Method for predicting and compensating positioning error of navigation system in electromagnetic interference environment

Abstract

The invention discloses a navigation system positioning error prediction and compensation method in an electromagnetic interference environment, and relates to the technical field of medical navigation and positioning. Comprising the following specific steps: acquiring reference data and positioning data of the electromagnetic navigation system in an electromagnetic interference environment to make an error data set; making a classification training set by using the error data set, and constructing and training a multi-layer perceptron classification model based on the classification training set; making a prediction training set by using the error data set, and constructing and training a multi-layer perceptron prediction model based on the prediction training set; acquiring positioning data of the electromagnetic navigation system during operation; in the operation, the positioning error of the electromagnetic navigation system during interference is predicted based on the multi-layer perceptron classification model and the multi-layer perceptron prediction model, and the positioning data of the electromagnetic navigation system is compensated based on the predicted positioning error. The error compensation method provided by the invention has high adaptability and strong robustness, and can effectively reduce the influence of intraoperative electromagnetic interference on the positioning precision of the electromagnetic navigation system.

Description

Technical Field

[0001] The present invention relates to the field of medical navigation and positioning technology, and specifically to a positioning error compensation method for an electromagnetic navigation system based on a multi-layer perceptron model, which is used to improve the positioning accuracy of the electromagnetic navigation system in a complex interference environment, and is particularly suitable for high-precision positioning and tracking of flexible surgical instruments during surgery. Background Art

[0002] The electromagnetic navigation system is a commonly used medical navigation technology that tracks the position and posture of surgical instruments by arranging magnetic field transmitters and sensors in space. It is widely used in minimally invasive surgery, endoscopic navigation, tumor ablation and other fields. However, the positioning accuracy of the electromagnetic navigation system is easily affected by metal objects and electromagnetic interference in the surrounding environment. This error may cause inaccurate positioning of surgical instruments, thus affecting the safety and effectiveness of the operation.

[0003] At present, the research on the positioning error of electromagnetic navigation system mainly focuses on physical shielding, signal filtering and algorithm optimization. Among them, the physical shielding method improves the positioning accuracy of the system by reducing external interference, but it is expensive and difficult to apply to complex surgical environments; the signal filtering method can eliminate noise to a certain extent, but its adaptability to dynamic interference is limited; and the error compensation method based on the algorithm corrects the positioning error by constructing a mathematical model, which is flexible and widely applicable, but the accuracy and real-time performance of the traditional algorithm in complex interference environments are still insufficient.

[0004] In recent years, with the rapid development of machine learning technology, the multilayer perceptron neural network model, as a typical artificial neural network, has been widely used in nonlinear data modeling and data prediction. Using the multilayer perceptron model to predict and compensate for the positioning error of the electromagnetic navigation system can not only overcome the influence of complex interference, but also achieve real-time correction while ensuring high accuracy, thus providing an efficient solution for the field of medical navigation. However, the technology is not yet fully mature, and further optimization is still needed in model training, interference data modeling, and system integration. Summary of the invention

[0005] In view of the problem that electromagnetic navigation systems may generate errors due to electromagnetic interference, the present invention provides a method for predicting positioning errors caused by electromagnetic interference in electromagnetic navigation systems based on a multi-layer perceptron neural network and compensating for positioning data of the electromagnetic navigation system based on the predicted positioning errors.

[0006] The above-mentioned purpose of the present invention is mainly achieved through the following scheme:

[0007] A method for predicting and compensating positioning errors of a navigation system in an electromagnetic interference environment, wherein the specific implementation process of the method comprises the following steps:

[0008] 1) Collect reference data and positioning data of the electromagnetic navigation system under electromagnetic interference environment to produce error data set;

[0009] 2) Use the error data set to create a classification training set, and build and train a multi-layer perceptron classification model based on the classification training set;

[0010] 3) Use the error data set to create a prediction training set, and build and train a multi-layer perceptron prediction model based on the prediction training set;

[0011] 4) Obtaining positioning data of the electromagnetic navigation system during surgery;

[0012] 5) Predict the positioning error of the electromagnetic navigation system during interference based on the multi-layer perceptron classification model and the multi-layer perceptron prediction model, and compensate the positioning data of the electromagnetic navigation system based on the predicted positioning error;

[0013] In the step 1), three positioning coils are used to prepare the error data set, wherein the positioning coil a and the positioning coil b are fixedly placed at different positions according to different surgical areas, and the positioning coil c is used to track the surgical instrument. When the position offset of the positioning coil a or the positioning coil b is greater than its own repeatable positioning accuracy, it is determined that electromagnetic interference occurs. When electromagnetic interference occurs, the data recorded in the error data set are the interference source code causing the electromagnetic interference, the reference data P of the positioning coil a, and the reference data P of the positioning coil a. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ), reference data P of positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) and the reference data P of the positioning coil c c (X c ,Y c ,Z c ) and positioning data P c ′(X′ c ,Y c ′,Z′ c );

[0014] The data acquisition range for making the error data set is the surgical area where the surgical instruments are more active in different surgeries. The positioning coil a and the positioning coil b can obtain reference data and positioning data from their own data at the same time. The reference data of the positioning coil c can be provided by the optical positioning system after the coordinate system conversion.

[0015] In the step 2), the interference source code causing electromagnetic interference in the error data set and the reference data P of the positioning coil a are used. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ) and the reference data P of the positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) to create a classification training set, the feature data of the classification training set is the reference data P of the positioning coil a a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ) and the reference data P of the positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ), the label data of the classification training set is the interference source code that causes electromagnetic interference;

[0016] The multi-layer perceptron classification model built and trained based on the classification training set includes two hidden layers consisting of 64 neurons. The activation functions of the first hidden layer and the second hidden layer are:

[0017] f(x)=max(0,x) and

[0018] The multi-layer perceptron classification model uses Adam as the optimizer, and its loss function is

[0019] In the step 3), the interference source code causing electromagnetic interference in the error data set and the reference data P of the positioning coil a are used. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ), reference data P of positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) and the reference data P of the positioning coil c c (X c ,Y c ,Z c ) and positioning data P c ′(X′ c ,Y c ′,Z′ c ) to make a prediction training set, the characteristic data of the prediction training set is the interference source code causing electromagnetic interference, the positioning data P of the positioning coil a a ′(X′ a ,Y a ′,Z′ a ), positioning data P of positioning coil b b ′(X′ b ,Y b ′,Z′ b ), positioning data P of positioning coil c c ′(X′ c ,Y c ′,Z′ c ), positioning error e of positioning coil a a (x a ,y a ,z a ) and the positioning error e of the positioning coil b b (x b ,y b ,z b ), the label data of the predicted training set is the positioning error e of the positioning coil c c (x c ,y c ,z c ), the positioning error calculation method is:

[0020]

[0021] The multi-layer perceptron prediction model built and trained based on the prediction training set includes three hidden layers consisting of 128 neurons. The activation functions of the first hidden layer and the second hidden layer are f(x)=max(0,x), and the activation function of the third hidden layer is f(x)=max(ax,x);

[0022] The multi-layer perceptron prediction model uses Adam as the optimizer, and its loss function is

[0023] In step 4), the intraoperative positioning coil a and the positioning coil b are fixedly placed at appropriate positions according to different surgical scopes, and the data of each positioning coil obtained during the operation is the reference data P of the positioning coil a. a (X a ,Y a ,Z a ), positioning data P of positioning coil a a ′(X′ a ,Y a ′,Z′ a ), positioning error e of positioning coil a a (x a ,y a ,z a ), reference data P of positioning coil b b (X b ,Y b ,Z b ), positioning data P of positioning coil b b ′(X′ b ,Y b ′,Z′ b ), positioning error e of positioning coil b b (x b ,y b ,z b ) and the positioning data P of the positioning coil c c ′(X′ c ,Y c ′,Z′ c );

[0024] In step 5), the multi-layer perceptron classification model is based on the reference data P of the intraoperative positioning coil a. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ), reference data P of positioning coil b b (X b ,Yb ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) to identify the interference source type and pass the interference source code to the multi-layer perceptron prediction model, which is based on the positioning data P of the intraoperative positioning coil a a ′(X′ a ,Y a ′,Z′ a ), positioning data P of positioning coil b b ′(X′ b ,Y b ′,Z′ b ), positioning data P of positioning coil c c ′(X′ c ,Y c ′,Z′ c ), positioning error e of positioning coil a a (x a ,y a ,z a ), positioning error e of positioning coil b b (x b ,y b ,z b ) and the interference source code to predict the positioning error of the positioning coil c and obtain the predicted positioning error Finally, the predicted positioning error based on the positioning coil c The positioning data of the positioning coil c is compensated by:

[0025]

[0026] In the formula This is the positioning data after compensation of the positioning coil c. Before compensation, the current interference level is judged to determine whether compensation is required. The judgment criteria are:

[0027] or

[0028] Where r represents the maximum repeatable positioning accuracy of the electromagnetic navigation system.

[0029] The beneficial effects of the present invention are: a method for predicting and compensating positioning errors of a navigation system in an electromagnetic interference environment of the present invention ensures the real-time performance of error compensation, has high adaptability and strong robustness for error compensation, can be integrated into the control software of the navigation system, and can also run independently outside the control software of the navigation system, can effectively solve the influence of electromagnetic interference during surgery on the navigation positioning accuracy, is suitable for traditional surgical instruments, flexible surgical instruments and other electromagnetic navigation scenarios, has broad application prospects and practical value, and in addition, the present invention provides a new solution to the problem of error compensation, and has a positive role in promoting the further development of navigation technology in the fields of medicine, industry, aviation, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] For ease of explanation, the present invention is described in detail by the following specific embodiments and accompanying drawings;

[0031] Figure 1 It is an overall flow chart of the positioning error prediction and compensation method of the present invention;

[0032] Figure 2 It is a structural diagram of the multi-layer perceptron classification model of the present invention;

[0033] Figure 3 It is a structural diagram of the multi-layer perceptron prediction model of the present invention;

[0034] Figure 4 It is a schematic diagram of sensor placement when making an error data set according to the present invention;

[0035] Figure 5 This is a schematic diagram of the placement of sensors during use during surgery according to the present invention. DETAILED DESCRIPTION

[0036] The embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0037] like Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 As shown, this specific implementation adopts the following technical solution: a method for predicting and compensating positioning errors of a navigation system in an electromagnetic interference environment, characterized in that the specific implementation process of the method includes the following steps:

[0038] 1) Collect reference data and positioning data of the electromagnetic navigation system under electromagnetic interference environment to produce error data set;

[0039] 2) Use the error data set to create a classification training set, and build and train a multi-layer perceptron classification model based on the classification training set;

[0040] 3) Use the error data set to create a prediction training set, and build and train a multi-layer perceptron prediction model based on the prediction training set;

[0041] 4) Obtaining positioning data of the electromagnetic navigation system during surgery;

[0042] 5) Predict the positioning error of the electromagnetic navigation system during interference based on the multi-layer perceptron classification model and the multi-layer perceptron prediction model, and compensate the positioning data of the electromagnetic navigation system based on the predicted positioning error;

[0043] In the step 1), three positioning coils are used to prepare the error data set, wherein the positioning coil a and the positioning coil b are fixedly placed at different positions according to different surgical areas, and the positioning coil c is used to track the surgical instrument. When the position offset of the positioning coil a or the positioning coil b is greater than its own repeatable positioning accuracy, it is determined that electromagnetic interference occurs. When electromagnetic interference occurs, the data recorded in the error data set are the interference source code causing the electromagnetic interference, the reference data P of the positioning coil a, and the reference data P of the positioning coil a. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ), reference data P of positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) and the reference data P of the positioning coil c c (X c ,Y c ,Z c ) and positioning data P c ′(X′ c ,Y c ′,Z′ c );

[0044] The data acquisition range for making the error data set is the surgical area where surgical instruments move more frequently in different surgeries. The positioning coils a and b should be placed close to the surgical area. Since the positioning coils a and b are fixedly placed, they can simultaneously obtain reference data and positioning data by themselves. The reference data is the undisturbed positioning data. The reference data of the positioning coil c can be provided by the optical positioning system after the coordinate system conversion.

[0045] In the step 2), the interference source code causing electromagnetic interference in the error data set and the reference data P of the positioning coil a are used. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ) and the reference data P of the positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) to create a classification training set, the feature data of the classification training set is the reference data P of the positioning coil a a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ) and the reference data P of the positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ), the label data of the classification training set is the interference source code that causes electromagnetic interference;

[0046] The multi-layer perceptron classification model built and trained based on the classification training set takes as input the reference data P of the positioning coil a. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ), reference data P of positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b), whose output is the interference source code that causes electromagnetic interference. The hidden layer of the multi-layer perceptron classification model consists of two hidden layers consisting of 64 neurons. The activation functions of the first hidden layer and the second hidden layer are:

[0047] f(x)=max(0,x) and

[0048] When training the multi-layer perceptron classification model, Adam is used as the optimizer to optimize it, and its loss function is

[0049] In the step 3), the interference source code causing electromagnetic interference in the error data set and the reference data P of the positioning coil a are used. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ), reference data P of positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) and the reference data P of the positioning coil c c (X c ,Y c ,Z c ) and positioning data P c ′(X′ c ,Y c ′,Z′ c ) to make a prediction training set, the characteristic data of the prediction training set is the interference source code causing electromagnetic interference, the positioning data P of the positioning coil a a ′(X′ a ,Y a ′,Z′ a ), positioning data P of positioning coil b b ′(X′ b ,Y b ′,Z′ b ), positioning data P of positioning coil c c ′(X′ c ,Y c ′,Z′ c ), positioning error e of positioning coil a a (x a ,y a ,z a) and the positioning error e of the positioning coil b b (x b ,y b ,z b ), the label data of the predicted training set is the positioning error e of the positioning coil c c (x c ,y c ,z c ), the positioning error calculation method is:

[0050]

[0051] The multi-layer perceptron prediction model built and trained based on the prediction training set takes as input the interference source code causing electromagnetic interference and the positioning data P of the positioning coil a. a ′(X′ a ,Y a ′,Z′ a ), positioning data P of positioning coil b b ′(X′ b ,Y b ′,Z′ b ), positioning data P of positioning coil c c ′(X′ c ,Y c ′,Z′ c ), positioning error e of positioning coil a a (x a ,y a ,z a ), positioning error e of positioning coil b b (x b ,y b ,z b ), whose output is the predicted positioning error of the positioning coil c The hidden layer of the multilayer perceptron prediction model consists of three hidden layers of 128 neurons. The activation function of the first and second hidden layers is f(x) = max(0, x), and the activation function of the third hidden layer is f(x) = max(ax, x);

[0052] When training the multi-layer perceptron classification model, Adam is used as the optimizer to optimize it, and its loss function is

[0053] In step 4), the intraoperative positioning coil a and the positioning coil b are fixedly placed at appropriate positions according to different surgical scopes, and the data of each positioning coil obtained during the operation is the reference data P of the positioning coil a. a (X a ,Y a ,Z a ), positioning data P of positioning coil aa ′(X′ a ,Y a ′,Z′ a ), positioning error e of positioning coil a a (x a ,y a ,z a ), reference data P of positioning coil b b (X b ,Y b ,Z b ), positioning data P of positioning coil b b ′(X′ b ,Y b ′,Z′ b ), positioning error e of positioning coil b b (x b ,y b ,z b ) and the positioning data P of the positioning coil c c ′(X′ c ,Y c ′,Z′ c );

[0054] In step 5), the multi-layer perceptron classification model is based on the reference data P of the intraoperative positioning coil a. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ), reference data P of positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) to identify the interference source type and pass the interference source code to the multi-layer perceptron prediction model, which is based on the positioning data P of the intraoperative positioning coil a a ′(X′ a ,Y a ′,Z′ a ), positioning data P of positioning coil b b ′(X′ b ,Y b ′,Z′ b ), positioning data P of positioning coil c c ′(X′ c ,Y c′,Z′ c ), positioning error e of positioning coil a a (x a ,y a ,z a ), positioning error e of positioning coil b b (x b ,y b ,z b ) and the interference source code to predict the positioning error of the positioning coil c and obtain the predicted positioning error Finally, the predicted positioning error based on the positioning coil c The positioning data of the positioning coil c is compensated by:

[0055]

[0056] In the formula For the positioning data after compensation of the positioning coil c, the current interference degree is judged before compensation to determine whether compensation is needed, so as to avoid the situation where the predicted positioning error is greater than the actual positioning error. The judgment criteria are:

[0057] or

[0058] Where r represents the maximum repeatable positioning accuracy of the electromagnetic navigation system.

[0059] According to an embodiment of the present invention, a method for predicting and compensating the positioning error of a navigation system in an electromagnetic interference environment is disclosed. A multi-layer perceptron neural network is used to predict the positioning error of an electromagnetic navigation system and compensate for the positioning data based on the predicted positioning error. Three electromagnetic positioning coils are used to collect data to produce an error data set, and a classification training set and a prediction training set are produced from the error data set. A multi-layer perceptron classification model is trained based on the classification training set to identify different interference sources. A multi-layer perceptron prediction model is trained based on the prediction training set and the positioning error is predicted to achieve compensation for the positioning data of the electromagnetic navigation system when electromagnetic interference occurs, thereby improving the positioning accuracy and robustness of the electromagnetic navigation system.

[0060] The above-mentioned embodiments are specific implementation methods of the present invention, and the present invention is described and illustrated in detail, but it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple improvements or replacements can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for predicting and compensating positioning errors of a navigation system in an electromagnetic interference environment, characterized in that: The specific implementation process of the method includes the following steps: 1) Collect reference data and positioning data of the electromagnetic navigation system under electromagnetic interference environment to produce error data set; 2) Use the error data set to create a classification training set, and build and train a multi-layer perceptron classification model based on the classification training set; 3) Use the error data set to create a prediction training set, and build and train a multi-layer perceptron prediction model based on the prediction training set; 4) Obtaining positioning data of the electromagnetic navigation system during surgery; 5) Predict the positioning error of the electromagnetic navigation system during the operation based on the multi-layer perceptron classification model and the multi-layer perceptron prediction model, and compensate the positioning data of the electromagnetic navigation system based on the predicted positioning error; In the step 1), three positioning coils are used to prepare the error data set, wherein the positioning coil a and the positioning coil b are fixedly placed at different positions according to different surgical scopes, and the positioning coil c is used to track the surgical instrument. When the position offset of the positioning coil a or the positioning coil b is greater than its own repeatable positioning accuracy, it is determined that electromagnetic interference occurs. When electromagnetic interference occurs, the data recorded in the error data set are the interference source code causing the electromagnetic interference, the reference data P of the positioning coil a, and the reference data P of the positioning coil a. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ), reference data P of positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) and the reference data P of the positioning coil c c (X c ,Y c ,Z c ) and positioning data P c ′(X′ c ,Y c ′,Z′ c ); In the step 2), the interference source code causing electromagnetic interference in the error data set and the reference data P of the positioning coil a are used. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ) and the reference data P of the positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) to create a classification training set, the feature data of the classification training set is the reference data P of the positioning coil a a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ) and the reference data P of the positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ), the label data of the classification training set is the interference source code that causes electromagnetic interference; The multi-layer perceptron classification model built and trained based on the classification training set includes two hidden layers consisting of 64 neurons. The activation functions of the first hidden layer and the second hidden layer are: f(x)=max(0,x) and The multi-layer perceptron classification model uses Adam as the optimizer, and its loss function is In the step 3), the interference source code causing electromagnetic interference in the error data set and the reference data P of the positioning coil a are used. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ), reference data P of positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) and the reference data P of the positioning coil c c (X c ,Y c ,Z c ) and positioning data P c ′(X′ c ,Y c ′,Z′ c ) to make a prediction training set, the characteristic data of the prediction training set is the interference source code causing electromagnetic interference, the positioning data P of the positioning coil a a ′(X′ a ,Y a ′,Z′ a ), positioning data P of positioning coil b b ′(X′ b ,Y b ′,Z′ b ), positioning data P of positioning coil c c ′(X′ c ,Y c ′,Z′ c ), positioning error e of positioning coil a a (x a ,y a ,z a ) and the positioning error e of the positioning coil b b (x b ,y b ,z b ), the label data of the predicted training set is the positioning error e of the positioning coil c c (x c ,y c ,z c ), the positioning error calculation method is: The multi-layer perceptron prediction model built and trained based on the prediction training set includes three hidden layers consisting of 128 neurons. The activation functions of the first hidden layer and the second hidden layer are f(x)=max(0,x), and the activation function of the third hidden layer is f(x)=max(ax,x); The multi-layer perceptron prediction model uses Adam as the optimizer, and its loss function is In step 4), the intraoperative positioning coil a and the positioning coil b are fixedly placed at appropriate positions according to different surgical scopes, and the data of each positioning coil obtained during the operation is the reference data P of the positioning coil a. a (X a ,Y a ,Z a ), positioning data P of positioning coil a a ′(X′ a ,Y a ′,Z′ a ), positioning error e of positioning coil a a (x a ,y a ,z a ), reference data P of positioning coil b b (X b ,Y b ,Z b ), positioning data P of positioning coil b b ′(X′ b ,Y b ′,Z′ b ), positioning error e of positioning coil b b (x b ,y b ,z b ) and the positioning data P of the positioning coil c c ′(X′ c ,Y c ′,Z′ c ); In step 5), the multi-layer perceptron classification model is based on the reference data P of the intraoperative positioning coil a. a (X a ,Y a ,Z a ) and positioning data P a ′(X′ a ,Y a ′,Z′ a ), reference data P of positioning coil b b (X b ,Y b ,Z b ) and positioning data P b ′(X′ b ,Y b ′,Z′ b ) to identify the interference source type and pass the interference source code to the multi-layer perceptron prediction model, which is based on the positioning data P of the intraoperative positioning coil a a ′(X′ a ,Y a ′,Z′ a ), positioning data P of positioning coil b b ′(X′ b ,Y b ′,Z′ b ), positioning data P of positioning coil c c ′(X′ c ,Y c ′,Z′ c ), positioning error e of positioning coil a a (x a ,y a ,z a ), positioning error e of positioning coil b b (x b ,y b ,z b ) and the interference source code to predict the positioning error of the positioning coil c and obtain the predicted positioning error Finally, the predicted positioning error based on the positioning coil c The positioning data of the positioning coil c is compensated by: In the formula is the positioning data after compensation for the positioning coil c.

Images