategy and a torque-current regression prediction model, which is used to generate virtual simulations in a concise, efficient, real-time and accurate manner. Circumferential rotational force feedback is generated in interventional procedures. The invention greatly restores the operation mode of the real interventional operation, improves the immersion of the operator, and can improve the effect of preoperative training of the virtual interventional operation system.

Description

An electromagnetic force feedback device and method for virtual interventional surgery system

technical field

The invention belongs to the technical field of virtual reality, and relates to an electromagnetic force feedback device and method for virtual interventional surgery, in particular to a topological structure of an electromagnetic coil array and corresponding surgical instruments designed based on the principle of magnetic levitation, and a Electromagnetic force feedback method of machine learning technology for accurate real-time generation of circumferential rotational force feedback in virtual interventional procedures.

Background technique

In the virtual vascular interventional surgery system, the user interacts with the virtual vascular scene by operating surgical instruments that are highly similar to the guide wire catheter, so as to obtain visual and tactile multi-sensory feedback, and experience the process of interventional surgery with high immersion ([Reference 1] ). The difficulty of the virtual interventional surgery system is to accurately generate the feedback force during the interventional surgery operation in real time, especially the circumferential rotation feedback force during the guide wire twisting process ([Reference 2]). However, most of the force feedback modules in the existing virtual interventional surgery systems are designed with contact-type mechanical guide rails and pulleys ([Document 3][Document 4]). The existence of mechanical guide rails greatly limits the freedom of operation, and there is a big gap with the actual implementation of interventional surgery, resulting in an unsatisfactory training effect of the virtual surgical system. With the maturity of magnetic levitation technology, its many advantages are gradually reflected, not only can it obtain a more flexible operating space, but also have the advantages of precise motion control and low energy consumption, in addition to eliminating friction and dynamic non-consumption in other driving methods. Linear hysteresis, etc. ([Literature 5]).

On the electromagnetic force feedback device based on magnetic levitation, Berkelman ([Literature 6] [Literature 7]) et al. first used a single electromagnetic coil as the basic unit with a specific topology of electromagnetic coil arrays and corresponding permanent magnets. The lever is used to obtain non-contact force-tactile perception, but the basis for the placement of the electromagnetic coil is not explained too much. The research group of Yuan Zhiyong of Wuhan University ([Reference 8]) applied electromagnetic force feedback technology to the force feedback module of virtual surgery system, especially for elastically deformable organs such as kidneys. The experimental results show that the operation experience based on the electromagnetic force feedback module is stronger than the mechanical immersion, but the research focuses on the stiffness perception of virtual tissue. So far, there is still a lack of systematic research on electromagnetic force feedback in complex virtual interventional surgery systems, and the technical difficulties lie in reproducing the surgical operation mode with high degree of reduction and accurately generating force feedback during the operation in real time.

To meet the real-time and accuracy of electromagnetic force feedback in complex virtual access surgery, it is necessary to establish a fast and efficient calculation model between the electromagnetic force and the excitation current of each coil. According to the characteristics of the topology structure and the operating instrument, the invention creates a concise and accurate calculation model in combination with methods such as model fusion, which ensures the real-time and accuracy of the force feedback.

references:

[Document 1]: Omisore O M, Han S P, Ren L X, et al. Towards Characterization and Adaptive Compensation of Backlash in a Novel Robotic Catheter System for Cardiovascular Interventions [J]. IEEE Transactions on Biomedical Circuits&Systems, 2018, PP(4): 1- 15.

[Document 2]: Shi Y, Zhou C, Xie L, et al. Research of the master-slave robotsurgical system with the function of force feedback [J]. Int J Med Robot, 2017: e1826.

[Document 3]: Guo J, Jin X, Guo S, et al. A Vascular Interventional SurgicalRobotic System Based on Force-Visual Feedback [J]. IEEE Sensors Journal, 2019, 19(23): 11081-11089.

[Document 4]: Guo J, Yu Y, Guo S, et al. Design and performance evaluation of novel master manipulator for the robot-assist catheter system [C]//IEEE International Conference on Mechatronics & Automation. IEEE, 2016: 937-924.

[Literature 5]: Kim Y, Parada G A, Liu S, et al. Ferromagnetic soft continuumrobots [J]. Science Robotics, 2019, 4(33): eaax7329.

[Document 6]: Berkelman P, Dzadovsky M. Magnet levitation and trajectory following motion control using a planar array of cylindrical coils [C]//ASME2008 Dynamic Systems and Control Conference. American Society of Mechanical Engineers, 2008: 923-930.

[Document 7]: Berkelman P, Dzadovsky M. Magnetic Levitation Over LargeTranslation and Rotation Ranges in All Directions [J]. IEEE/ASME Transactionson Mechatronics, 2013, 18(1): 44-52.

[Document 8]: Tong Q, Yuan Z, Liao X, et al. Magnetic Levitation HapticAugmentation for Virtual Tissue Stiffness Perception [J]. IEEE Transactions on Visualization and Computer Graphics, 2018, 24(12): 3123-3136.

SUMMARY OF THE INVENTION

Aiming at the problem that the force feedback module in the existing virtual interventional operation system is difficult to produce an immersive and realistic operation sense, the present invention designs a set of electromagnetic coil arrays including a highly symmetrical electromagnetic coil array and multiple A device with a degree of freedom surgical joystick to reproduce the key force feedback in interventional procedures - circumferential rotational force feedback. In addition, in view of the problem that the electromagnetic force haptic module is difficult to dynamically generate real-time accurate force feedback, an electromagnetic force feedback method suitable for this set of devices is proposed, which can be used for simple, efficient, real-time and accurate generation of circumferential rotational force feedback in virtual interventional surgery. .

The technical scheme adopted by the device of the present invention is: an electromagnetic force feedback device for a virtual interventional operation system, which is characterized in that it includes an electromagnetic coil array and a surgical instrument; the electromagnetic coil array includes four identical electromagnetic coils, Four protective bases and chassis slide rails; the four identical electromagnetic coils are placed on the same plane, and the centers of the bottom surfaces that are closer to each other in each electromagnetic coil are connected in turn to form a square, and the centers of the squares are two pairs of each other. is the intersection point of the bottom surface of the diagonal coil; the four identical electromagnetic coils are fixed on the protection base respectively; the four protection bases are all set on the chassis slide rail, which can be The chassis moves on the slide rail; the surgical instrument is a multi-degree-of-freedom surgical operating rod.

Preferably, the angle of the protection base is adjustable. The angle is the angle between the bottom surface of the electromagnetic coil and the horizontal plane, and the angle adjustment range is 0° to 90°, which can increase the secondary development and flexibility of the device.

Preferably, the surgical instrument includes a rigid operating rod, a first permanent magnet, and a second permanent magnet; the first permanent magnet is disposed on the top of the rigid operating rod, and the second permanent magnet is fixedly disposed on the first permanent magnet The upper part of the permanent magnet is in a cross shape with the first permanent magnet. The two cooperate with each other to highly simulate and restore the mechanical properties of the guide wire in the actual interventional operation.

Preferably, the first permanent magnet and the second permanent magnet are both long cylindrical permanent magnets, and the limitation of the manufacturing process of permanent magnets on the market makes the long cylindrical permanent magnets meet the requirement that the magnetization surface is smaller.

Preferably, the rigid operating rod is made of polyethylene, the radius of the bottom surface is 9mm, and the height is 200mm. These parameters can be adjusted according to the actual situation. The operating rod should be as slender as possible and the material with lower density should be selected, so as to be more consistent with the nature of the catheter in the actual interventional operation.

Preferably, the radius of the bottom surface of the first permanent magnet is 8mm, and the height is 15mm. These parameters can be adjusted according to the actual situation. The first permanent magnet should be as light as possible and fit with the tip of the guide wire in the actual interventional operation, so as to maintain the purity of the force feedback.

Preferably, the radius of the bottom surface of the second permanent magnet is 6mm, and the height is 35mm. These parameters can be adjusted according to the actual situation, and the second permanent magnet should be as slender as possible, so as to obtain a stronger circumferential feedback force while keeping the weight light.

Preferably, the material of the electromagnetic coil is Copper, the distance between the pair of electromagnetic coils that are diagonal to each other is 74 mm, and the number of turns of a single electromagnetic coil is 1024. These parameters can be adjusted according to the actual situation. According to Bio-Savart's law, the closer the distance to the current source, the greater the magnetic field strength; the greater the cross-sectional current of the electromagnetic coil, the greater the magnetic field strength. Therefore, the distance between the coil pairs should be as small as possible but slightly larger than the length of the second permanent magnet. The number of turns of the coil is calculated to be around 1000, which will make the cross-sectional current larger, thereby generating a strong electromagnetic force under the joint promotion of the two factors. feedback force.

The technical scheme adopted by the method of the present invention is: an electromagnetic force feedback method for a virtual interventional operation system, which is characterized in that it includes the following steps:

Step 1: Carry out modeling and simulation of materials such as the electromagnetic coil and the second permanent magnet in the electromagnetic force feedback device to obtain an electromagnetic force feedback model;

Determine the basic operating principle of the electromagnetic force feedback device: the electromagnetic coils in the diagonal positions of each other are given excitation currents of the same magnitude and direction;

Determine the optimal current distribution strategy of the electromagnetic force feedback device: For the angle θ in the positive direction of the X-axis in the Cartesian coordinate system formed by the N pole of the second permanent magnet and the center of the bottom surface of the coil which is a diagonal line to each other, a three-dimensional tuple (x, y, z) to determine the optimal distribution strategy of the current in the electromagnetic coil array;

x=[θ/90]mod2

y=|(45·[θ/45])mod2-θmod45|

z=[θ/180]

Among them, in the three-dimensional tuple (x, y, z), x determines the electromagnetic coil pair to which the current is applied, y is the included angle of the equivalent mapping in the minimum symmetric interval, and z is used to indicate the current direction in the coil pair; each θ Corresponds to a unique triple, that is, a unique optimal current distribution strategy;

Step 2: Build a torque-current regression prediction model;

Adjust the angle θ between the N pole of the second permanent magnet and the positive direction of the X-axis of the Cartesian coordinate system in the electromagnetic force feedback model established in step 1 and the electromagnetic coil excitation current value I, and calculate the torque value T of the second permanent magnet rotating around the axis; Thereby, several sets of offline data are obtained to form an offline data set;

The BPNN network and GRNN network for regression prediction are trained using offline data sets, and the neural network with the best performance is used as a sub-model, and a strong torque-current prediction model is constructed by the model fusion method, so that it can be based on the N pole and the second permanent magnet. The included angle θ in the positive direction of the X-axis of the Cartesian coordinate system and the torque value T can quickly and accurately calculate the excitation current value I of the coil array;

Step 3: Generate the current distribution pattern of the electromagnetic coil array;

Calculate the triplet (x, y, z) according to θ, and then input the y and torque value T in the triplet through the torque-current regression prediction model to solve the excitation current value I, and finally form a new triplet (x, I, z), the triplet is used to guide the distribution of the current of each solenoid in the solenoid array, thereby generating the corresponding circumferential rotational force feedback.

Preferably, the specific implementation of step 2 includes the following sub-steps:

Step 2.1: continuously adjust the rotation angle θ of the second permanent magnet and the electromagnetic coil excitation current value I, and solve to obtain the corresponding torque data T to obtain several sets of offline data to form an offline data set;

Step 2.2: Build a BPNN network including an input layer, a hidden layer, and an output layer; wherein, the input value of the network is the torque value T that the second permanent magnet should receive, and the N pole and the rectangular coordinate system of the second permanent magnet on the surgical instrument The included angle θ in the positive direction of the X axis, the output value is the excitation current value I of the electromagnetic coil;

Step 2.3: Build a GRNN network including an input layer, a model layer, a summation layer, and an output layer; wherein, the input value of the network is the torque value T that the second permanent magnet should receive and the N pole of the second permanent magnet on the surgical instrument. The included angle θ in the positive direction of the X-axis of the Cartesian coordinate system, the output value is the excitation current value I of the electromagnetic coil;

Step 2.4: Train BPNN network and GRNN network;

In the offline training process of the BPNN network, after normalizing the offline dataset, the k-fold training method is used to divide the offline dataset into a training set and a validation set and conduct network training; The square error is used as a loss function. When the error of the training network on the validation set is less than the threshold, the training is stopped and the BPNN network is saved as one of the sub-models of the model fusion, which is recorded as the BPNN network prediction sub-model;

In the offline training process of the GRNN network, an appropriate amount of training data is selected according to the running time of the machine and an initial value of the hyperparameter is randomly generated to determine the network structure and parameters of the GRNN, and several pieces of data outside the training set in the data set are selected as Validation set; use the validation set for GRNN network training and dynamically adjust the value of hyperparameters according to the error during the training process; when the mean square error of the training network on the validation set is less than the threshold, save the hyperparameter and training set as Another sub-model of model fusion, denoted as GRNN network prediction sub-model;

Step 2.5: Take another M pieces of new data as prediction sets and input them into two sub-models for current prediction;

Using the extreme value idea in mathematical statistics, the sub-model weights that minimize the variance of the deviation between the predicted value and the actual value are solved, and the torque-current regression prediction model is finally obtained as follows:

k ₁ +k ₂ =1

和

分别是BPNN网络预测子模型和GRNN网络预测子模型所对应的预测值，θ为第二永磁体(203)的N极与直角坐标系中X轴正方向的夹角，t为第二永磁体(203)的力矩数值，k₁和k₂则对应两个子模型预测结果的权值，

为最终强预测模型的预测值。in

and

are the prediction values corresponding to the BPNN network prediction sub-model and the GRNN network prediction sub-model, respectively, θ is the angle between the N pole of the second permanent magnet (203) and the positive direction of the X-axis in the Cartesian coordinate system, and t is the second permanent magnet The moment value of (203), k ₁ and k ₂ correspond to the weights of the prediction results of the two sub-models,

is the predicted value of the final strong prediction model.

The invention extracts the basic operation principle of the device by analyzing the motion principle of the rigid body rotating around the fixed axis, so as to ensure the feasibility of the method; uses the optimal current distribution strategy analyzed and summarized by the finite element technology to ensure the high efficiency and simplicity of the method; The torque-current prediction model based on neural network can quickly and accurately calculate the dynamic excitation current, ensuring that the electromagnetic force feedback method can accurately generate circumferential rotational force feedback in interventional surgery in real time.

Compared with the prior art, the present invention has the following innovations and advantages:

(1) The use of magnetic levitation-based electromagnetic force feedback technology to generate key force feedback in virtual interventional surgery can effectively reduce system energy consumption, and can almost completely eliminate mechanical friction interference and dynamic nonlinear hysteresis during operation, etc. Complete precise motion control.

(2) The electromagnetic force feedback device designed for the virtual interventional operation system highly restores the operation environment and operation mode of the interventional operation. Compared with the existing mechanical force feedback module based on the guide rail and the pulley, the device gives the operator more freedom. The operation space is not limited to sliding back and forth on the guide rail, which increases the operator's immersion and improves the effect of preoperative training.

(3) The electromagnetic force feedback method proposed according to the electromagnetic force feedback device has strong universality. When the parameters of the established electromagnetic force feedback device (the size and material of the electromagnetic coil, the size and position of the long cylindrical permanent magnet of the operating rod) are changed , this electromagnetic force feedback method is still applicable.

(4) Compared with the existing three-dimensional analytical calculation methods, the proposed torque-current regression prediction model is more concise and efficient, saves a lot of intermediate deduction and calculation processes, and directly establishes the relationship between the two key physical quantities of torque and current, so that the generation of Real-time accurate circumferential rotational force feedback is possible.

Description of drawings

1 is a schematic diagram of an electromagnetic coil array according to an embodiment of the present invention;

2 is a cross-sectional view of an electromagnetic coil according to an embodiment of the present invention;

3 is a schematic diagram of a surgical instrument according to an embodiment of the present invention;

Fig. 4 is the working flow chart of the embodiment of the present invention;

FIG. 5 is a schematic diagram of a circumferential rotational force feedback according to an embodiment of the present invention;

6 is an experimental data diagram of the basic operating principle of the verification device according to the embodiment of the present invention;

7 and 8 are experimental data diagrams for determining an optimal current distribution strategy according to an embodiment of the present invention;

FIG. 9 is a schematic structural diagram of a torque-current regression prediction model according to an embodiment of the present invention.

Detailed ways

In order to facilitate the understanding and implementation of the present invention by those of ordinary skill in the art, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention, but not to limit it. this invention.

Please refer to FIGS. 1 , 2 and 3 , an electromagnetic force feedback device for a virtual interventional surgical system provided by the present invention includes an electromagnetic coil array 1 and a surgical instrument 2 ; the electromagnetic coil array includes four identical electromagnetic coils 101 , four protective bases 102, and chassis slide rails 103; four identical electromagnetic coils 101 are placed on the same plane, and the centers of the bottom surfaces that are closer in each electromagnetic coil 101 are connected in turn to form a square, and the center of the square is The two pairs of diagonal coils are the intersection of the bottom surfaces of the circle centers; the four identical electromagnetic coils 101 are respectively fixed on the protective base 102; the four protective bases 102 are all set on the chassis slide rail 103, which can be Move on the chassis slide rail 103; the surgical instrument 2 is a multi-degree-of-freedom surgical operating rod.

The electromagnetic coil array 1 of this embodiment includes four identical electromagnetic coils 101 and their matching adjustable bases 102, and a chassis slide rail 103 for the electromagnetic coils to slide in a diagonal direction. Among them, the electromagnetic coil skeleton, the adjustable coil base, and the chassis with slide rails are all made of aluminum alloy materials. The electromagnetic coil winding material is Copper, the distance between the diagonal coil pairs is 74mm, and the number of turns of a single coil is 1024. The cross-sectional view is shown in Figure 2. The angle of the protection base 102 in this embodiment is adjustable. The angle is the angle between the bottom surface of the electromagnetic coil and the horizontal plane, and the angle adjustment range is 0° to 90°. When the electromagnetic coil array is energized according to a specific principle, the corresponding magnetic field can be generated to interact with the surgical instrument to generate force feedback in virtual interventional surgery.

Please refer to FIG. 3 , the surgical instrument 2 of this embodiment includes a rigid operating rod 201 , a first permanent magnet 202 , and a second permanent magnet 203 ; the first permanent magnet 202 is provided with the top of the upper end of the rigid operating rod 201 , and the second permanent magnet 203 is fixedly provided In the upper part of the first permanent magnet 202 , the first permanent magnet 202 is crossed with the first permanent magnet 202 . Both the first permanent magnet 202 and the second permanent magnet 203 are long cylindrical permanent magnets.

The surgical instrument 2 of this embodiment is integrated into a rigid bamboo dragonfly-shaped operating rod. Although the real guide wire catheter is made of non-rigid materials, relative to the strength of the internal environment such as blood vessels, it can be considered that the guide wire catheter is in the process of operation. The part in contact with the blood vessel will not be significantly deformed, so the material of the rigid operating rod 201 is set to polyethylene, the radius of the bottom surface is 9mm, and the height is 200mm. The material grade of the first permanent magnet 202 at the tip of the operating rod is NdFe35, the radius of the bottom surface is 8mm, and the height is 15mm, which can generate a feedback force in the axial direction of the surgical instrument 2 when it interacts with the magnetic field excited by the single coil or the triple coil; magnetization The second permanent magnet 203 whose surface is two bottom surfaces (the rectangular permanent magnet cannot form two magnetic poles on the narrow surface due to the limitation of the manufacturing process) has a material grade of NdFe35, a bottom surface radius of 6 mm and a height of 35 mm, which can provide surgical instruments 2 The degree of freedom in the circumferential direction interacts with the specific magnetic field excited by the electromagnetic coil array 1 of the present invention to generate a circumferential rotational feedback force.

The workflow of the entire virtual interventional surgery system in this example is shown in Figure 4, where the workflow of the electromagnetic force feedback device is:

(1) The user operates the surgical instrument 2 to move to the position shown in FIG. 5 , that is, the body center of the second permanent magnet 203 on the surgical instrument coincides with the center of the electromagnetic coil array 1 , assuming that the surgical instrument 2 interacts with the virtual scene and deforms , resulting in a counterclockwise rotational feedback force around the Z-axis.

(2) The PC side calculates the magnitude and direction of the rotational moment of the surgical instrument 2 around the Z axis according to the degree of deformation, and the edge-embedded end with binocular vision positioning function obtains the N pole of the second permanent magnet 203 in the surgical instrument at this time. The included angle θ with the positive direction of the X-axis of the Cartesian coordinate system, and this parameter is passed to the PC side.

(3) In the PC terminal, the torque value T and the angle y mapped by the optimal current strategy are used as input to the torque-current regression prediction model constructed in advance, and each electromagnetic coil in the electromagnetic coil array 1 is quickly and accurately obtained. The excitation current of 101 is transmitted to the underlying embedded in the form of triples.

(4) The bottom embedded terminal controls the drive circuit of the electromagnetic coil 101 to generate the corresponding electromagnetic coil excitation current according to the received current data, so that the electromagnetic coil array 1 generates a corresponding magnetic field so that the second permanent magnet 203 on the surgical instrument generates a circumferential direction Rotational feedback force.

In order to meet the requirement of simple, efficient, real-time and accurate generation of circumferential rotational force feedback in virtual interventional surgery, an electromagnetic force feedback method suitable for the device was proposed. The electromagnetic force feedback method guarantees the feasibility, efficiency, simplicity and real-time accuracy of the device from the basic operating principles of the device, the optimal current distribution strategy and the torque-current calculation model. In general, the electromagnetic force feedback method can be used as an algorithm, which can be obtained according to the two input data on the PC side: the torque value T and the angle θ between the N pole of the second permanent magnet 203 and the positive direction of the X-axis of the Cartesian coordinate system. The optimal distribution method of the excitation current in the whole electromagnetic coil array 1, so as to meet the requirement of circumferential rotation force feedback.

Aiming at the proposed electromagnetic force feedback device, the present invention designs a corresponding electromagnetic force feedback generation method, which is used for concise, efficient, real-time and accurate generation of circumferential rotational force feedback in virtual interventional operations, including the following steps:

Step 1: Perform modeling and simulation of the electromagnetic coil 101 and the second permanent magnet 203 in the electromagnetic force feedback device with materials of equal proportions and sizes to obtain an electromagnetic force feedback model;

In this embodiment, first, finite element analysis software is used to model the electromagnetic coil 101 and the second permanent magnet 203 in the electromagnetic force feedback device (the other parts of the device are not magnetically conductive, so they can be ignored), and the electromagnetic coil 101 and the second permanent magnet 203 are established. The geometric model of the permanent magnet 203 is set as Copper and NdFe35, respectively. Then set the air space of the entire model to offset 50% in the x and y directions, and 100% in the z direction, and use the zero tangent vector boundary condition, set the cross section of the electromagnetic coil 101 as the excitation current conduction surface, and set the second permanent The electromagnetic force of the magnet 203 and the moment of its rotation around the positive direction of the Z-axis are used as solution parameters.

Determine the basic operating principle of the electromagnetic force feedback device: the electromagnetic coils 101 in the diagonal positions of each other are given excitation currents with the same magnitude and direction;

When the device operates according to the above principles, it can meet the two conditions for the generation of circumferential rotation feedback force:

(1) The resultant force of the second permanent magnet 203 as a whole by the electromagnetic force is 0, so as to ensure that it does not produce any movement other than the circumferential rotation around the axis, thereby ensuring the purity of the circumferential rotation force and the sense of touch.

(2) The rotational moment of the second permanent magnet 203 around the fixed axis Z-axis is not 0 and the value is relatively considerable, thereby ensuring the effectiveness of the circumferential rotational force tactile sensation.

Determine the optimal current distribution strategy of the electromagnetic force feedback device;

On the basis of determining the basic operating principles of the electromagnetic force feedback device, the current distribution strategy in the electromagnetic coil array 1 is further refined, so that the strategy has both high efficiency and simplicity:

(1) High efficiency: When the excitation current is passed through the same size, the electromagnetic coil pair that generates a larger torque is selected as the field source.

(2) Simplicity: The topology of the electromagnetic coil array 1 is analyzed, and the entire working area of the device is mapped into the minimum symmetry interval by using the characteristics of relative position and absolute position.

Based on the above description of the two characteristics of the optimal current distribution strategy, the following mathematical expression of the optimal current distribution strategy is proposed: For the right angle formed by the N pole of the second permanent magnet 203 and the center of the bottom surface of the coil that are diagonal to each other A three-dimensional tuple (x, y, z) is proposed to determine the optimal distribution strategy of the current in the electromagnetic coil array 1. The calculation method of the three elements is as follows:

x=[θ/90]mod2

y=|(45·[θ/45])mod2-θmod45|

z=[θ/180]

Among them, in this three-dimensional tuple, x determines the electromagnetic coil pair to which the current is applied, y is the included angle of the equivalent mapping in the minimum symmetric interval, and z is used to indicate the current direction in the coil pair. Each θ corresponds to a unique triple, that is, a unique optimal current distribution strategy.

In this embodiment, the correctness of the basic operating principle is further verified;

Using the electromagnetic force feedback model established in step 1, and according to the basic operating principle of the electromagnetic force feedback device, the excitation currents with the same magnitude and direction are distributed to the electromagnetic coils 101 that are diagonal to each other, and the second permanent magnet 203 is analyzed after solving. The value of the resultant force of the electromagnetic force and the value of the torque around the axis of rotation are used to verify the correctness of the basic operating principle.

In this embodiment, when the included angle between the N pole of the second permanent magnet 203 on the surgical instrument 2 and the positive direction of the X-axis of the Cartesian coordinate system is 0, according to the basic operating principle of the device, the coils 1 and 3 are centered respectively leading to the starting point. The value is 0, the cut-off value is 2048A, and the step size is 204.8A of excitation current, and the results in Figure 6 are obtained after solving using the finite element calculation software. It can be seen that the result obtained by the simulation satisfies the two conditions for the rigid body to rotate around the fixed axis within the error range: the resultant force of the electromagnetic force on the second permanent magnet 203 is 0, and the value of its rotational torque around the axis is considerable. Then the feasibility of the basic operating principle is verified.

In this embodiment, the optimal current distribution strategy is further determined;

Continue to use the electromagnetic force feedback model established in step 1, where the angular range for each pair of coils to satisfy the high efficiency will be determined and the simplicity of the optimal current distribution strategy will be verified.

Determination of high efficiency: In all ranges of any minimum symmetry interval, two pairs of electromagnetic coils (1, 3 and 2, 4 coil pairs) that are diagonal to each other are respectively led to the excitation current of the same size, and the simulation solution is compared. When different coil pairs work, the torque value of the second permanent magnet 203 around the rotating shaft determines the working angle range of the two pairs of coils within the minimum symmetry interval, so as to ensure the high efficiency of the strategy.

Verification of simplicity: when the second permanent magnet 203 rotates to the θ position, solve the torque value T generated according to the basic operating principle of the electromagnetic force feedback device and the torque value generated under the guidance of the triplet determined by the optimal current distribution strategy. T, compare the two data to verify the simplicity of the strategy.

In this embodiment, the angle range that satisfies the high efficiency of the optimal current distribution strategy is first determined. According to the linear relationship between the current and the electromagnetic field generated by it, it is only necessary to solve the effect of the excitation current on the torque in the two coil pairs under a certain value of current. contribution size. When the two sets of coil pairs are connected to the excitation current of 2048A according to the basic operating principle of the device, the simulation solution is carried out, and the result is shown in Figure 7. It can be seen that in this embodiment, when θ is in the range of 0°-45°, the torque values generated by the excitation currents in the coil pairs 1 and 3 far from the second permanent magnet 203 are higher than the torque values generated by the excitation currents in the coil pairs 2 and 4. big. Therefore, in the subsequent current setting, the coil pair farther away from the second permanent magnet 203 is selected as the field source for current distribution.

In the process of simplicity verification, according to θ and the triplet (x, y, z) calculated from it, use the model established in step 1 to solve the problem according to the basic operating principle of the device before the mapping and the basic operating principle of the device after the mapping. The torque value T obtained by the optimal current distribution strategy is compared, and the specific results are shown in Figure 8. It can be seen from the data in the figure that θ takes any value from 0° to 360°, and the triplet (x, y, z) is obtained through the optimal current distribution strategy, and the torque value T before mapping is driven by the excitation current of random size. _θ is basically the same as the torque value T _y obtained by setting the working coil pair and the current direction according to the triple (x, y, z) after mapping, indicating that the second permanent magnet 203 uses the symmetry of the coil array in the entire working range. It is possible to map torque, current and other numerical values within the minimum symmetry interval of 0°-45°. Therefore, the simplicity of the optimal current distribution strategy is verified.

Using the finite element software in step 1 for calculation and verification takes a long time and cannot be calculated online. Therefore, it is necessary to build a small-scale calculation model for relatively real-time and accurate online calculation. Therefore, the regression prediction model based on neural network is selected for prediction.

Step 2: Build a torque-current regression prediction model;

Adjust the angle θ between the N pole of the second permanent magnet 203 and the positive direction of the X-axis of the rectangular coordinate system in the electromagnetic force feedback model established in step 1 and the value I of the excitation current of the electromagnetic coil, and calculate the torque value of the second permanent magnet 203 rotating around the axis T; thereby obtaining several sets of offline data to form an offline data set;

The BPNN network and GRNN network for regression prediction are trained using offline data sets, and the neural network with the best performance is used as a sub-model, and a strong torque-current prediction model is constructed by the model fusion method, so that the N pole of the second permanent magnet 203 can be calculated according to the N pole of the second permanent magnet 203. Quickly and accurately calculate the excitation current value I of the coil array with the angle θ and the torque value T in the positive direction of the rectangular coordinate system;

Includes the following sub-steps:

Step 2.1: Continuously adjust the rotation angle θ of the second permanent magnet 203 and the excitation current value I of the electromagnetic coil 101, and solve to obtain the corresponding torque data T to obtain several sets of offline data to form an offline data set; the data in the offline data set of this embodiment The number of bars is 966.

Step 2.2: Build a BPNN network including an input layer, two hidden layers, and an output layer. Among them, the input value of the network is the torque value T that the second permanent magnet 203 should receive and the angle θ between the N pole of the second permanent magnet 203 on the surgical instrument 2 and the positive direction of the X-axis of the rectangular coordinate system, the nodes before and after the two hidden layers The numbers are 7 and 8 respectively, the activation functions are Sigmoid and Relu, respectively, and the output value is the excitation current value I of the electromagnetic coil.

Step 2.3: Build a GRNN network including an input layer, a pattern layer, a summation layer, and an output layer. The input and output parameters are the same as those in step 2.2. After a lot of pre-experiments, the number of nodes in the mode layer is set to 322, and the number of nodes in the summation layer is the output dimension + 1, which is 2.

Step 2.4: Train BPNN network and GRNN network;

In the offline training process of the BPNN network, after normalizing the offline dataset, the k-fold training method is used to divide the offline dataset into a training set and a validation set and conduct network training; The square error is used as a loss function. When the error of the training network on the validation set is less than the threshold, the training is stopped and the BPNN network is saved as one of the sub-models of the model fusion, which is recorded as the BPNN network prediction sub-model;

In this embodiment, k is taken as 4. Then the mean square error of the excitation current of the electromagnetic coil is used as the loss function. When the training error of the model on the validation set is less than 42A, the training is stopped and the model is saved.

In the offline training process of the GRNN network, an appropriate amount of training data is selected according to the running time of the machine and an initial value of the hyperparameter is randomly generated to determine the network structure and parameters of the GRNN, and several pieces of data outside the training set in the data set are selected as Validation set; use the validation set for GRNN network training and dynamically adjust the value of hyperparameters according to the error during the training process; when the mean square error of the training network on the validation set is less than the threshold, save the hyperparameter and training set as Another sub-model of model fusion, denoted as GRNN network prediction sub-model;

Because of the particularity of the GRNN network, the network does not need to train the weights between nodes. In other words, when the training set and hyperparameter values are determined, the entire network structure is fixed. Every time a prediction is made, the test data needs to be operated on with the data in the training set. In this embodiment, 322 pieces of data are evenly taken from the data set as the training set of the GRNN network, and another 100 pieces of data in the data set are taken as the verification set, and the values of the hyperparameters are continuously adjusted so that the mean square error is less than 42A. After experiments, the hyperparameter δ is set to 0.5.

Step 2.5: Take another 15 new pieces of data as prediction sets and input them into the two sub-models for current prediction. Using the extreme value idea in mathematical statistics, the sub-model weights that minimize the variance of the deviation between the predicted value and the actual value are solved. The final torque-current regression prediction model is as follows:

和

分别是BPNN网络预测子模型和GRNN网络预测子模型所对应的预测值，θ为第二永磁体203的N极与直角坐标系中X轴正方向的夹角，t为第二永磁体(203)的力矩数值，

为最终强预测模型的预测值，整个强力矩-电流回归预测模型结构如图9所示。in

and

are respectively the prediction values corresponding to the BPNN network prediction sub-model and the GRNN network prediction sub-model, θ is the angle between the N pole of the second permanent magnet 203 and the positive direction of the X-axis in the Cartesian coordinate system, and t is the second permanent magnet (203 ) torque value,

For the predicted value of the final strong prediction model, the structure of the entire strong torque-current regression prediction model is shown in Figure 9.

Step 3: Generate the current distribution pattern of the electromagnetic coil array;

In this electromagnetic force feedback method, after obtaining the position parameter θ and torque parameter T of the surgical instrument 2, the triplet (x, y, z) is calculated by substituting the optimal current distribution strategy into θ, and then the model is predicted by torque-current regression. Enter the y and torque values T in the triplet to solve the excitation current value I, and finally form a new triplet (x,I,z), and transfer the triplet to the underlying embedded processor to guide the electromagnetic coil array The distribution of the current of each electromagnetic coil 101 in 1, thereby generating the corresponding circumferential rotational force feedback.

Starting from the real-time and accuracy of the key force feedback in the virtual interventional surgery system, this embodiment realizes a set of electromagnetic coil array topology and surgical instruments according to the structural characteristics of the electromagnetic force feedback device, and proposes a corresponding device based on the device. The simple, efficient, real-time and accurate electromagnetic force feedback method is verified by AnsoftMaxwell simulation software. Then, a large amount of simulation data is obtained offline through the finite element calculation method to train multiple regression prediction models, and finally the model fusion method is used to build a strong torque-current prediction model to quickly and accurately predict the current. The experimental results of this embodiment show that the prediction model can predict the current with a 3% error and a frequency higher than 40 Hz, and combined with the electromagnetic force feedback method, the key force feedback in the interventional operation can be restored in real time and accurately.

The electromagnetic force feedback method and device for the virtual interventional operation system designed by the present invention can not only generate the circumferential rotation feedback force in the interventional operation accurately in real time, but also avoid the shortcomings in the guide rail-pulley type force feedback module, such as mechanical friction caused by errors, excessive equipment scale, and high energy consumption. In particular, the present invention has achieved a great improvement in the sense of immersion in operation. Both in the operation space and the operation method, the procedure of the interventional operation is basically and completely reproduced, and combined with the visual effect of the virtual scene and the tactile sense of electromagnetic force feedback. The preoperative training effect of the virtual interventional surgery system is greatly improved.

It is worth mentioning that a relatively novel calculation method of physical quantities is proposed in the electromagnetic force feedback method of the present invention, which is different from the previous calculation method based on three-dimensional analysis. This method omits the intermediate complex derivation steps. The neural network performs model fusion, which directly establishes the connection between the key physical quantity torque and current, and also has great credibility in the calculation accuracy, which greatly reduces the calculation time and memory consumption.

It should be understood that the parts not described in detail in this specification belong to the prior art; the above description of the preferred embodiments is relatively detailed, and therefore should not be considered as a limitation on the protection scope of the patent of the present invention. Under the inspiration of the present invention, without departing from the scope of protection of the claims of the present invention, substitutions or modifications can also be made, which all fall within the scope of protection of the present invention. Requirements shall prevail.

Claims (7)

1. An electromagnetic force feedback method facing a virtual interventional operation system adopts an electromagnetic force feedback device facing the virtual interventional operation system;

the method is characterized in that:

the device comprises an electromagnetic coil array (1) and a surgical instrument (2);

the electromagnetic coil array comprises four identical electromagnetic coils (101), four protection bases (102) and a chassis sliding rail (103); the four identical electromagnetic coils (101) are arranged on the same plane, the circle centers of the bottom surfaces of the electromagnetic coils (101) which are closer to each other are sequentially connected to form a square, and the center of the square is the intersection point of the two pairs of bottom surface circle center connecting lines of the diagonal coils; the four identical electromagnetic coils (101) are respectively and fixedly arranged on the protection base (102); the four protection bases (102) are all arranged on the chassis slide rail (103) and can move on the chassis slide rail (103);

the surgical instrument (2) is a surgical operating rod with multiple degrees of freedom;

the surgical instrument (2) comprises a rigid operating rod (201), a first permanent magnet (202) and a second permanent magnet (203);

the first permanent magnet (202) is arranged at the top of the upper end of the rigid operating rod (201), and the second permanent magnet (203) is fixedly arranged at the middle upper part of the first permanent magnet (202) and is crossed with the first permanent magnet (202);

the method comprises the following steps:

step 1: modeling simulation of materials with equal proportion and the like is carried out on an electromagnetic coil (101) and a second permanent magnet (203) in the electromagnetic force feedback device to obtain an electromagnetic force feedback model;

determining the basic operation principle of the electromagnetic force feedback device: the electromagnetic coils (101) which are arranged at the diagonal positions give excitation currents with the same magnitude and direction;

determining an optimal current distribution strategy of the electromagnetic force feedback device: a three-dimensional tuple (X, y, z) is provided for an included angle theta in the positive direction of the X axis in a rectangular coordinate system formed by connecting the N pole of the second permanent magnet (203) and the centers of the bottom surfaces of the coils which are diagonal to each other, so as to determine the optimal distribution strategy of the current in the electromagnetic coil array (1);

x＝[θ/90]mod2

y＝|(45·[θ/45])mod2-θmod45|

z＝[θ/180]

in the three-dimensional tuple (x, y, z), x determines a solenoid coil pair applying current, y is an included angle of equivalent mapping in a minimum symmetric interval, and z is used for indicating the current direction in the coil pair; each theta corresponds to a unique triplet, namely a unique optimal current distribution strategy;

and 2, step: constructing a torque-current regression prediction model;

adjusting the positive direction included angle theta of the N pole of the second permanent magnet (203) in the electromagnetic force feedback model established in the step 1 and the X axis of the rectangular coordinate system and the exciting current value I of the electromagnetic coil, and calculating the torque value T of the second permanent magnet (203) rotating around the axis; thereby obtaining a plurality of groups of offline data to form an offline data set;

training a BPNN network and a GRNN network for regression prediction by using an offline data set, constructing a strong torque-current prediction model by using a neural network with optimal performance as a sub-model through a model fusion method, and thus quickly and accurately calculating an excitation current value I of a coil array according to an included angle theta between the N pole of a second permanent magnet (203) and the positive direction of the X axis of a rectangular coordinate system and a torque value T;

the specific implementation of the step 2 comprises the following sub-steps:

step 2.1: continuously adjusting the rotation angle theta of the second permanent magnet (203) and the excitation current value I of the electromagnetic coil (101), and solving to obtain corresponding torque data T so as to obtain a plurality of groups of offline data to form an offline data set;

step 2.2: constructing a BPNN network comprising an input layer, a hidden layer and an output layer; the input value of the network is a torque value T to which the second permanent magnet (203) is supposed to be subjected and an included angle theta between the N pole of the second permanent magnet (203) on the surgical instrument (2) and the positive direction of the X axis of the rectangular coordinate system, and the output value is an electromagnetic coil excitation current value I;

step 2.3: building a GRNN network comprising an input layer, a mode layer, a summation layer and an output layer; the input value of the network is a torque value T to which the second permanent magnet (203) is supposed to be subjected and an included angle theta between the N pole of the second permanent magnet (203) on the surgical instrument (2) and the positive direction of the X axis of the rectangular coordinate system, and the output value is an electromagnetic coil excitation current value I;

step 2.4: training a BPNN network and a GRNN network;

in the off-line training process of the BPNN, after off-line data sets are subjected to normalization processing, dividing the off-line data sets into a training set and a verification set by using a k-fold training method, and performing network training; then, taking the mean square error of the exciting current of the electromagnetic coil as a loss function, stopping training and storing the BPNN as one of the model fusion submodels when the error of the network on the verification set is smaller than a threshold value after training, and marking the model fusion submodel as a BPNN prediction submodel;

in the off-line training process of the GRNN, selecting a proper amount of training data according to the running time of a machine and randomly generating a super-parameter initial value so as to determine the network structure and parameters of the GRNN, and simultaneously selecting a plurality of pieces of data out of a training set in a data set as a verification set; carrying out GRNN network training by using a verification set and dynamically adjusting the value of the hyper-parameter according to an error in the training process; when the mean square error of the network on the verification set is smaller than a threshold value after training, the hyper-parameters and the training set are stored and used as a sub-model for fusing another model, and the sub-model is marked as a GRNN network prediction sub-model;

step 2.5: taking M brand new data as a prediction set and respectively inputting the data into the two submodels for current prediction;

by using an extreme value thought in mathematical statistics, solving a sub-model weight value which enables the variance of the deviation between the predicted value and the actual value to be minimum, and finally obtaining a moment-current regression prediction model as follows:

k ₁ +k ₂ ＝1

wherein

And

the BPNN network predictor model and the GRNN network predictor model correspond toTheta is an included angle between the N pole of the second permanent magnet (203) and the positive direction of the X axis in the rectangular coordinate system, t is a moment value of the second permanent magnet (203), and k is ₁ And k ₂ The weight values of the results are predicted corresponding to the two submodels,

the predicted value of the final strong prediction model is obtained;

and step 3: generating a current distribution pattern for the electromagnetic coil array;

and calculating a triad (x, y, z) according to theta, inputting the y and moment values T in the triad through a moment-current regression prediction model to solve an excitation current value I, finally forming a new triad (x, I, z), and guiding the current distribution of each electromagnetic coil (101) in the electromagnetic coil array (1) by using the triad so as to generate corresponding circumferential rotating force feedback.

2. The electromagnetic force feedback method for the virtual interventional surgical system as set forth in claim 1, wherein: the angle of the protection base (102) is adjustable, wherein the angle is an included angle between the bottom surface of the electromagnetic coil and the horizontal plane, and the angle adjusting range is 0-90 degrees.

3. The electromagnetic force feedback method for the virtual interventional surgical system as set forth in claim 1, wherein: the first permanent magnet (202) and the second permanent magnet (203) are both long cylindrical permanent magnets.

4. The electromagnetic force feedback method for virtual interventional surgery systems as defined in claim 1, characterized in that: the rigid operating rod (201) is made of polyethylene, the radius of the bottom surface is 9mm, and the height is 200 mm.

5. The electromagnetic force feedback method for the virtual interventional surgical system as set forth in claim 1, wherein: the radius of the bottom surface of the first permanent magnet (202) is 8mm, and the height of the first permanent magnet is 15 mm.

6. The electromagnetic force feedback method for virtual interventional surgery systems as defined in claim 1, characterized in that: the radius of the bottom surface of the second permanent magnet (203) is 6mm, and the height of the second permanent magnet is 35 mm.

7. The electromagnetic force feedback method for virtual interventional surgical systems as defined in any one of claims 1 to 6, wherein: the material of the electromagnetic coil (101) is Copper, the distance between the pair of the electromagnetic coils (101) which are diagonal to each other is 74mm, and the number of coil turns of a single electromagnetic coil (101) is 1024.

Images