A network data transmission method and system for network virtual surgery

Abstract

The present invention discloses a network data transmission method and system for network virtual surgery, comprising: obtaining working online features of the network virtual surgery, inputting the working online features into a pre-trained back propagation neural network to obtain four monitoring output features; calculating a data compression rate, a prediction compensation window size, a number of redundant retransmissions, and a local interpolation smoothing factor based on the four monitoring output features; compressing the working data of the network virtual surgery according to the data compression rate and then performing data transmission; during the data transmission process of the network virtual surgery, compensating for delays based on a Kalman predictor of the prediction compensation window size, triggering redundant retransmissions according to the number of redundant retransmissions to combat packet loss, and substituting the local interpolation smoothing factor into an interpolation filter to smooth network jitter; and combining the back propagation neural network with a virtual surgery force feedback system to achieve adaptive adjustment of delay, jitter, and packet loss problems in the network virtual surgery.

Description

Technical Field

[0001] The present invention belongs to the technical field of network data transmission, and in particular relates to a network data transmission method and system for network virtual surgery. Background Art

[0002] Virtual surgery systems pursue high authenticity and stability. They provide a more realistic operating experience for virtual surgery by simulating multimodal feedback in real surgery. Therefore, a network data transmission method for network virtual surgery is of great significance for improving surgical training effects and surgical safety.

[0003] To address the issue of aortic tumor growth, existing technologies use the finite element method (FEM) to establish a multiscale fluid-structure interaction model. This approach achieves tactile rendering through physical modeling, but suffers from large data volumes and long computation times, making it difficult to meet real-time requirements. The computational complexity of multiscale fluid-structure interaction leads to significant system response delays, especially in scenarios with frequent interactions. Localized lag effects can reduce the real-time and realism of force feedback.

[0004] The multiscale fluid-structure interaction model lacks an adaptive compensation mechanism for dynamic hysteresis errors and cannot adjust output weights based on the operator's real-time actions. Furthermore, the platform's stability and refresh rate need to be further improved to prevent signal interruptions or data loss. Summary of the Invention

[0005] The present invention provides a network data transmission method and system for network virtual surgery, which combines a back propagation neural network with a virtual surgery force feedback system to achieve adaptive adjustment of delay, jitter and packet loss problems in network virtual surgery.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A first aspect of the present invention provides a network data transmission method for network virtual surgery, comprising:

[0008] The online working features of the network virtual surgery are obtained and input into the pre-trained back-propagation neural network to obtain four monitoring output features. The data compression rate, prediction compensation window size, number of redundant retransmissions, and local interpolation smoothing factor are calculated based on the four monitoring output features.

[0009] The working data of network virtual surgery is compressed according to the data compression rate and then transmitted. During the data transmission process of network virtual surgery, a Kalman predictor based on the predicted compensation window size compensates for delays, redundant retransmissions are triggered according to the number of redundant retransmissions to combat packet loss, and the local interpolation smoothing factor is substituted into the interpolation filter to smooth network jitter.

[0010] Furthermore, the back propagation neural network includes an input layer, a first hidden layer, a second hidden layer and an output layer; the activation function in the first hidden layer and the second hidden layer is set to function; the activation function of the output layer is set to function.

[0011] Furthermore, the training process of the back propagation neural network includes:

[0012] Obtaining a work training feature of the network virtual surgery and inputting it into the input layer of the back propagation neural network, performing feature extraction on the work training feature through the first hidden layer to obtain a first intermediate feature, and inputting the first intermediate feature into the second hidden layer to obtain a second intermediate feature;

[0013] The second intermediate features are clustered through the output layer to obtain the training output features, and the first training loss value of the back propagation neural network is calculated based on the training output features; the weight of the back propagation neural network is optimized using the first training loss value.

[0014] Based on the training output characteristics, the transmission parameters of the network virtual surgery are set and data transmission is performed, and the working monitoring characteristics of the data transmission process in the network virtual surgery are recorded; at each set interval, a second training loss value is calculated based on the working monitoring characteristics of two adjacent records, and the second training loss value is used to optimize the weight of the back propagation neural network;

[0015] The work monitoring features are used as work training features to retrain the back propagation neural network, and the training process of the back propagation neural network is repeated until a preset threshold is reached to output the trained back propagation neural network.

[0016] Furthermore, the work online features, work monitoring features and work training features include end-to-end delay during network virtual surgery , Packet Loss Rate , bandwidth utilization , network jitter , Equipment movement speed , the proportion of key areas of tissue deformation and force feedback error .

[0017] Furthermore, feature extraction is performed on the working training features through the first hidden layer to obtain the first intermediate features, specifically including:

[0018]

[0019] In the formula, Represents the first hidden layer The first intermediate feature output by the node, is the activation function; is the number of nodes in the input layer; is the node number in the input layer; Represents the input layer nodes to the first hidden layer The connection weight of each node, Represents the input layer The training features of the work received by the nodes, Represents the first hidden layer The bias term of each node.

[0020] Furthermore, inputting the first intermediate feature into the second hidden layer to obtain the second intermediate feature specifically includes:

[0021]

[0022] In the formula, The second hidden layer The second intermediate feature output by the node, is the number of nodes in the first hidden layer; is the node number in the first hidden layer; Represents the first hidden layer nodes to the second hidden layer The connection weight of each node, Represents the first hidden layer The first intermediate feature output by the node, The second hidden layer The bias term of each node.

[0023] Furthermore, the second intermediate features are clustered through the output layer to obtain training output features, specifically including:

[0024]

[0025]

[0026]

[0027]

[0028] In the formula, 、 、 and are the training output features output by each node in the output layer; is the activation function; is the number of nodes in the second hidden layer; is the node number in the second hidden layer; The second hidden layer The connection weight from the node to the first node in the output layer, The second hidden layer The connection weight from the node to the second node in the output layer, The second hidden layer The connection weight from the node to the third node in the output layer, The second hidden layer The connection weight from the node to the fourth node in the output layer; The second hidden layer The second intermediate feature output by the node; is the bias term of the first node in the output layer, is the bias term of the second node in the output layer, is the bias term of the third node in the output layer, is the bias term of the fourth node in the output layer.

[0029] Furthermore, the first training loss value of the back propagation neural network is calculated based on the training output features, specifically including:

[0030]

[0031]

[0032] , ,

[0033]

[0034] in, represents the loss function, represents the weight coefficient of effective delay, Indicates the effective delay, represents the weight coefficient of force feedback error, represents the predicted force feedback value, Indicates the actual force feedback value, Represents the weight coefficient of the compression rate fluctuation term, represents the rate of change of compression ratio over time, Indicates the data compression ratio, Indicates time, The weight coefficient of the penalty term representing the number of redundant retransmissions, represents the penalty term for the number of redundant retransmissions, is the number of redundant retransmissions; The end-to-end delay during network virtual surgery, is the local interpolation smoothing factor, To predict the weight of the compensation window size, Compensate window size for prediction; 、 、 and are the training output features output by each node in the output layer.

[0035] Furthermore, at each set time interval, a second training loss value is calculated based on the work monitoring features of two adjacent records, specifically including:

[0036] Obtaining the work monitoring features of two adjacent records at set intervals, and using the work monitoring features as training monitoring samples, wherein the training monitoring samples include the actual lag time, data compression rate, force feedback value, and redundant retransmission times of the network virtual surgery;

[0037] The second training loss value is calculated based on the work monitoring features, and the expression formula is:

[0038] ;

[0039] ;

[0040] In the formula, 、 、 Respectively 、 、 Back propagation neural network weights after training with sliding sampling window; represents the learning rate, The number of training monitoring samples within the sliding replenishment window; For the The gradient of the loss value in the training iteration, Indicates the The loss value in training iterations; represents the L2 regularization coefficient; To set weights; For the The actual lag time after backpropagating the neural network adjustment strategy for training iterations; Indicates the The force feedback value is predicted in training iterations, Indicates the The actual force feedback value in the training iteration; is the unit time of each training iteration of the back-propagation neural network; and Respectively 、 Data compression ratio in training iterations; For the The number of redundant retransmissions in training iterations.

[0041] Furthermore, it also includes: when the proportion of the key area of ​​tissue deformation in network virtual surgery is greater than When the data compression rate is limited to When the speed of the instrument in the network virtual surgery exceeds the set speed threshold, the prediction compensation window size is increased according to the set ratio. When the force feedback error in network virtual surgery exceeds the set error threshold, the number of redundant retransmissions is increased according to the set ratio.

[0042] A second aspect of the present invention provides a network data transmission system for network virtual surgery, comprising:

[0043] An acquisition module, used to obtain the working online features of network virtual surgery;

[0044] The prediction module is used to input the working online features into the pre-trained back-propagation neural network to obtain four monitoring output features; based on the four monitoring output features, the data compression rate, the prediction compensation window size, the number of redundant retransmissions, and the local interpolation smoothing factor are calculated;

[0045] The execution module compresses the working data of the network virtual surgery according to the data compression rate and then executes data transmission; during the data transmission process of the network virtual surgery, the Kalman predictor based on the predicted compensation window size compensates for the delay, redundant retransmission is triggered according to the number of redundant retransmissions to combat packet loss, and the local interpolation smoothing factor is substituted into the interpolation filter to smooth network jitter.

[0046] Furthermore, a training module is included, and the training module trains the back propagation neural network, specifically including:

[0047] Obtaining a work training feature of the network virtual surgery and inputting it into the input layer of the back propagation neural network, performing feature extraction on the work training feature through the first hidden layer to obtain a first intermediate feature, and inputting the first intermediate feature into the second hidden layer to obtain a second intermediate feature;

[0048] The second intermediate features are clustered through the output layer to obtain the training output features, and the first training loss value of the back propagation neural network is calculated based on the training output features; the weight of the back propagation neural network is optimized using the first training loss value.

[0049] Based on the training output characteristics, the transmission parameters of the network virtual surgery are set and data transmission is performed, and the working monitoring characteristics of the data transmission process in the network virtual surgery are recorded; at each set interval, a second training loss value is calculated based on the working monitoring characteristics of two adjacent records, and the second training loss value is used to optimize the weight of the back propagation neural network;

[0050] The work monitoring features are used as work training features to retrain the back propagation neural network, and the training process of the back propagation neural network is repeated until a preset threshold is reached to output the trained back propagation neural network.

[0051] The third aspect of the present invention provides an electronic device comprising a storage medium and a processor; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the network data transmission method for network virtual surgery described in the first aspect.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] The present invention combines a back-propagation neural network with a virtual surgery force feedback system to achieve adaptive adjustment of delay, jitter, and packet loss problems in networked virtual surgery, thereby maintaining optimal performance under different network conditions and surgical scenarios and enhancing the robustness of the system.

[0054] The present invention can effectively compensate for the delay in network transmission and ensure the real-time performance of the force feedback signal through a Kalman predictor based on the prediction compensation window size; substitute the local interpolation smoothing factor into the interpolation filter to smooth the impact of network jitter on the force feedback signal, further improving the stability and accuracy of the force feedback; trigger redundant retransmission according to the number of redundant retransmissions, which can effectively combat the problem of network packet loss.

[0055] This invention significantly enhances the system's adaptability and stability by dynamically adjusting neural network weights and transmission parameters through real-time monitoring of network status and surgical operation changes. Using operational monitoring features as training data, the network is continuously iteratively optimized, enabling the system to better adapt to complex network environments and surgical requirements, improving force feedback authenticity and overall performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a flow chart of a network data transmission method for network virtual surgery provided by Example 1 of the present invention;

[0057] Figure 2 1 is a structural diagram of the back propagation neural network provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0058] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0059] Example 1

[0060] like Figures 1 to 2 As shown, this embodiment provides a network data transmission method for network virtual surgery, including:

[0061] Construct the input layer, first hidden layer, second hidden layer and output layer of the back propagation neural network; in this embodiment, the input layer has 7 nodes, the first hidden layer and the second hidden layer have 12 nodes, and the output layer has 4 nodes; the activation function in the first hidden layer and the second hidden layer is set to function; the activation function of the output layer is set to A mixture of functional and linear activations.

[0062] Training the back propagation neural network specifically includes:

[0063] Obtaining work training characteristics, including end-to-end delay during network virtual surgery , Packet Loss Rate , bandwidth utilization , network jitter , Equipment movement speed , the proportion of key areas of tissue deformation and force feedback error

[0064] The work training features are input into the input layer of the back propagation neural network, and the work training features are extracted through the first hidden layer to obtain the first intermediate features, which are expressed as follows:

[0065]

[0066] In the formula, Represents the first hidden layer The first intermediate feature output by the node, is the activation function; is the number of nodes in the input layer. ; is the node number in the input layer; Represents the input layer nodes to the first hidden layer The connection weight of each node, Represents the input layer The training features of the work received by the nodes, Represents the first hidden layer The bias term of each node.

[0067] The first intermediate feature is input into the second hidden layer to obtain the second intermediate feature; the expression formula is:

[0068]

[0069] In the formula, The second hidden layer The second intermediate feature output by the node, is the number of nodes in the first hidden layer. In this embodiment ; is the node number in the first hidden layer; Represents the first hidden layer nodes to the second hidden layer The connection weight of each node, Represents the first hidden layer The first intermediate feature output by the node, The second hidden layer The bias term of each node.

[0070] The second intermediate features are clustered through the output layer to obtain the training output features, specifically including:

[0071]

[0072]

[0073]

[0074]

[0075] In the formula, is the training output feature output by the first node of the output layer, is the training output feature output by the second node of the output layer, is the training output feature output by the third node of the output layer, The training output feature output by the fourth node of the output layer; is the activation function; is the number of nodes in the second hidden layer. ; is the node number in the second hidden layer; The second hidden layer The connection weight from the node to the first node in the output layer, The second hidden layer The connection weight from the node to the second node in the output layer, The second hidden layer The connection weight from the node to the third node in the output layer, The second hidden layer The connection weight from the node to the fourth node in the output layer; The second hidden layer The second intermediate feature output by the node; is the bias term of the first node in the output layer, is the bias term of the second node in the output layer, is the bias term of the third node in the output layer, is the bias term of the fourth node in the output layer.

[0076] Calculate the first training loss value of the back propagation neural network based on the training output features; the expression formula is:

[0077]

[0078]

[0079] , ,

[0080]

[0081] in, represents the loss function, represents the weight coefficient of effective delay, Indicates the effective delay, represents the weight coefficient of force feedback error, represents the predicted force feedback value, Indicates the actual force feedback value, represents the weight coefficient of the compression rate fluctuation term, represents the rate of change of compression ratio over time, Indicates the data compression ratio, Indicates time, The weight coefficient of the penalty term representing the number of redundant retransmissions, represents the penalty term for the number of redundant retransmissions, is the number of redundant retransmissions; The end-to-end delay during network virtual surgery, is the local interpolation smoothing factor, To predict the weight of the compensation window size, Compensate window size for prediction; is the training output feature output by the third node of the output layer, is the training output feature output by the first node of the output layer, is the training output feature output by the 4th node of the output layer, It is the training output feature output by the second node of the output layer.

[0082] The first training loss value is used to optimize the weights of the back propagation neural network, specifically including:

[0083]

[0084]

[0085]

[0086]

[0087] In the formula, represents the gradient of data compression rate; represents the gradient of the prediction compensation window size, represents the learning rate; represents the gradient of the number of redundant retransmissions; represents the gradient of the local interpolation smoothing factor;

[0088] Calculate the gradients of the nodes in the second hidden layer and the first hidden layer, specifically including:

[0089]

[0090]

[0091] In the formula, The second hidden layer The gradient of the node, Represents the output layer The gradient of the node, The second hidden layer nodes to the output layer The connection weight of each node, represents the indicator function, when Take 1 when it is, otherwise take 0;

[0092] Represents the first hidden layer The gradient of the node, The second hidden layer The gradient of the node, Represents the first hidden layer nodes to the second hidden layer The weight of each node; represents the indicator function, when Take 1 when it is, otherwise take 0;

[0093]

[0094]

[0095]

[0096] In the formula, represents the learning rate, represents the L2 regularization coefficient.

[0097] Setting transmission parameters of the network virtual surgery based on the training output characteristics and performing data transmission, and recording work monitoring characteristics of the data transmission process in the network virtual surgery;

[0098] The second training loss value is calculated based on the work monitoring features of two adjacent records at each set interval, specifically including:

[0099] Obtaining the work monitoring features of two adjacent records at set intervals, and using the work monitoring features as training monitoring samples, wherein the training monitoring samples include the actual lag time, data compression rate, force feedback value, and redundant retransmission times of the network virtual surgery;

[0100] The second training loss value is calculated based on the work monitoring features, and the expression formula is:

[0101] ;

[0102] ;

[0103] In the formula, 、 、 Respectively 、 、 Back propagation neural network weights after training with sliding sampling window; represents the learning rate, The number of training monitoring samples within the sliding replenishment window; For the The gradient of the loss value in the training iteration, Indicates the The loss value in training iterations; represents the L2 regularization coefficient; To set weights; For the The actual lag time after backpropagating the neural network adjustment strategy for training iterations; Indicates the The force feedback value is predicted in training iterations, Indicates the The actual force feedback value in the training iteration; and Respectively 、 Data compression ratio in training iterations; is the unit time of each training iteration of the back-propagation neural network; For the The number of redundant retransmissions in training iterations.

[0104] The second training loss value is used to optimize the weights of the back propagation neural network; the work monitoring features are used as work training features to retrain the back propagation neural network, and the training process of the back propagation neural network is repeated until a preset threshold is reached to output the trained back propagation neural network.

[0105] This embodiment significantly enhances the system's adaptability and stability by dynamically adjusting neural network weights and transmission parameters through real-time monitoring of network status and surgical operation changes. Using operational monitoring features as training data, the network is continuously iteratively optimized, enabling the system to better adapt to complex network environments and surgical requirements, improving force feedback realism and overall performance.

[0106] Obtain online features of network virtual surgery, including end-to-end delay during network virtual surgery , Packet Loss Rate , bandwidth utilization , network jitter , Equipment movement speed , the proportion of key areas of tissue deformation and force feedback error .

[0107] Input the working online features into the pre-trained back-propagation neural network to obtain four monitoring output features; calculate the data compression rate, prediction compensation window size, redundant retransmission times, and local interpolation smoothing factor based on the four monitoring output features;

[0108] The working data of network virtual surgery is compressed according to the data compression rate and then transmitted. During the data transmission process of network virtual surgery, a Kalman predictor based on the predicted compensation window size compensates for delays, redundant retransmissions are triggered according to the number of redundant retransmissions to combat packet loss, and the local interpolation smoothing factor is substituted into the interpolation filter to smooth network jitter.

[0109] When the proportion of key tissue deformation areas in network virtual surgery is greater than When the data compression rate is limited to When the speed of the instrument in the network virtual surgery exceeds the set speed threshold, the prediction compensation window size is increased according to the set ratio. When the force feedback error in network virtual surgery exceeds the set error threshold, the number of redundant retransmissions is increased according to the set ratio.

[0110] Through the Kalman predictor based on the prediction compensation window size, the delay in network transmission can be effectively compensated to ensure the real-time performance of the force feedback signal; the local interpolation smoothing factor is substituted into the interpolation filter to smooth the impact of network jitter on the force feedback signal, further improving the stability and accuracy of the force feedback; redundant retransmission is triggered according to the number of redundant retransmissions, which can effectively combat the problem of network packet loss.

[0111] This embodiment combines a backpropagation neural network with a virtual surgery force feedback system to achieve adaptive adjustment of delay, jitter, and packet loss problems in networked virtual surgery, thereby maintaining optimal performance under different network conditions and surgical scenarios and enhancing the robustness of the system.

[0112] Example 2

[0113] This embodiment provides a network data transmission system for network virtual surgery. The network data transmission system can execute the network data transmission method described in Example 1. The network data transmission system includes:

[0114] An acquisition module, used to obtain the working online features of network virtual surgery;

[0115] The prediction module is used to input the working online features into the pre-trained back-propagation neural network to obtain four monitoring output features; based on the four monitoring output features, the data compression rate, the prediction compensation window size, the number of redundant retransmissions, and the local interpolation smoothing factor are calculated;

[0116] The execution module compresses the working data of the network virtual surgery according to the data compression ratio and then transmits the data. During the data transmission process of the network virtual surgery, a Kalman predictor based on the predicted compensation window size compensates for delays, redundant retransmissions are triggered based on the number of redundant retransmissions to combat packet loss, and a local interpolation smoothing factor is substituted into the interpolation filter to smooth network jitter.

[0117] Training module, trains the back-propagation neural network.

[0118] The training module trains the back propagation neural network, specifically including:

[0119] Obtaining a work training feature of the network virtual surgery and inputting it into the input layer of the back propagation neural network, performing feature extraction on the work training feature through the first hidden layer to obtain a first intermediate feature, and inputting the first intermediate feature into the second hidden layer to obtain a second intermediate feature;

[0120] The second intermediate features are clustered through the output layer to obtain the training output features, and the first training loss value of the back propagation neural network is calculated based on the training output features; the weight of the back propagation neural network is optimized using the first training loss value.

[0121] Based on the training output characteristics, the transmission parameters of the network virtual surgery are set and data transmission is performed, and the working monitoring characteristics of the data transmission process in the network virtual surgery are recorded; at each set interval, a second training loss value is calculated based on the working monitoring characteristics of two adjacent records, and the second training loss value is used to optimize the weight of the back propagation neural network;

[0122] The work monitoring features are used as work training features to retrain the back propagation neural network, and the training process of the back propagation neural network is repeated until a preset threshold is reached to output the trained back propagation neural network.

[0123] Example 3

[0124] This embodiment provides an electronic device, including a storage medium and a processor; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the network data transmission method for network virtual surgery described in Example 1.

[0125] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0126] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0127] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0128] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0129] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A network data transmission method for network virtual surgery, characterized in that: include: Obtain online features of the network virtual surgery, and input the online features into the pre-trained back-propagation neural network to obtain four monitoring output features; The data compression rate, prediction compensation window size, redundant retransmission times and local interpolation smoothing factor are calculated based on the four monitoring output characteristics; The working data of the network virtual surgery is compressed according to the data compression ratio and then transmitted. During the data transmission process of the network virtual surgery, a Kalman predictor based on the predicted compensation window size compensates for delays, redundant retransmissions are triggered based on the number of redundant retransmissions to combat packet loss, and a local interpolation smoothing factor is substituted into the interpolation filter to smooth network jitter. The back propagation neural network includes an input layer, a first hidden layer, a second hidden layer and an output layer; The training process of the back propagation neural network includes: Obtaining a work training feature of the network virtual surgery and inputting it into the input layer of the back propagation neural network, performing feature extraction on the work training feature through the first hidden layer to obtain a first intermediate feature, and inputting the first intermediate feature into the second hidden layer to obtain a second intermediate feature; The second intermediate features are clustered through the output layer to obtain the training output features. The first training loss value of the back propagation neural network is calculated based on the training output features. The expression formula is: ； ； ， ， ； ； in, represents the loss function, represents the weight coefficient of effective delay, Indicates the effective delay, represents the weight coefficient of force feedback error, represents the predicted force feedback value, Indicates the actual force feedback value, represents the weight coefficient of the compression rate fluctuation term, represents the rate of change of compression ratio over time, Indicates the data compression ratio, Indicates time, The weight coefficient of the penalty term representing the number of redundant retransmissions, represents the penalty term for the number of redundant retransmissions, is the number of redundant retransmissions; The end-to-end delay during network virtual surgery, is the local interpolation smoothing factor, To predict the weight of the compensation window size, Compensate window size for prediction; 、 、 and are the training output features output by each node in the output layer; Optimizing the weights of the back propagation neural network using the first training loss value; Based on the training output characteristics, the transmission parameters of the network virtual surgery are set and data transmission is performed, and the working monitoring characteristics of the data transmission process during the network virtual surgery are recorded; at each set interval, a second training loss value is calculated based on the working monitoring characteristics of two adjacent records, and the second training loss value is used to optimize the weights of the back propagation neural network; The work monitoring features are used as work training features to retrain the back propagation neural network, and the training process of the back propagation neural network is repeated until a preset threshold is reached to output the trained back propagation neural network.

2. The network data transmission method according to claim 1, characterized in that: The work online features, work monitoring features and work training features include end-to-end delay during network virtual surgery , Packet Loss Rate , bandwidth utilization , network jitter , Equipment movement speed , the proportion of key areas of tissue deformation and force feedback error .

3. The network data transmission method according to claim 1, wherein: Extracting the working training features through the first hidden layer to obtain a first intermediate feature, and inputting the first intermediate feature into the second hidden layer to obtain a second intermediate feature, specifically comprising: ； In the formula, Represents the first hidden layer The first intermediate feature output by the node, is the activation function; is the number of nodes in the input layer; is the node number in the input layer; Represents the input layer nodes to the first hidden layer The connection weight of each node, Represents the input layer The training features of the work received by the nodes, Represents the first hidden layer The bias term of each node; ； In the formula, The second hidden layer The second intermediate feature output by the node, is the number of nodes in the first hidden layer; is the node number in the first hidden layer; Represents the first hidden layer nodes to the second hidden layer The connection weight of each node, The second hidden layer The bias term of each node.

4. The network data transmission method according to claim 1, wherein: The second intermediate features are clustered through the output layer to obtain the training output features, specifically including: ； ； ； ； In the formula, 、 、 and are the training output features output by each node in the output layer; is the activation function; is the number of nodes in the second hidden layer; is the node number in the second hidden layer; 、 、 and The second hidden layer The connection weights from each node to each node in the output layer; The second hidden layer The second intermediate feature output by the node; 、 、 and is the bias term of each node in the output layer.

5. The network data transmission method according to claim 1, wherein: The second training loss value is calculated based on the work monitoring features of two adjacent records at each set interval, specifically including: Obtaining the work monitoring features of two adjacent records at set intervals, and using the work monitoring features as training monitoring samples, wherein the training monitoring samples include the actual lag time, data compression rate, force feedback value, and redundant retransmission times of the network virtual surgery; The second training loss value is calculated based on the work monitoring features, and the expression formula is: ； ； In the formula, 、 、 Respectively 、 、 Back propagation neural network weights after training with sliding sampling window; represents the learning rate, Monitor the number of samples for training within the sliding sampling window; For the The gradient of the loss value in the training iteration, Indicates the The loss value in training iterations; represents the L2 regularization coefficient; To set weights; For the The actual lag time after backpropagating the neural network adjustment strategy for training iterations; Indicates the The force feedback value is predicted in training iterations, Indicates the The actual force feedback value in the training iteration; and Respectively 、 Data compression ratio in training iterations; is the unit time of each training iteration of the back-propagation neural network; For the The number of redundant retransmissions in training iterations.

6. The network data transmission method according to claim 1, characterized in that: Also includes: When the proportion of key tissue deformation areas in network virtual surgery is greater than When the data compression rate is limited to ; When the instrument movement speed in network virtual surgery exceeds the set speed threshold, the prediction compensation window size is increased according to the set ratio; when the force feedback error in network virtual surgery exceeds the set error threshold, the number of redundant retransmissions is increased according to the set ratio.

7. A network data transmission system for network virtual surgery, characterized in that: include: An acquisition module, used to obtain the working online features of network virtual surgery; The prediction module is used to input the working online features into the pre-trained back-propagation neural network to obtain four monitoring output features; The data compression rate, prediction compensation window size, redundant retransmission times and local interpolation smoothing factor are calculated based on the four monitoring output characteristics; The execution module compresses the working data of the network virtual surgery according to the data compression ratio and then transmits the data. During the data transmission process of the network virtual surgery, a Kalman predictor based on the predicted compensation window size compensates for delays, redundant retransmissions are triggered based on the number of redundant retransmissions to combat packet loss, and a local interpolation smoothing factor is substituted into the interpolation filter to smooth network jitter. The training process of the back propagation neural network includes: Obtaining a work training feature of the network virtual surgery and inputting it into the input layer of the back propagation neural network, performing feature extraction on the work training feature through the first hidden layer to obtain a first intermediate feature, and inputting the first intermediate feature into the second hidden layer to obtain a second intermediate feature; The second intermediate features are clustered through the output layer to obtain the training output features. The first training loss value of the back propagation neural network is calculated based on the training output features. The expression formula is: ； ； ， ， ； ； in, represents the loss function, represents the weight coefficient of effective delay, Indicates the effective delay, represents the weight coefficient of force feedback error, represents the predicted force feedback value, Indicates the actual force feedback value, represents the weight coefficient of the compression rate fluctuation term, represents the rate of change of compression ratio over time, Indicates the data compression ratio, Indicates time, The weight coefficient of the penalty term representing the number of redundant retransmissions, represents the penalty term for the number of redundant retransmissions, is the number of redundant retransmissions; The end-to-end delay during network virtual surgery, is the local interpolation smoothing factor, To predict the weight of the compensation window size, Compensate window size for prediction; 、 、 and are the training output features output by each node in the output layer; Optimizing the weights of the back propagation neural network using the first training loss value; Based on the training output characteristics, the transmission parameters of the network virtual surgery are set and data transmission is performed, and the working monitoring characteristics of the data transmission process during the network virtual surgery are recorded; at each set interval, a second training loss value is calculated based on the working monitoring characteristics of two adjacent records, and the second training loss value is used to optimize the weights of the back propagation neural network; The work monitoring features are used as work training features to retrain the back propagation neural network, and the training process of the back propagation neural network is repeated until a preset threshold is reached to output the trained back propagation neural network.

8. An electronic device comprising a storage medium and a processor; the storage medium is used to store instructions; characterized in that, The processor is configured to operate according to the instructions to execute the network data transmission method for network virtual surgery according to any one of claims 1 to 6.

Images