Ultrasound-guided nerve block anesthesia puncture method based on artificial intelligence

Abstract

The invention provides an ultrasound-guided nerve block anesthesia puncture method based on artificial intelligence, and the method comprises the steps: S11, collecting a target nerve block image through an ultrasonic probe, and positioning a to-be-punctured point in the target nerve block image; s12, identifying tissues except the point to be punctured in the target nerve block image, marking the tissues as dangerous areas, and planning a path from the skin to the point to be punctured by taking avoidance of the dangerous areas as a principle; s13, electromyographic signals of the arms of the operator are collected in real time, and an intention prediction model between the electromyographic signals and hand actions is established; and S14, determining a space coordinate system position of a puncture needle point based on the to-be-punctured point, associating the puncture needle point with the intention prediction model, monitoring whether the puncture needle point punctures along the planned path, and if not, sending out an early warning signal. The precision, the safety, the efficiency and the reliability of puncture are improved, the operation difficulty and the complication risk are reduced, and meanwhile, the development of personalized medical treatment is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and in particular, to an ultrasound-guided nerve block anesthesia puncture method based on artificial intelligence. Background Art

[0002] As a visualization technology, ultrasound technology has developed rapidly in recent years. The ultrasound visualization technology combines the dual advantages of non-invasiveness and visibility, can display the anatomical structure of local tissues, has a good guiding effect on the direction and depth of the anesthesia puncture needle insertion, and reduces the occurrence of anesthesia complications. The ultrasound-guided nerve block can reduce the blindness of the operation and improve the success rate, and is currently widely recognized. The ultrasound-guided nerve block can be used in different parts of the human body, including the neck, trunk, limbs, etc., and displays the images of the nerves and accompanying blood vessels in the above parts, providing a prerequisite for the ultrasound-guided nerve block.

[0003] The clarity and resolution of ultrasound images directly affect the operator's ability to identify nerves and their surrounding tissues. At present, although ultrasound technology has been able to clearly display nerves and surrounding structures, traditional puncture methods highly rely on doctors' experience and surface anatomical landmarks, lack precise image guidance, resulting in insufficient puncture accuracy, and are prone to accidentally injure important surrounding tissues (such as blood vessels, nerves, etc.), increasing the risk of postoperative complications. Secondly, the traditional method cannot real-time monitor the path and position of the puncture needle, and it is difficult to dynamically adjust the puncture path to adapt to complex anatomical structures, significantly increasing the puncture difficulty and risk. In addition, there is a lack of real-time feedback and warning mechanism for the doctor's operation intention during the puncture process. Once the puncture needle deviates from the predetermined path, it is difficult to correct it in time, further reducing the safety and success rate of the operation. Finally, the traditional method has high requirements for doctors' experience, and it is difficult for novice doctors or non-professional operators to quickly master, resulting in a long learning curve and restricting the wide application of nerve block technology. Summary of the Invention

[0004] The purpose of the present invention is to provide an ultrasound-guided nerve block anesthesia puncture method based on artificial intelligence, which can accurately locate the target puncture point and other surrounding tissue cells, generate a puncture planning path from the skin surface to the target nerve, and improve the puncture accuracy.

[0005] To achieve the above invention purpose, the present invention provides an ultrasound-guided nerve block anesthesia puncture method based on artificial intelligence, and the method includes:

[0006] S11. Collect the target nerve block image through an ultrasound probe, and locate the puncture point in the target nerve block image;

[0007] S12. Identify the tissues in the target nerve block image other than the puncture point to be punctured, including but not limited to blood vessels, fascia, and nerves, mark the tissues that need to be avoided as dangerous areas, and plan the path from the skin to the puncture point to be punctured based on the principle of avoiding dangerous areas, so as to obtain the planned path from the skin to the puncture point to be punctured;

[0008] S13. Collect the electromyogram signals of the operator's arm in real time, preprocess and extract the features of the electromyogram signals, and train the neural network through the extracted features to establish an intention prediction model between the electromyogram signals and the hand movements;

[0009] S14. Determine the spatial coordinate system position of the puncture needle tip based on the puncture point to be punctured, monitor the real-time trajectory of the puncture needle in the intraoperative target nerve block image, associate the puncture needle tip with the intention prediction model, and monitor whether the puncture needle tip punctures along the planned path. If the distance between the needle tip trajectory and the planned path exceeds the preset threshold, an early warning signal is issued.

[0010] Further, in step S11, locating the puncture point to be punctured in the target nerve block image specifically includes:

[0011] S21. Use a high-order partial differential equation with an adaptive adjustment coefficient to remove the image noise;

[0012] S22. Use the superposition integral transform to visually enhance the tissues in the image;

[0013] S23. Use a network structure including local and global information feedback to locate the puncture point to be punctured, and mark the located puncture point to be punctured with color values;

[0014] S24. Design a dynamic adjustment model depending on the image gradient and local brightness characteristics for the image contrast differences of different patients.

[0015] Further, in step S12, marking the tissues that need to be avoided as dangerous areas specifically includes:

[0016] S31. Perform three-dimensional reconstruction on the target nerve block image to establish a three-dimensional model with tissues as units;

[0017] S32. Based on the three-dimensional model of the tissues, determine the recognition units through segmentation or clustering methods, traverse each recognition unit to extract feature points, and construct a feature point matrix, where the recognition units are composed of tissues with the same attributes;

[0018] S33. Use machine learning to construct an avoidance recognition model for the recognition units that need to be marked and avoided;

[0019] S34. Assign the same marked color value to the feature point matrix corresponding to the recognition unit that needs to be avoided through the avoidance recognition model, and the tissue corresponding to the recognition unit assigned the same marked color value is the dangerous area.

[0020] Further, in step S12, the path from the skin to the puncture point is planned on the principle of avoiding the dangerous area, specifically including:

[0021] S41. Determine the puncture area within the preset angle range on the skin surface corresponding to the target nerve block image of the puncture point to be;

[0022] S42. Obtain the dangerous area and the puncturable area within the puncture area, calculate the risk distance that needs to avoid the dangerous area, and obtain the optional path from the skin to the puncturable area corresponding to the puncture point to be;

[0023] S43. Screen the optional path through the reinforcement algorithm with the minimum puncture depth and the maximum risk distance as the optimization objectives to obtain at least three optional planned paths.

[0024] Further, in step S13, the electromyographic signal is preprocessed and feature extracted, specifically including:

[0025] Filter and denoise the collected electromyographic signal;

[0026] Extract features from the preprocessed electromyographic signal, and the features include time domain features, frequency domain features, and time-frequency domain features.

[0027] Further, in step S13, the neural network is trained through the extracted features to establish an intention prediction model between the electromyographic signal and the hand movement, specifically including:

[0028] S51. Normalize the obtained time domain features, frequency domain features, and time-frequency domain features and then label them to obtain a label data set;

[0029] S52. Train the label data set through the neural network, calculate the gap between the predicted value and the true value through the mean square error, and update the network weights through the optimization algorithm SGD;

[0030] S53. Use the training data set to iteratively train the convolutional neural network, continuously adjust the weights and reduce the loss function value;

[0031] S54. Evaluate the performance of the trained convolutional neural network through the validation data set and perform hyperparameter tuning and optimization. After verifying that the generalization ability of the convolutional neural network meets the standard using the test data set, output the intention prediction model between the electromyographic signal and the hand movement.

[0032] Further, in step S14, determining the spatial coordinate system position of the puncture needle tip based on the puncture point to be punctured specifically includes:

[0033] Establishing a three-dimensional coordinate system based on the puncture point to be punctured, obtaining the relative positions of each pixel point in the target nerve block image with respect to the three-dimensional coordinate system, and positioning the puncture needle tip based on the relative positions.

[0034] Further, in step S14, monitoring the real-time trajectory of the puncture needle in the target nerve block image during the operation specifically includes:

[0035] S61. Identifying the puncture needle in the target nerve block image based on a neural network and performing feature enhancement processing;

[0036] S62. Obtaining the edge contour of the puncture needle body through a contour recognition algorithm and marking the color value of the edge contour;

[0037] S63. Superimposing the puncture needle after edge contour recognition on the target nerve block image in real time for display.

[0038] Compared with the prior art, the beneficial effects of the present invention are:

[0039] An ultrasound-guided nerve block anesthesia puncture method based on artificial intelligence provided by the present invention accurately locates the acquired target nerve block image to obtain the target puncture point to be punctured. At the same time, by identifying and marking dangerous areas such as blood vessels and nerves, and planning the puncture path on the principle of avoiding these areas, the risks of nerve injury, blood vessel injury, etc. caused by mis-puncture are reduced. Through electromyogram signal monitoring and an intention prediction model, the action intention of the operator is fed back in real time, thereby monitoring the real-time trajectory of the puncture needle and comparing it with the planned path. Once the needle tip trajectory deviates from the planned path, a warning signal is sent to remind the operator to adjust in time, enhancing the safety of the operation. The present invention improves the accuracy and real-time performance of processing the target nerve block image, so that the operator can guide the puncture needle to reach the target puncture area in real time and accurately during the operation, ensuring the success of anesthesia and the safety of the patient. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for description in the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0041] Figure 1 An ultrasound-guided nerve block anesthesia puncture method based on artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present invention, rather than limiting the present invention. Additionally, it should be noted that for the convenience of description, only the parts related to the present invention rather than all the structures are shown in the accompanying drawings.

[0043] Referring to Figure 1 , this embodiment provides an ultrasound-guided nerve block anesthesia puncture method based on artificial intelligence, and the method includes:

[0044] S11. Collect the target nerve block image through an ultrasound probe, and locate the puncture point to be punctured in the target nerve block image. In this embodiment, the target nerve block image - femoral nerve of the groin area of an obese patient is collected through an ultrasound probe, that is, the ultrasound probe is placed at a position where the femoral artery or femoral nerve can be displayed, and the puncture point to be punctured is located through the collected target nerve block image. The puncture point to be punctured can be given a color mark to assist the operator in finding the nerve block point that needs to be anesthetized.

[0045] In step S11, locating the puncture point to be punctured in the target nerve block image specifically includes:

[0046] S21. Remove the image noise by using a high-order partial differential equation with an adaptive adjustment coefficient.

[0047] S22. Perform visual enhancement on the tissues in the image by using a superposition integral transform.

[0048] S23. Locate the puncture point to be punctured by using a network structure including local and global information feedback, and perform color value marking on the located puncture point to be punctured.

[0049] S24. Design a dynamic adjustment model depending on the image gradient and local brightness characteristics for the image contrast difference of different patients.

[0050] In this embodiment, in ultrasound-guided nerve block, since ultrasound images are often interfered by noise, which affects the operator's recognition of nerves, using a high-order partial differential equation with an adaptive adjustment coefficient can effectively remove the noise in the image, while retaining the details and structures in the image, making the boundaries of nerves and blood vessels clearer. The visibility of nerves and blood vessels in the ultrasound image is further improved through a superposition integral transform, and visual enhancement is performed on the tissues in the image, making the key structures more prominent in the image, improving the contrast and clarity of the image, helping the operator to more clearly observe the boundary between the nerve and the surrounding tissues, and improving the clarity of nerves and blood vessels in the image, so as to more accurately locate the puncture point to be punctured.

[0051] The deep neural network extracts local and global features from the enhanced ultrasound images. Each layer of the deep neural network is used to extract features of different scales and complexities, including edges, textures, and high-level semantic information. By combining local details and global information, it can more accurately determine the puncture point location. At the same time, the color value marking can be intuitively displayed to the operator, reducing operation errors. Due to the subtle differences in physiological structures among different patients, such as the anatomical positions of nerves, the density and distribution of surrounding tissues, etc., these differences will lead to significant differences in the contrast and brightness of the images. By dynamically adjusting the model, the contrast and brightness of the images can be optimized according to the specific image features of each patient, so as to more clearly display the target nerve block image and the puncture point to be punctured.

[0052] S12. Identify the tissues in the target nerve block image other than the puncture point to be punctured, including but not limited to blood vessels, fascia, and nerves. Mark the tissues that need to be avoided as dangerous areas. Plan the path from the skin to the puncture point to be punctured based on the principle of avoiding dangerous areas, and obtain the planned path from the skin to the puncture point. In this embodiment, by identifying and marking important tissues such as blood vessels and nerves as dangerous areas, the path planning can avoid these areas, significantly reducing the risk of damaging important tissues during the puncture process. Precise avoidance of dangerous areas can effectively reduce complications such as bleeding and nerve injury caused by mis-puncture, and improve the safety of the operation.

[0053] In step S12, marking the tissues that need to be avoided as dangerous areas specifically includes:

[0054] S31. Perform three-dimensional reconstruction on the target nerve block image to establish a three-dimensional model with tissues as units.

[0055] S32. Based on the three-dimensional model of the tissues, determine the recognition units by segmentation or clustering methods, traverse each recognition unit to extract feature points, and construct a feature point matrix. The recognition units are composed of tissues with the same attributes.

[0056] S33. Construct an avoidance recognition model through machine learning for the recognition units that need to be avoided and have been marked.

[0057] S34. Assign the same marked color value to the feature point matrix corresponding to the recognition units that need to be avoided through the avoidance recognition model. The tissues corresponding to the recognition units with the same marked color value are dangerous areas.

[0058] In this embodiment, a model based on tissues is established through three-dimensional reconstruction, which can more intuitively display the spatial relationships of tissues such as nerves, blood vessels, and fascia. By using segmentation or clustering methods to determine the recognition units and extracting feature points to construct a matrix, the details of the target nerve and its surrounding tissues can be more clearly shown. During nerve block surgery, the tissues around the target nerve are complex and diverse, and it is difficult to accurately identify and distinguish these tissues by traditional methods. A large number of image data are learned by a machine learning model to automatically identify and mark the dangerous areas that need to be avoided. At the same time, there are slight differences in the anatomical structures of different patients, and the machine learning model can dynamically adjust the recognition strategy according to the specific image features of each patient, so as to better adapt to individual differences. By assigning the same marker color value, the recognition units identified by the avoidance recognition model as needing to be avoided can be prominently displayed in the image, enabling the operator to quickly identify and avoid these dangerous areas during the operation.

[0059] In step S12, the path from the skin to the puncture point to be punctured is planned based on the principle of avoiding dangerous areas, which specifically includes:

[0060] S41. Determine the puncture area within the preset angle range on the skin surface corresponding to the target nerve block image from the puncture point to be punctured.

[0061] S42. Obtain the dangerous areas and the puncturable areas within the puncture area, calculate the risk distance for avoiding the dangerous areas, and obtain the optional paths from the skin to the puncturable areas corresponding to the puncture point to be punctured.

[0062] S43. Screen the optional paths through a reinforcement algorithm with the minimum puncture depth and the maximum risk distance as the optimization objectives to obtain at least three optional planned paths.

[0063] In this embodiment, during nerve block anesthesia, the puncture needle needs to avoid blood vessels, nerves, and other important tissues to prevent complications caused by accidental puncture, such as bleeding, nerve injury, etc. At the same time, the position of the nerve in the human body is not completely fixed, and there are tissues such as blood vessels and muscles around it. By determining the puncture area within the preset angle range on the skin surface corresponding to the target nerve block image from the puncture point to be punctured, a suitable angle and skin position for puncture can be located.

[0064] During nerve block operations, the puncture needle may accidentally damage nerves, blood vessels, or other important structures, such as blood vessel injury, hematoma formation, nerve injury, etc. By identifying dangerous areas and calculating risk distances, these areas can be effectively avoided, reducing the occurrence of complications. For example, in brachial plexus nerve block, the relationship between the nerve and blood vessel is close, and the puncture needle may accidentally damage the blood vessel, leading to hematoma. By using ultrasound guidance to identify the relative positions of the blood vessel and nerve, the puncture needle can be prevented from directly damaging the blood vessel. Reinforcement learning starts from the initial state (skin surface), gradually selects the optimal actions, generates multiple optional paths, and takes the minimum puncture depth and the maximum risk distance as the optimization objectives, reducing the moving distance of the puncture needle in the tissue, reducing the risk of tissue injury, and at the same time reducing the pain and discomfort of the patient.

[0065] S13. Real-time collect the electromyographic signals of the operator's arm, preprocess and extract features from the electromyographic signals, and train the neural network through the extracted features to establish an intention prediction model between the electromyographic signals and the hand movements. In this embodiment, the electromyographic signals can be collected by attaching surface electrodes to the muscle surfaces of the operator's forearm and hand, or all methods capable of collecting electromyographic signals in the prior art can be applicable, and the present invention does not limit this. The electromyographic signals can directly reflect the brain's control intention of the muscles. The generation of the electromyographic signals precedes limb movement. By establishing an intention prediction model, the trajectory and direction of the operator's hand movements can be predicted in advance, so as to more accurately predict the path of the puncture needle, reduce the puncture deviation caused by small hand tremors or misoperations, and ensure that the puncture needle accurately reaches the target position according to the preset path.

[0066] In step S13, preprocessing and feature extraction of the electromyographic signals specifically include:

[0067] Filter and denoise the collected electromyographic signals.

[0068] Extract features from the preprocessed electromyographic signals, and the features include time-domain features, frequency-domain features, and time-frequency domain features.

[0069] In this embodiment, the noise outside the frequency band of the electromyographic signals, such as high-frequency noise and low-frequency baseline drift, is removed by filtering, and the noise within the frequency band is removed by combining methods such as wavelet transform, which can significantly improve the signal-to-noise ratio of the electromyographic signals and make the signals clearer. The extraction of time-domain features includes mean, variance, root mean square value, etc., and these features reflect the intensity and pattern of muscle activity. The extraction of frequency-domain features calculates the power spectral density of the electromyographic signals through fast Fourier transform and analyzes its frequency distribution. For example, the spectrum of the electromyographic signals during isometric contraction is mainly concentrated in the low-frequency band, while during isotonic contraction, it is concentrated in the high-frequency band. The extraction of time-frequency domain features uses wavelet transform to capture the time-domain and frequency-domain information of the signals and describe the characteristics of the electromyographic signals.

[0070] In step S13, the neural network is trained with the extracted features to establish an intention prediction model between the EMG signals and the hand movements, which specifically includes:

[0071] S51. Normalize the obtained time-domain features, frequency-domain features, and time-frequency domain features and then label them to obtain a labeled dataset.

[0072] S52. Train the labeled dataset through the neural network, calculate the gap between the predicted value and the true value through the mean square error, and update the network weights through the optimization algorithm SGD.

[0073] S53. Use the training dataset to iteratively train the convolutional neural network, continuously adjust the weights and reduce the loss function value.

[0074] S54. Evaluate the performance of the trained convolutional neural network through the validation dataset and perform hyperparameter adjustment and optimization. After verifying that the generalization ability of the convolutional neural network meets the standard using the test dataset, output the intention prediction model between the EMG signals and the hand movements.

[0075] In this embodiment, the normalized time-domain features, frequency-domain features, and time-frequency domain features are labeled, that is, the normalized time-domain features, frequency-domain features, and time-frequency domain features are associated with the corresponding hand movement labels. For example, the hand movement label is "clenching the fist". The labeled dataset is divided into a training set, a validation set, and a test set. The training set is used for model training, the validation set is used for hyperparameter adjustment, and the test set is used for evaluating the generalization ability of the model. The convolutional neural network is used as the model architecture, the predicted value is calculated through forward propagation, and the mean square error is used as the loss function to measure the gap between the predicted value and the true value. The network weights are updated using stochastic gradient descent, and the weights are continuously adjusted through iterative training to reduce the loss function value. The performance of the model is evaluated using the validation dataset, and the hyperparameters are adjusted according to the loss value and accuracy of the validation set; the generalization ability of the model is verified using the test dataset to ensure that the model can accurately predict on unseen data, and finally the optimized intention prediction model is output.

[0076] S14. Determine the position of the puncture needle tip in the spatial coordinate system based on the puncture point to be punctured, monitor the real-time trajectory of the puncture needle in the intraoperative target nerve block image, associate the puncture needle tip with the intention prediction model, and monitor whether the puncture needle tip punctures along the planned path. If the distance between the needle tip trajectory and the planned path exceeds the preset threshold, a warning signal is issued. In this embodiment, by monitoring the trajectory of the puncture needle tip in real time and comparing it with the planned path, it is ensured that the puncture needle advances accurately along the predetermined path, reducing puncture deviation caused by hand tremors, operation errors, or complex anatomical structures. When the needle tip trajectory deviates from the planned path by more than the preset threshold, a warning signal is issued in a timely manner. For example, a vibration warning is triggered when the deviation is greater than 5 degrees to remind the operator to adjust the operation path, avoid accidental injury to nerves, blood vessels, or other important tissues by the puncture needle, and reduce the surgical risk.

[0077] In step S14, determining the position of the puncture needle tip in the spatial coordinate system based on the puncture point to be punctured specifically includes:

[0078] Establish a three-dimensional coordinate system based on the puncture point to be punctured, obtain the relative positions of each pixel point in the target nerve block image with respect to the three-dimensional coordinate system, and position the puncture needle tip based on the relative positions.

[0079] In this embodiment, the puncture point to be punctured is used as the origin of the three-dimensional coordinate system. According to the anatomical structure or image features of the surgical area, three appropriate orthogonal directions are selected as the X, Y, and Z axes. For example, the natural direction of the anatomical structure (such as the direction of a nerve or blood vessel) is selected to define the coordinate axes. The coordinate system is defined by the marked feature points in the image and two other non-collinear points. By calculating the relative position relationship between them, the unit vectors of the coordinate axes are determined. The pixel point coordinates in the target nerve block image are converted into three-dimensional coordinates in the world coordinate system through camera calibration or known geometric relationships, and the relative positions of each pixel point with respect to the puncture point to be punctured are calculated, including distance and direction. During the operation, the position information of the puncture needle tip is transmitted back in real time by the ultrasonic probe, mapped to the established three-dimensional coordinate system, and its relative position with respect to the origin is calculated. The real-time position of the puncture needle tip is compared with the planned path to determine whether it deviates from the predetermined trajectory.

[0080] In step S14, monitoring the real-time trajectory of the puncture needle in the intraoperative target nerve block image specifically includes:

[0081] S61. Identify the puncture needle in the target nerve block image based on a neural network and perform feature enhancement processing.

[0082] S62. Obtain the edge contour of the puncture needle body through a contour recognition algorithm and mark the color value of the edge contour.

[0083] S63. Display the puncture needle after edge contour recognition in real time by superimposing it on the target nerve block image.

[0084] In this embodiment, the neural network is used to identify and enhance the features of the puncture needle in the target nerve block image, making the puncture needle clearer in the image, reducing the misidentification caused by noise or artifacts, and improving the recognition accuracy of the puncture needle. The edge contour of the puncture needle body is obtained through the contour recognition algorithm, and the extracted edge contour is marked with color values to more intuitively display the position of the puncture needle in the image. The puncture needle after edge contour recognition is superimposed on the target nerve block image in real time for real-time display, helping the operator to more accurately control the path of the puncture needle, ensuring that it advances along the preset path, and avoiding accidentally injuring nerves, blood vessels or other important tissues.

[0085] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. An ultrasound-guided nerve block anesthesia puncture method based on artificial intelligence, characterized in that The method includes: S11. Collecting a target nerve block image through an ultrasound probe and locating a puncture point to be punctured in the target nerve block image; S12. Identifying tissues in the target nerve block image other than the puncture point to be punctured, including but not limited to blood vessels, fascia, and nerves, marking the tissues that need to be avoided as dangerous areas, and planning the path from the skin to the puncture point to be punctured based on the principle of avoiding dangerous areas to obtain a planned path from the skin to the puncture point to be punctured; S13. Real-time collecting the electromyogram signals of the operator's arm, preprocessing and feature extracting the electromyogram signals, and training a neural network through the extracted features to establish an intention prediction model between the electromyogram signals and hand movements; S14. Determining the spatial coordinate system position of the puncture needle tip based on the puncture point to be punctured, monitoring the real-time trajectory of the puncture needle in the intraoperative target nerve block image, associating the puncture needle tip with the intention prediction model, monitoring whether the puncture needle tip punctures along the planned path, and if the distance between the needle tip trajectory and the planned path exceeds a preset threshold, an early warning signal is issued.

2. The ultrasound-guided nerve block anesthesia puncture method based on artificial intelligence according to claim 1, wherein In step S11, locating the puncture point to be punctured in the target nerve block image specifically includes: S21. Removing image noise by using a high-order partial differential equation with an adaptive adjustment coefficient; S22. Using a superposition integral transform to visually enhance the tissues in the image; S23. Locating the puncture point to be punctured by using a network structure containing local and global information feedback, and color-coding the located puncture point to be punctured; S24. Designing a dynamic adjustment model depending on the image gradient and local brightness characteristics for the image contrast differences of different patients.

3. The method for ultrasound-guided nerve block anesthesia puncture based on artificial intelligence according to claim 1, wherein In step S12, marking the tissues that need to be avoided as dangerous areas specifically includes: S31. Performing three-dimensional reconstruction on the target nerve block image to establish a three-dimensional model with tissues as units; S32. Based on the three-dimensional model of the tissues, determining recognition units through segmentation or clustering methods, traversing each recognition unit to extract feature points, and constructing a feature point matrix, where the recognition units are composed of tissues with the same attributes; S33. Constructing an avoidance recognition model for the recognition units marked to be avoided through machine learning; S34. Assigning the same marked color value to the feature point matrix corresponding to the recognition units that need to be avoided through the avoidance recognition model, and the tissues corresponding to the recognition units assigned the same marked color value are dangerous areas.

4. The method for ultrasound-guided nerve block anesthesia puncture based on artificial intelligence according to claim 1, characterized in that In step S12, planning the path from the skin to the puncture point to be punctured based on the principle of avoiding dangerous areas specifically includes: S41. Determining a puncture area within a preset angle range on the skin surface corresponding to the puncture point to be punctured in the target nerve block image; S42. Obtaining the dangerous areas and puncturable areas within the puncture area, calculating the risk distance for avoiding the dangerous areas, and obtaining the optional paths from the skin to the puncturable areas corresponding to the puncture point to be punctured; S43. Screening the optional paths through a reinforcement algorithm with the minimum puncture depth and the maximum risk distance as optimization objectives to obtain at least three optional planned paths.

5. The method for ultrasound-guided nerve block anesthesia puncture based on artificial intelligence according to claim 1, wherein, In step S13, preprocessing and feature extracting the electromyogram signals specifically includes: Performing filtering and denoising processing on the collected electromyogram signals; Extract features from the preprocessed electromyogram signals, where the features include time-domain features, frequency-domain features, and time-frequency domain features.

6. The method for ultrasound-guided nerve block anesthesia puncture based on artificial intelligence according to claim 5, wherein In step S13, train a neural network with the extracted features to establish an intention prediction model between the electromyogram signals and hand movements, specifically including: S51. Normalize and label the obtained time-domain features, frequency-domain features, and time-frequency domain features to obtain a labeled dataset. S52. Train the labeled dataset through a neural network, calculate the gap between the predicted value and the true value using the mean square error, and update the network weights through the optimization algorithm SGD. S53. Use the training dataset to iteratively train the convolutional neural network, continuously adjust the weights and reduce the loss function value. S54. Evaluate the performance of the trained convolutional neural network through the validation dataset, perform hyperparameter tuning and optimization. After verifying that the generalization ability of the convolutional neural network meets the standard using the test dataset, output the intention prediction model between the electromyogram signals and hand movements.

7. The method for ultrasound-guided nerve block anesthesia puncture based on artificial intelligence according to claim 1, wherein In step S14, determine the spatial coordinate system position of the puncture needle tip based on the puncture point to be punctured, specifically including: Establish a three-dimensional coordinate system based on the puncture point to be punctured, obtain the relative positions of each pixel point in the target nerve block image with respect to the three-dimensional coordinate system, and locate the puncture needle tip based on the relative positions.

8. The method for ultrasound-guided nerve block anesthesia puncture based on artificial intelligence according to claim 1, wherein, In step S14, monitor the real-time trajectory of the puncture needle in the target nerve block image during the operation, specifically including: S61. Identify the puncture needle in the target nerve block image based on the neural network and perform feature enhancement processing. S62. Obtain the edge contour of the puncture needle body through the contour recognition algorithm and mark the color value of the edge contour. S63. Display the puncture needle after edge contour recognition in real time by superimposing it on the target nerve block image.

Images