AR-guided infusion port puncture navigation system

Abstract

The invention belongs to the field of medical technology and information technology, and discloses an AR-guided infusion port puncture navigation system, which collects and marks historical data of a large number of puncture operations, and uses a deep learning algorithm to train a model so as to identify a correct operation mode. The visual angle, motion trail, gesture change, force and space coordinate data of an operator are captured in real time, rapidly analyzed and input into a model for real-time judgment, a multi-sensory feedback mechanism is integrated on an AR interface, visual, auditory and tactile feedback is provided in real time according to a model judgment result, the operator is guided to adjust actions, and the real-time performance of the operator is improved. The feedback data in the operation process are continuously collected, the feedback effect is regularly evaluated, the model is iteratively optimized, and a continuously improved closed loop mechanism is formed, so that the accuracy and efficiency of medical operation are improved, the medical risk is reduced, the method can be widely applied to clinical teaching, skill training and actual operation, and the medical service quality is remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the field of medical technology and information technology, and specifically is an AR-guided infusion port puncture navigation system. Background Art

[0002] In the past few decades, the field of medical education and practice has undergone significant technological innovation, among which the application of augmented reality (AR) technology is particularly noteworthy. AR technology brings unprecedented possibilities to the medical field by superimposing computer-generated information on the user's real-world view.

[0003] In recent years, AR technology has begun to be applied to medical training to provide an immersive learning experience, but existing systems mainly focus on visual simulation and lack real-time personalized guidance and multi-sensory feedback, making it difficult to effectively improve operational accuracy.

[0004] To this end, the present invention provides an AR-guided infusion port puncture navigation system. By integrating AR technology and deep learning algorithms, an intelligent medical operation assistance system is constructed that can monitor operation details in real time, provide personalized guidance and objectively evaluate operation levels. It significantly improves the accuracy and efficiency of operations, reduces medical risks, and provides medical students with an efficient and safe skills training platform. Summary of the invention

[0005] In order to make up for the deficiencies of the prior art, at least one technical problem raised in the background technology is solved.

[0006] The technical solution adopted by the present invention to solve its technical problems is: the AR-guided infusion port puncture navigation system described in the present invention collects and annotates a large amount of historical data of puncture operations, uses a deep learning algorithm to train a model to identify the correct operation mode, and captures the operator's perspective, motion trajectory, gesture changes, strength and spatial coordinate data in real time, quickly analyzes and inputs them into the model for real-time judgment, integrates a multi-sensory feedback mechanism on the AR interface, provides visual, auditory and tactile feedback in real time according to the model judgment results, guides the operator to adjust his movements, continuously collects feedback data during the operation, regularly evaluates the feedback effect, and iteratively optimizes the model to form a closed-loop mechanism of continuous improvement to improve the accuracy and efficiency of medical operations.

[0007] The purpose of the present invention can be achieved by the following technical solution. An AR-guided infusion port puncture navigation system includes the following steps: Step 1: Collect a large amount of historical puncture operation data, including data on successful and failed cases, extract key action feature data from the data, annotate the data in detail, distinguish correct operation steps from potential errors, and provide "right" and "wrong" labels for model learning; Use deep learning algorithms to build models and input labeled data for training, so that the models can identify correct and incorrect operation modes and optimize accuracy and efficiency evaluation standards; Step 2: Capture the operator's working information (including perspective information, motion information, hand motion trajectory, gesture changes, puncture force, pressure distribution and spatial coordinate data), quickly process the captured real-time data, obtain corresponding feature data, and use the model to analyze these feature data in real time; Step 3: Integrate visual, auditory, and tactile feedback mechanisms on the AR interface to synchronize with the user's operations in real time; Input the results of real-time analysis into the trained model, allowing the model to determine the accuracy of the current operation in real time and identify whether deviations or errors occur; Based on the judgment results of the model, the system provides intuitive visual feedback (color coding: green - correct, red - wrong; arrow indication: indicating the correct position and angle), auditory prompts (voice guidance: guiding the user in every step of the operation, especially providing immediate verbal correction suggestions when errors are found) and tactile feedback (using the vibration of force feedback gloves) through the AR interface to guide the user to adjust the operation to achieve the best effect; Step 4: Record the feedback provided by the system during each operation and the actual response of the user, including the operation data after the user corrects the behavior; Regularly evaluate the quality of feedback provided by the system and the improvement of user operations, and measure the effect through quantitative indicators; Based on the new round of feedback data collected, the model is iteratively trained and tuned to continuously improve its recognition accuracy and feedback effectiveness, forming a closed-loop continuous optimization mechanism.

[0008] As a further solution of the present invention: key action feature data, including operation path and puncture point coordinates, time series, important events (successful puncture, wrong puncture and puncture deviation) and force size and direction.

[0009] As a further solution of the present invention: mapping the coordinates of all puncture paths and puncture points into a unified range [0, 1]; Set the operation start time to 0 and the end time to 1, and convert the time data corresponding to all operations into the time percentage relative to the start of the operation, so that the duration of all operations is mapped to the same scale, that is, between [0, 1]; Each event of successful puncture, wrong puncture and puncture deviation is converted into a vector, where the position of the corresponding event is 1 and the rest are 0; Align the processed data according to the operation samples and integrate them into a complete visual data set; Obtain force data (force magnitude and direction) from the sensor and align the force data with the visual data according to the operation sample.

[0010] As a further solution of the present invention: the processed puncture path and puncture point data are constructed into an array of shape (N, M, D), where N is the number of puncture operation samples, each sample has coordinate records of M time points, and D is the coordinate dimension (usually 2 or 3, corresponding to two-dimensional or three-dimensional space); The processed time data is constructed into an array of shape (N, M), where each row represents the operation time series of a sample; The encoding of each event is formed into a matrix of shape (N, C), where C is the number of event types and each row represents the event sequence of a sample during the operation.

[0011] As a further solution of the present invention: the model is a multimodal fusion model, a model architecture consisting of a convolutional neural network (CNN) + a recurrent neural network (RNN), and the training set is used to train the model; Among them, the convolutional neural network algorithm processes visual image data and extracts spatial features; the recurrent neural network algorithm processes time series data, namely force feedback data, to capture the time dependency in the action sequence.

[0012] As a further solution of the present invention: the model selects the ResNet (residual network) model in the convolutional neural network (CNN) as a visual pre-training model, and uses the input of visual data to train the model to extract spatial features (including puncture point position, path, angle, and subtle changes in the skin surface) from the visual record of the puncture process for precise positioning and path planning; A recurrent neural network (RNN) model is used as a force pre-training model. By inputting force signals, the model outputs a set of vectors representing force characteristics, which are used to reveal key information during the puncture process (including force, speed, rhythm, and coordination).

[0013] As a further solution of the present invention: the acquired raw time data is processed into time series data points to obtain the moving speed of the operator's hand or the puncture needle when working; Based on the velocity data, the acceleration of the operator's hand or the puncture needle when working is obtained; The maximum force during puncture is identified in the acquired force feedback data, and the average force value is obtained according to the corresponding time series.

[0014] As a further solution of the present invention: the processed data is used as input to a trained model for real-time analysis to predict the accuracy and effectiveness of the operator's current operation and determine whether there is a deviation or error. Specifically: After real-time data is processed, the prediction value and deviation value are obtained through the model. When the prediction value is greater than the standard threshold, y=1 is predicted; otherwise, y=0 is predicted. When y=1, it indicates that the current operation is correct, and when y=0, it indicates that the current operation is wrong. Based on the "correct" or "wrong" result of the actual operation and the deviation value between the actual operation and the target operation, the result information of whether there is deviation and error in the operator's current operation is generated.

[0015] As a further solution of the present invention: based on the judgment result of the model, intuitive visual feedback, auditory prompts and tactile feedback are provided through the AR interface to guide the user to adjust the operation to achieve the best effect.

[0016] As a further solution of the present invention: recording all interaction data during each operation, using the interaction record of the sensor and the user interface, tracking the user's operation correction after receiving feedback, and recording the final result of each operation; Compare the operator's operation data before receiving feedback with the operation data after receiving feedback. If the deviation and error information in the operation data after receiving feedback are less than the operation data before receiving feedback, it means that the feedback has brought about obvious operation improvement; if the deviation and error information in the operation data after receiving feedback is greater than or equal to the operation data before receiving feedback, it means that the feedback has not brought about operation improvement; Get the operator's complete and correct operation time for each time, build a plane coordinate system, set the operator's number of times as the x-axis, and the operation time as the y-axis, substitute the number of times and the operation time into the coordinate system, and analyze to get the trend line of the operation time changing with the number of times. By observing the slope of the trend line, if the trend line tilts downward, it means that the feedback has played a continuous role in promoting the operator's skill proficiency.

[0017] The beneficial effects of the present invention are as follows: 1. The AR-guided infusion port puncture navigation system described in the present invention, by integrating AR technology and deep learning algorithms, can monitor and guide medical operations in real time, greatly improving the accuracy and efficiency of surgery, while significantly reducing medical risks, providing doctors with instant feedback and personalized suggestions, and ensuring the safety and effectiveness of the medical process.

[0018] 2. The AR-guided infusion port puncture navigation system described in the present invention uses simulation technology to allow medical students to hone their skills without actual risks, thereby accelerating their professional growth. The system's built-in evaluation mechanism can objectively evaluate the operation level and provide a specific path for skill improvement, which not only reduces the cost of traditional practical teaching, but also improves resource utilization.

[0019] 3. The AR-guided infusion port puncture navigation system described in the present invention provides a scientific basis for medical decision-making through data-driven decision support, promotes the optimal allocation of medical resources, and can effectively improve the level of medical services in remote areas and promote medical equity. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The present invention will be further described below in conjunction with the accompanying drawings.

[0021] Figure 1 A flow chart of an AR-guided infusion port puncture navigation system; Figure 2 A flowchart of processing raw data in an AR-guided infusion port puncture navigation system according to an embodiment of the present invention; Figure 3 It is an operation logic diagram of an AR-guided infusion port puncture navigation system according to an embodiment of the present invention; Figure 4 Schematic diagram of the structure of an AR-guided infusion port puncture navigation system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0023] Example 1 like Figure 1-Figure 4 As shown, an AR-guided infusion port puncture navigation system according to an embodiment of the present invention is specifically: Step 1: The data collection training module is used to collect a large amount of historical puncture operation data, including data on successful and failed cases, extract key action feature data from the data, annotate the data in detail, distinguish correct operation steps from potential errors, and provide "right" and "wrong" labels for model learning; Use deep learning algorithms to build models and input labeled data for training, so that the models can identify correct and incorrect operation modes and optimize accuracy and efficiency evaluation standards; Step 2: The real-time data analysis module is used to capture the information of the operator during work (including visual angle information, motion information, hand motion trajectory, gesture changes, puncture force, pressure distribution and spatial coordinate data), quickly process the captured real-time data, obtain corresponding feature data, and use the model to analyze these feature data in real time; Step 3: The real-time feedback module is used to integrate visual, auditory, and tactile feedback mechanisms on the AR interface, which are synchronized with the user's operations in real time; Input the results of real-time analysis into the trained model, allowing the model to determine the accuracy of the current operation in real time and identify whether deviations or errors occur; Based on the judgment results of the model, the system provides intuitive visual feedback (color coding: green - correct, red - wrong; arrow indication: indicating the correct position and angle), auditory prompts (voice guidance: guiding the user in every step of the operation, especially providing immediate verbal correction suggestions when errors are found) and tactile feedback (using the vibration of force feedback gloves) through the AR interface to guide the user to adjust the operation to achieve the best effect; Step 4: The feedback evaluation module is used to record the feedback provided by the system during each operation and the actual response of the user, including the operation data after the user corrects the behavior; Regularly evaluate the quality of feedback provided by the system and the improvement of user operations, and measure the effect through quantitative indicators; Based on the new round of feedback data collected, the model is iteratively trained and tuned to continuously improve its recognition accuracy and feedback effectiveness, forming a closed-loop continuous optimization mechanism.

[0024] Example 2 like Figure 1-Figure 2 As shown, based on Example 1, a medical expert performs a standard puncture operation in a simulated environment, and simultaneously records video data and sensor data of the entire process; Collect data on puncture operations in real medical scenarios while ensuring patient safety and privacy; The collected data were integrated into a dataset that included cases of varying difficulty levels, different variations in patient anatomy, and examples of correct and incorrect procedures; The video is annotated frame by frame to mark the puncture path, puncture point, and the location and time data of important events (successful puncture, incorrect puncture, and puncture offset) for preprocessing. Specifically: The path coordinates, time series and event labels of all operations are sorted into corresponding data structures, and the coordinates of all puncture paths and puncture points are mapped to a unified range [0, 1]; Set the operation start time to 0 and the end time to 1, and convert the time data corresponding to all operations into the time percentage relative to the start of the operation, so that the duration of all operations is mapped to the same scale, that is, between [0, 1]; Each event of successful puncture, wrong puncture and puncture deviation is converted into a vector, where the position of the corresponding event is 1 and the rest are 0; The processed puncture path and puncture point data are constructed into an array of shape (N, M, D), where N is the number of puncture operation samples, each sample has M time point coordinate records, and D is the coordinate dimension (usually 2 or 3, corresponding to two-dimensional or three-dimensional space); The processed time data is constructed into an array of shape (N, M), where each row represents the operation time series of a sample; The encoding of each event is formed into a matrix of shape (N, C), where C is the number of event types and each row represents the event sequence of a sample during the operation; Align the above processed data according to the operation samples and integrate them into a complete visual data set; Obtain force data (force magnitude and direction) from the sensor and align the force data with the visual data set according to the operation samples; Integrate the visual data set and the corresponding force data into a complete data set, and divide it into training set, validation set and test set; Build a multimodal fusion model, a model architecture consisting of a convolutional neural network (CNN) + a recurrent neural network (RNN), and use the training set to train the model; Among them, the convolutional neural network algorithm processes visual image data and extracts spatial features; the recurrent neural network algorithm processes time series data, namely force feedback data, to capture the time dependency in the action sequence; Specific: The ResNet (residual network) model in the convolutional neural network (CNN) was selected as the visual pre-training model. By inputting visual data, the model was trained to extract spatial features (including puncture point location, path, angle, and subtle changes in the skin surface) from the visual records of the puncture process. These spatial features can be used for precise positioning and path planning. The recurrent neural network (RNN) model is used as a force pre-training model. By inputting force signals, the output is a set of vectors representing force characteristics. These feature vectors can reveal key information during the puncture process (including force, speed, rhythm, and coordination). After pre-training the above model through the training set, the validation set is used to optimize the model parameters to find the optimal parameter configuration, and the test set is used to evaluate the accuracy and precision of the model.

[0025] Example 3 like Figure 1-Figure 3 As shown, based on Example 1, the acquired original time data is processed into time series data points, and all time series data points are integrated into a set { }, through the formula Calculate the movement speed of the operator's hand or puncture needle when working ,in represents the measured value at time t, is the sampling interval between two adjacent time points; Based on the speed data, the formula Calculate the acceleration of the operator's hand or puncture needle when working ; The movement speed and acceleration data of the operator's hand or puncture needle are input into the model to analyze the operator's operation consistency and control precision; The maximum force during puncture is identified in the acquired force feedback data, and the average force value is obtained by using the mean calculation formula according to the corresponding time series; The maximum force and average force during puncture are input into the model to analyze whether the operator has applied excessive or insufficient force, as well as the overall stability of the operator's force application; Obtain the curve smoothness, linearity, and curvature data of the puncture path during puncture by the operator, and input these data into the model to analyze the accuracy of the operator's operation and the conformity of the path; Obtain the time consumed to complete a specific operation phase (including positioning, needle insertion, and adjustment) and input it into the model to analyze the operator's efficiency and rhythm control during the operation; Obtain information about specific events (puncture success, puncture failure, or the moment when correction is required), process it into event vectors and input them into the model to analyze the characteristics of sudden force changes and speed changes to automatically mark these key points; The processed data is used as input to the trained model for real-time analysis to predict the accuracy and effectiveness of the operator's current operation and determine whether there are deviations or errors. Specifically: The set of actual features obtained after real-time data processing is recorded as X=( ), where each Corresponding to a feature, a weight is preset for each feature, denoted as θ=( ,..., ), Indicates the prediction tendency in the absence of input features; By formula To calculate the probability that the target variable y is equal to 1 when the input feature x is given, and get the predicted output of the model ;when When it is greater than the standard threshold, predict y=1; otherwise, predict y=0; when y=1, it means the current operation is correct, and when y=0, it means the current operation is wrong; By formula Get the deviation between the actual operation and the standard state ,in Indicated in the feature The deviation value is 0, and a set of influence coefficients are preset ( , ,…, ) to correspond to the impact of each feature on the deviation value; Based on the "correct" or "wrong" result of the actual operation and the deviation value between the actual operation and the target operation, the result information of whether there is deviation and error in the operator's current operation is generated.

[0026] Example 4 like Figure 1-Figure 3 As shown, based on Example 1, real-time feedback is generated based on whether there are deviations and erroneous result information during the operator's current operation, and visual, auditory, and tactile feedback mechanisms are integrated on the AR interface to synchronize with the user's operation in real time. Specifically: Based on the judgment results of the model, the system provides intuitive visual feedback through the AR interface (color coding: green - correct, red - wrong; arrow indication: indicating the correct position, path and angle), auditory prompts (voice guidance: providing immediate verbal correction suggestions when errors are found, sound reminders: different error types correspond to different sound effects, and different sound effects are used to quickly provide feedback to the operator) and tactile feedback (using the vibration of force feedback gloves to remind the user when the operator makes an error or deviates from the standard, and simulate the correct force or direction sense), guiding the user to adjust the operation to achieve the best effect.

[0027] Example 5 Based on Example 1, the feedback quality provided by the system and the improvement of the user's operation are regularly evaluated, and the effect is measured by quantitative indicators, specifically: Automatically record all interaction data during each operation, including visual, auditory, and tactile feedback information provided by the AR interface, as well as the user's actual response to these feedbacks, including the magnitude, direction, and speed of the user's adjustment operation; Use sensors and user interface interaction records to track the user's operational corrections after receiving feedback, including the time, method, and effect of the corrections; Record the final results of each operation, including success or failure, time taken, deviations and error messages during the operation; Compare the operator's operation data before receiving feedback with the operation data after receiving feedback. If the deviation and error information in the operation data after receiving feedback are less than the operation data before receiving feedback, it means that the feedback has brought about obvious operation improvement; if the deviation and error information in the operation data after receiving feedback is greater than or equal to the operation data before receiving feedback, it means that the feedback has not brought about operation improvement; Get the operator's complete and correct operation time for each time, build a plane coordinate system, set the operator's number of times as the x-axis, and the operation time as the y-axis, substitute the number of times and the operation time into the coordinate system, and analyze to get the trend line of the operation time changing with the number of times. By observing the slope of the trend line, if the trend line tilts downward, it means that the feedback has played a continuous role in promoting the operator's skill proficiency.

[0028] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An AR-guided infusion port puncture navigation system, characterized by: Step 1: Collect a large amount of historical puncture operation data, including data on successful and failed cases, extract key action feature data from the data, annotate the data in detail, distinguish correct operation steps from potential errors, and provide "right" and "wrong" labels for model learning; Use deep learning algorithms to build models and input labeled data for training, so that the models can identify correct and incorrect operation modes and optimize accuracy and efficiency evaluation standards; Step 2: Capture the operator's working information (including perspective information, motion information, hand motion trajectory, gesture changes, puncture force, pressure distribution and spatial coordinate data), quickly process the captured real-time data, obtain corresponding feature data, and use the model to analyze these feature data in real time; Step 3: Integrate visual, auditory, and tactile feedback mechanisms on the AR interface to synchronize with the user's operations in real time; Input the results of real-time analysis into the trained model, allowing the model to determine the accuracy of the current operation in real time and identify whether deviations or errors occur; Based on the judgment results of the model, the system provides intuitive visual feedback (color coding: green - correct, red - wrong; arrow indication: indicating the correct position and angle), auditory prompts (voice guidance: guiding the user in every step of the operation, especially providing immediate verbal correction suggestions when errors are found) and tactile feedback (using the vibration of force feedback gloves) through the AR interface to guide the user to adjust the operation to achieve the best effect; Step 4: Record the feedback provided by the system during each operation and the actual response of the user, including the operation data after the user corrects the behavior; Regularly evaluate the quality of feedback provided by the system and the improvement of user operations, and measure the effect through quantitative indicators; Based on the new round of feedback data collected, the model is iteratively trained and tuned to continuously improve its recognition accuracy and feedback effectiveness, forming a closed-loop continuous optimization mechanism.

2. The AR-guided infusion port puncture navigation system according to claim 1, characterized in that: The key action feature data include operation path and puncture point coordinates, time series, important events (successful puncture, incorrect puncture and puncture deviation) and force magnitude and direction.

3. The AR-guided infusion port puncture navigation system according to claim 2, characterized in that: Map the coordinates of all puncture paths and puncture points to a uniform range of [0, 1]; Set the start time of the operation to 0 and the end time to 1, and convert the time data corresponding to all operations into the percentage of time relative to the start of the operation, so that the duration of all operations is mapped to the same scale, that is, between [0, 1]; Each event of successful puncture, incorrect puncture, and puncture deviation is converted into a vector, in which the position of the corresponding event is 1 and the rest are 0; Align the processed data according to the operation samples and integrate them into a complete visual data set; Obtain force data (force magnitude and direction) from the sensor and align the force data with the visual data according to the operation sample.

4. The AR-guided infusion port puncture navigation system according to claim 3, characterized in that: The processed puncture path and puncture point data are constructed into an array of shape (N, M, D), where N is the number of puncture operation samples, each sample has M time point coordinate records, and D is the coordinate dimension (usually 2 or 3, corresponding to two-dimensional or three-dimensional space); The processed time data is constructed into an array of shape (N, M), where each row represents the operation time series of a sample; The encoding of each event is formed into a matrix of shape (N, C), where C is the number of event types and each row represents the event sequence of a sample during the operation.

5. The AR-guided infusion port puncture navigation system according to claim 1, characterized in that: The model is a multimodal fusion model, which has a model architecture consisting of a convolutional neural network (CNN) + a recurrent neural network (RNN), and the training set is used to train the model; Among them, the convolutional neural network algorithm processes visual image data and extracts spatial features; the recurrent neural network algorithm processes time series data, namely force feedback data, to capture the time dependency in the action sequence.

6. The AR-guided infusion port puncture navigation system according to claim 5, characterized in that: The model selects the ResNet (residual network) model in the convolutional neural network (CNN) as a visual pre-training model, and uses the input visual data to train the model to extract spatial features (including the location of the puncture point, path, angle, and subtle changes in the skin surface) from the visual record of the puncture process for precise positioning and path planning; A recurrent neural network (RNN) model is used as a force pre-training model. By inputting force signals, the model outputs a set of vectors representing force characteristics, which are used to reveal key information during the puncture process (including force, speed, rhythm, and coordination).

7. The AR-guided infusion port puncture navigation system according to claim 1, characterized in that: The acquired raw time data is processed into time series data points to obtain the movement speed of the operator's hand or puncture needle when working; Based on the velocity data, the acceleration of the operator's hand or the puncture needle when working is obtained; The maximum force during puncture is identified in the acquired force feedback data, and the average force value is obtained according to the corresponding time series.

8. The AR-guided infusion port puncture navigation system according to claim 1, characterized in that: The processed data is used as input to the trained model for real-time analysis to predict the accuracy and effectiveness of the operator's current operation and determine whether there is any deviation or error. Specifically: After real-time data is processed, the prediction value and deviation value are obtained through the model. When the prediction value is greater than the standard threshold, y=1 is predicted; otherwise, y=0 is predicted. When y=1, it indicates that the current operation is correct, and when y=0, it indicates that the current operation is wrong. Based on the "correct" or "wrong" result of the actual operation and the deviation value between the actual operation and the target operation, the result information of whether there is deviation and error in the operator's current operation is generated.

9. The AR-guided infusion port puncture navigation system according to claim 1, characterized in that: Based on the model's judgment results, the AR interface provides intuitive visual feedback, auditory prompts, and tactile feedback to guide users to adjust operations to achieve the best results.

10. The AR-guided infusion port puncture navigation system according to claim 1, characterized in that: Record all interaction data during each operation, use the interaction records of sensors and user interfaces to track the user's operation corrections after receiving feedback, and record the final results of each operation; Compare the operator's operation data before receiving feedback with the operation data after receiving feedback. If the deviation and error information in the operation data after receiving feedback are less than the operation data before receiving feedback, it means that the feedback has brought about obvious operation improvement; if the deviation and error information in the operation data after receiving feedback is greater than or equal to the operation data before receiving feedback, it means that the feedback has not brought about operation improvement; Get the operator's complete and correct operation time for each time, build a plane coordinate system, set the operator's number of times as the x-axis, and the operation time as the y-axis, substitute the number of times and the operation time into the coordinate system, and analyze to get the trend line of the operation time changing with the number of times. By observing the slope of the trend line, if the trend line tilts downward, it means that the feedback has played a continuous role in promoting the operator's skill proficiency.

Images