Venipuncture path guiding method and system based on real-time image

Abstract

The invention discloses a venipuncture path guiding method and system based on real-time images, and relates to the technical field of puncture guiding. A venipuncture path guiding system based on real-time images comprises a patient difference compensation module, a real-time image analysis module and a venipuncture guiding module. Patient physiological parameter data and puncture part sign data are introduced, deep analysis is carried out in combination with a patient individual difference analysis model, a personalized puncture guide scheme can be provided according to individual differences of different patients, and puncture accuracy and safety are remarkably improved; the real-time image features, the puncture difference auxiliary features and the venipuncture guide model are combined for path planning, and the optimal puncture path can be calculated and provided in real time, so that the puncture operation is optimized, the damage to surrounding tissues is reduced, and the puncture success rate is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of puncture guidance, and in particular to a method and system for guiding a venous puncture path based on real-time images. Background Art

[0002] Puncture guidance technology is a key technology used in clinical medicine for precise positioning and guiding puncture, especially in the fields of tumor puncture and tissue biopsy. At present, commonly used puncture guidance technologies include ultrasound guidance, CT guidance and MRI guidance. Although these technologies can provide relatively accurate positioning, there are still some disadvantages, such as limited image resolution, insufficient real-time feedback and greater difficulty in operation. Especially in complex anatomical environments, differences in anatomical structure or skin characteristics of different patients may affect image quality, thereby reducing puncture accuracy and increasing errors. For example, for the elderly, obese patients or patients with difficult-to-reach veins, traditional puncture methods have operational risks and are prone to puncture failure, damage to blood vessels or nerves, and other problems. Therefore, there is an urgent need for a puncture guidance method that can analyze and process image data of the puncture site in real time, provide accurate three-dimensional positioning and path planning, and optimize the puncture path and angle according to the specific anatomical characteristics of the patient to reduce puncture errors and damage to surrounding tissues. Summary of the invention

[0003] The present invention aims to provide a method and system for guiding the path of venous puncture based on real-time images, which analyzes image data in real time and guides puncture by considering individual differences in patient characteristics.

[0004] A method for guiding a venous puncture path based on real-time imaging comprises the following steps: Obtaining a patient physiological parameter dataset X and a puncture site sign dataset Y; performing analysis based on the patient physiological parameter dataset X, the puncture site sign dataset Y and the patient individual difference analysis model to obtain the patient puncture difference auxiliary features; the patient individual difference analysis model includes a physiological parameter feature analysis layer, a site sign data analysis layer, a simulated puncture site layer and a result output layer, and screening the patient physiological parameter features by fusing the MLP and random forest models, and extracting the puncture site sign data features in combination with the adaptive convolution kernel technology and performing patient puncture difference auxiliary feature analysis to eliminate individual differences; Among them, the patient physiological parameter data set X, X = {X n |n=1,2,…,N},X n represents the nth patient physiological parameter data, N is the total number of patient physiological parameter data; puncture site sign data set Y, Y={Y m |m=1,2,…,M},Y m represents the mth item of puncture site sign data, where M is the total number of puncture site sign data; Acquire a real-time image of the patient's puncture site; perform image processing based on the real-time image of the patient's puncture site, the patient's puncture difference auxiliary features and the real-time analysis model of the puncture image to obtain the real-time image features of the puncture site; the real-time analysis model of the puncture image is used to use a composite convolution structure to speed up the image processing speed and ensure real-time performance; Real-time path planning is performed based on the patient's puncture difference auxiliary features, real-time image features of the puncture site and the venous puncture guidance model to obtain the patient's real-time venous puncture path guidance strategy; real-time venous puncture auxiliary guidance is performed on the patient based on the patient's real-time venous puncture path guidance strategy.

[0005] As a preferred technical solution of the present invention, the physiological parameter feature analysis layer in the patient individual difference analysis model is used to perform feature screening analysis on the patient physiological parameter data set X to obtain the patient physiological parameter data feature set X'; Specific steps for feature screening analysis: The physiological parameter feature analysis layer is constructed based on the fusion of MLP model and random forest model to analyze the patient's physiological parameter data X n Perform feature screening and extraction to obtain the patient's physiological parameter data feature X n ' and feature selection label Q n ; Based on the patient's physiological parameter data feature X n ' and feature selection label Q n Construct patient physiological parameter data feature set X'; feature selection label Q n For patient physiological parameter data feature X n 'Screen and determine whether to keep; The site sign data analysis layer is used to perform feature analysis on the puncture site sign data set Y to obtain the puncture site sign data feature set Y'; Specific steps for feature analysis: Use the adaptive convolution kernel in the CNN model to analyze the puncture site sign data Y m Perform feature extraction to obtain the puncture site sign data feature Y m '; The simulated puncture site layer is used to perform simulation optimization based on the patient's physiological parameter data feature set X' and the puncture site sign data feature set Y' to obtain the patient's puncture difference auxiliary features; The result output layer is used to output auxiliary features of patient puncture differences.

[0006] As a preferred technical solution of the present invention, the specific steps of performing simulation optimization in the simulated puncture site layer include: Based on the patient's physiological parameter data feature set X', the puncture site sign data feature set Y' and digital twin technology, a patient simulation puncture feedback model is constructed; Construct K patient simulation puncture parameter individuals L k , each patient simulates the puncture parameter individual L k represents the puncture parameters used by the patient simulation puncture feedback model; K patient simulation puncture parameter individuals L k Combination, get the iterative population of patient simulation puncture parameters; set the maximum number of iterations; patient simulation puncture parameter individual L k The fitness is S k ; Based on the sparrow search algorithm, the iterative population of patient simulation puncture parameters is divided into a finder population and a follower population; the finder search direction vector and the follower search direction vector are introduced; the finder search direction vector and the follower search direction vector are constructed based on the particle swarm algorithm; In the population iteration process, the finder search direction vector is used to guide the finder population to update its position, and the follower search direction vector is used to guide the follower population to update its position; at the same time, some individuals are randomly selected from the patient simulation puncture parameter iteration population as the disturbance item population; the disturbance item population updates its position based on the average position of the finder population; When the maximum number of iterations is reached, the patient simulation puncture parameter individual corresponding to the current maximum fitness is output, which is the optimal patient simulation puncture parameter individual; based on the optimal patient simulation puncture parameter individual and the patient simulation puncture feedback model, the patient puncture difference auxiliary features are output.

[0007] As a preferred technical solution of the present invention, the real-time analysis model of puncture images includes an image preprocessing layer, an image feature analysis layer and an image feature output layer; The image preprocessing layer is used to preprocess the real-time image of the puncture site of the patient to obtain a preprocessed real-time image of the puncture site of the patient; The image feature analysis layer is used to extract image features based on the pre-processed real-time image of the puncture site of the patient and the auxiliary features of the patient's puncture difference, so as to obtain the real-time image features of the puncture site; The image feature output layer is used to output real-time image features of the puncture site.

[0008] As a preferred technical solution of the present invention, the image feature analysis layer is constructed by a composite structure of three convolutional layers and a fully connected composite layer; In a convolutional layer composite structure, there are convolutional layers, normalization layers, activation function layers and maximum pooling layers; The real-time image of the patient's puncture site is preprocessed by a composite structure of three convolutional layers to obtain a real-time multi-dimensional feature map of the patient's puncture site; In the fully connected composite layer, feature integration is performed based on the real-time multi-dimensional feature map of the patient's puncture site and the auxiliary features of the patient's puncture difference to obtain the real-time image features of the puncture site; the fully connected composite layer is composed of three fully connected layers.

[0009] As a preferred technical solution of the present invention, the venous puncture guidance model includes a three-dimensional model optimization layer, a path planning layer and a strategy output layer; The three-dimensional model optimization layer optimizes the patient simulation puncture feedback model based on the real-time image features of the puncture site and the auxiliary features of the patient puncture difference to obtain an optimized three-dimensional patient simulation puncture feedback model; The path planning layer is used to perform path planning based on the optimized three-dimensional patient simulation puncture feedback model and the improved A* algorithm to obtain the patient's real-time venous puncture path guidance strategy; The strategy output layer is used to output the patient's real-time venipuncture path guidance strategy.

[0010] As a preferred technical solution of the present invention, the specific steps of performing path planning in the path planning layer include: The heuristic function of the A* algorithm is improved to obtain the improved heuristic function F(i)=g(i)+h 1 (i) *W 1 +h 2 (i) *W 2 ; where g(i) represents the actual distance cost from the starting point of the venipuncture path to the end point of the venipuncture path, and h 1 (i) represents the predicted distance cost from the end point of the venipuncture path to the optimal solution point of the reverse search, h 2 (i) represents the predicted distance cost from the end point of the venipuncture path to the midpoint of the venipuncture path; W 1 Indicates h 1 (i) Dynamic weight value; W 2 Indicates h 2 (i) dynamic weight value; i represents the end point of the current venipuncture path; The improved heuristic function F(i) and the optimized three-dimensional patient simulation puncture feedback model are used for dynamic path planning to obtain the patient's real-time intravenous puncture path guidance strategy.

[0011] A venous puncture path guidance system based on real-time imaging, comprising: The patient difference compensation module includes a data acquisition unit and a difference analysis unit; the data acquisition unit is used to acquire the patient's physiological parameter data set X and the puncture site sign data set Y; wherein the patient's physiological parameter data set X, X={X n |n=1,2,…,N},X nrepresents the nth patient physiological parameter data, N is the total number of patient physiological parameter data; puncture site sign data set Y, Y={Y m |m=1,2,…,M},Y m represents the mth item of puncture site sign data, where M is the total number of puncture site sign data; the difference analysis unit is used to perform analysis based on the patient physiological parameter data set X, the puncture site sign data set Y and the patient individual difference analysis model to obtain the patient puncture difference auxiliary feature; the patient individual difference analysis model includes a physiological parameter feature analysis layer, a site sign data analysis layer, a simulated puncture site layer and a result output layer, and the patient physiological parameter feature is screened by fusing the MLP and random forest models, and the puncture site sign data features are extracted in combination with the adaptive convolution kernel technology and the patient puncture difference auxiliary feature analysis is performed to eliminate individual differences; The real-time image analysis module includes an image analysis unit; the image analysis unit is used to obtain a real-time image of the patient's puncture site; image processing is performed based on the real-time image of the patient's puncture site, the patient's puncture difference auxiliary features and the puncture image real-time analysis model to obtain the real-time image features of the puncture site; the puncture image real-time analysis model is used to use a composite convolution structure to speed up the image processing speed and ensure real-time performance; The venous puncture guidance module includes a puncture guidance unit; the puncture guidance unit is used to perform real-time path planning based on the patient's puncture difference auxiliary characteristics, the real-time image characteristics of the puncture site and the venous puncture guidance model to obtain the patient's real-time venous puncture path guidance strategy; and perform real-time venous puncture auxiliary guidance on the patient according to the patient's real-time venous puncture path guidance strategy.

[0012] The present invention has the following advantages: 1. The present invention introduces the patient's physiological parameter data and puncture site sign data, and combines it with the patient's individual difference analysis model for in-depth analysis, so as to provide personalized puncture guidance schemes according to the individual differences of different patients, and significantly improve the accuracy and safety of puncture; the physiological parameter data is feature screened by combining MLP and random forest models, and the puncture site sign data features are extracted using adaptive convolution kernel technology, so that the model can eliminate individual differences during the processing process and optimize puncture path planning; the path planning is carried out by combining real-time image features, puncture difference auxiliary features and intravenous puncture guidance models, which can calculate and provide the best puncture path in real time, thereby optimizing the puncture operation, reducing damage to surrounding tissues, and improving the puncture success rate.

[0013] 2. The present invention constructs a patient simulation puncture feedback model based on the patient physiological parameter feature set X' and the puncture site sign data feature set Y', which can perform accurate simulation according to the individual differences of each patient, ensure personalized optimization of the puncture path and parameters, thereby improving the success rate of puncture and reducing unnecessary damage; the iterative optimization process of the patient simulation puncture parameter individual Lk, combined with the sparrow search algorithm and the particle swarm algorithm, continuously updates the parameter configuration, so that the simulated puncture parameters gradually approach the optimal solution. This adaptive adjustment mechanism can effectively respond to the dynamic changes of patients during the puncture process and provide flexible and efficient path guidance; the sparrow search algorithm divides the patient simulation puncture parameter individuals into a discoverer population and a follower population. Through the collaborative search of the discoverer population and the follower population, the diversity and convergence speed of the optimization process are enhanced, while avoiding the problem of falling into the local optimal solution and improving the global performance of the optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a schematic structural diagram of a real-time imaging-based venous puncture path guidance system used in an embodiment of the present invention. DETAILED DESCRIPTION

[0015] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0016] Embodiment 1, a method for guiding a venous puncture path based on real-time images, comprising the following steps: Obtain the patient's physiological parameter data set X and the puncture site sign data set Y; where the patient's physiological parameter data set X, X={X n |n=1,2,…,N},X n represents the nth patient physiological parameter data, N is the total number of patient physiological parameter data; puncture site sign data set Y, Y={Y m |m=1,2,…,M},Y m represents the mth item of puncture site sign data, where M is the total number of puncture site sign data; The patient physiological parameter data set contains various data related to the patient's overall health status, the function of the body's systems, and the impact on the puncture operation. Specific data categories may include age, weight, height, fat layer thickness, skin thickness, etc.; the puncture site sign data set contains anatomical features and real-time imaging data related to the puncture site. Specific data categories may include vascular location, vascular diameter, tissue type, puncture depth, tissue elasticity, etc.; these data can be obtained through medical imaging equipment (such as ultrasound, CT, MRI) and physiological monitoring equipment (such as blood glucose meter, electrocardiogram), providing basic information for individual difference analysis models; Based on the patient physiological parameter data set X, the puncture site sign data set Y and the patient individual difference analysis model, the auxiliary features of the patient puncture difference are obtained; the patient individual difference analysis model includes a physiological parameter feature analysis layer, a site sign data analysis layer, a simulated puncture site layer and a result output layer, and the patient physiological parameter features are screened by fusing the MLP and random forest models, and the puncture site sign data features are extracted by combining the adaptive convolution kernel technology and the auxiliary feature analysis of the patient puncture difference is performed to eliminate individual differences; The physiological parameter feature analysis layer in the patient individual difference analysis model is used to perform feature screening analysis on the patient physiological parameter data set X to obtain the patient physiological parameter data feature set X'; Specific steps for feature screening analysis: The physiological parameter feature analysis layer is constructed based on the fusion of MLP model and random forest model to analyze the patient's physiological parameter data X n Perform feature screening and extraction to obtain the patient's physiological parameter data feature X n ' and feature selection label Q n ; Based on the patient's physiological parameter data feature X n ' and feature selection label Q n Construct patient physiological parameter data feature set X'; feature selection label Q n For patient physiological parameter data feature X n 'Screen and determine whether to keep; The site sign data analysis layer is used to perform feature analysis on the puncture site sign data set Y to obtain the puncture site sign data feature set Y'; Specific steps for feature analysis: Use the adaptive convolution kernel in the CNN model to analyze the puncture site sign data Y m Perform feature extraction to obtain the puncture site sign data feature Y m '; The simulated puncture site layer is used to perform simulation optimization based on the patient's physiological parameter data feature set X' and the puncture site sign data feature set Y' to obtain the patient's puncture difference auxiliary features; The result output layer is used to output the auxiliary features of patient puncture differences; By integrating physiological parameter data and puncture site sign data, the model can eliminate individual differences between different patients and provide a customized puncture path for each patient, thereby improving the puncture success rate and reducing unnecessary damage; the physiological parameter feature analysis layer and the site sign data analysis layer use MLP, random forest and CNN adaptive convolution kernel technologies to screen and extract data features, enhancing the model's ability to capture details, thereby providing more accurate input data for path planning; through simulation optimization of simulated puncture sites, adjustments can be made based on the actual anatomical characteristics and puncture differences of each patient, reducing errors and operational difficulties during the puncture process, and reducing the risk of accidental injury to surrounding tissues; the output of patient puncture difference auxiliary features provides support for real-time path planning and guidance, and can adjust and guide the puncture path in real time, improve operability during surgery, and help doctors make decisions quickly; The specific steps for simulation optimization in the simulated puncture site layer include: Based on the patient's physiological parameter data feature set X', the puncture site sign data feature set Y' and digital twin technology, a patient simulation puncture feedback model is constructed; Construct K patient simulation puncture parameter individuals L k , each patient simulates the puncture parameter individual L k represents the puncture parameters used by the patient simulation puncture feedback model; K patient simulation puncture parameter individuals L k combination, obtain the iterative population of patient simulation puncture parameters; set the maximum number of iterations; the maximum number of iterations is set by professional technicians according to actual conditions; the individual L of patient simulation puncture parameters k The fitness is S k ; Fitness S k The calculation method is the inverse of minimizing tissue damage during puncture: Based on the sparrow search algorithm, the iterative population of patient simulation puncture parameters is divided into a finder population and a follower population; the finder search direction vector and the follower search direction vector are introduced; the finder search direction vector and the follower search direction vector are constructed based on the particle swarm algorithm; In the population iteration process, the finder search direction vector is used to guide the finder population to update its position, and the follower search direction vector is used to guide the follower population to update its position; at the same time, some individuals are randomly selected from the patient simulation puncture parameter iteration population as the disturbance item population; the disturbance item population updates its position based on the average position of the finder population; When the maximum number of iterations is reached, the patient simulation puncture parameter individual corresponding to the current maximum fitness is output, which is the optimal patient simulation puncture parameter individual; based on the optimal patient simulation puncture parameter individual and the patient simulation puncture feedback model, the patient puncture difference auxiliary feature is output; Through the patient simulation puncture feedback model constructed based on the patient physiological parameter feature set X' and the puncture site sign data feature set Y', accurate simulation can be performed according to the individual differences of each patient to ensure personalized optimization of the puncture path and parameters, thereby improving the success rate of puncture and reducing unnecessary damage; the individual L of the patient simulation puncture parameter k The iterative optimization process combines the sparrow search algorithm and the particle swarm algorithm to continuously update the parameter configuration, so that the simulated puncture parameters gradually approach the optimal solution. This adaptive adjustment mechanism can effectively cope with the dynamic changes of patients during the puncture process and provide flexible and efficient path guidance. The sparrow search algorithm divides the individual simulated puncture parameters of patients into a discoverer population and a follower population. Through the collaborative search of the discoverer population and the follower population, the diversity and convergence speed of the optimization process are enhanced, while avoiding the problem of falling into the local optimal solution and improving the global performance of the optimization. Introducing the disturbance term population and updating the position based on the average position of the discoverer population can effectively break the limitation of the local optimal solution and improve the exploratory nature of the search process, so that the model can find more suitable puncture paths and parameters. Acquire a real-time image of the patient's puncture site; perform image processing based on the real-time image of the patient's puncture site, the patient's puncture difference auxiliary features and the real-time analysis model of the puncture image to obtain the real-time image features of the puncture site; the real-time analysis model of the puncture image is used to use a composite convolution structure to speed up the image processing speed and ensure real-time performance; The real-time analysis model of puncture images includes an image preprocessing layer, an image feature analysis layer and an image feature output layer; The image preprocessing layer is used to preprocess the real-time image of the puncture site of the patient to obtain a preprocessed real-time image of the puncture site of the patient; The image feature analysis layer is used to extract image features based on the pre-processed real-time image of the puncture site of the patient and the auxiliary features of the patient's puncture difference, so as to obtain the real-time image features of the puncture site; The image feature output layer is used to output the real-time image features of the puncture site; The image feature analysis layer is constructed by a composite structure of three convolutional layers and a fully connected composite layer; In a convolutional layer composite structure, there are convolutional layers, normalization layers, activation function layers and maximum pooling layers; The real-time image of the patient's puncture site is preprocessed by a composite structure of three convolutional layers to obtain a real-time multi-dimensional feature map of the patient's puncture site; In the fully connected composite layer, feature integration is performed based on the real-time multi-dimensional feature map of the patient's puncture site and the auxiliary features of the patient's puncture difference to obtain the real-time image features of the puncture site; the fully connected composite layer consists of three fully connected layers; The composite convolution structure accelerates the image feature extraction process and can process high-dimensional puncture site image data to ensure that real-time image analysis is completed within a limited time to meet the time requirements of actual clinical operations. Through parallel processing and efficient convolution structure, it can respond quickly and provide real-time feedback on puncture path adjustment to avoid puncture errors caused by delays. The multi-layer convolution structure improves the ability to recognize puncture site details, especially in complex anatomical environments, and can accurately extract key features to support more accurate puncture path planning. The patient puncture difference auxiliary features combined with deep learning feature extraction of image data help eliminate individual differences, optimize puncture plans for each patient's specific situation, and ensure personalized and precise treatment. The fully connected composite layer improves the overall understanding of the puncture site by integrating the patient puncture difference auxiliary features with real-time image data, and better grasps the relationship between the patient's physiological characteristics and the puncture environment. Perform real-time path planning based on the patient's puncture difference auxiliary features, the real-time image features of the puncture site and the venous puncture guidance model to obtain the patient's real-time venous puncture path guidance strategy; perform real-time venous puncture auxiliary guidance on the patient based on the patient's real-time venous puncture path guidance strategy; The venipuncture guidance model includes a three-dimensional model optimization layer, a path planning layer, and a strategy output layer; The three-dimensional model optimization layer optimizes the patient simulation puncture feedback model based on the real-time image features of the puncture site and the auxiliary features of the patient puncture difference to obtain an optimized three-dimensional patient simulation puncture feedback model; the optimized three-dimensional patient simulation puncture feedback model can display the actual venous puncture area and venous puncture target point; The path planning layer is used to perform path planning based on the optimized three-dimensional patient simulation puncture feedback model and the improved A* algorithm to obtain the patient's real-time venous puncture path guidance strategy; The strategy output layer is used to output the patient's real-time venipuncture path guidance strategy; The specific steps of path planning in the path planning layer include: The heuristic function of the A* algorithm is improved to obtain the improved heuristic function F(i)=g(i)+h 1 (i) *W 1 +h 2 (i) *W 2 ; where g(i) represents the actual distance cost from the starting point of the venipuncture path to the end point of the venipuncture path, and h 1 (i) represents the predicted distance cost from the end point of the venipuncture path to the optimal solution point of the reverse search, h 2 (i) represents the predicted distance cost from the end point of the venipuncture path to the midpoint of the venipuncture path; W 1 Indicates h 1 (i) Dynamic weight value; W 2 Indicates h2 (i) Dynamic weight value; W 1 and W 2 The specific value of is set by professional technicians according to actual conditions; i represents the end point of the current venous puncture path; The improved heuristic function F(i) and the optimized three-dimensional patient simulation puncture feedback model are used for dynamic path planning to obtain the patient's real-time venous puncture path guidance strategy; The three-dimensional model optimization layer combines the patient puncture difference auxiliary features and real-time image features to optimize the patient simulation puncture feedback model, which can accurately display the puncture area and target point and clearly identify the actual puncture path. Through the accurate three-dimensional feedback model, the path can be dynamically adjusted during the puncture process to avoid mispuncture and ensure more accurate puncture operations. The improved heuristic function introduces dynamic weights, combines the actual path distance, the predicted end point distance and the predicted path midpoint distance, so that the path planning is more in line with the needs of the actual scene. The dynamic adjustment of the weight value in the heuristic function can meet the puncture needs of different patients, adjust the path planning strategy in real time, and improve the intelligence and flexibility of path selection. In the process of path planning, dynamic path planning combines real-time imaging data, auxiliary features of individual patient differences and optimized three-dimensional models to ensure that each path adjustment can be optimized according to the actual situation, ensuring the flexibility of real-time response during the puncture process; through the improved A* algorithm, the optimal path can be quickly calculated under variable physiological parameters and anatomical structures, reducing the probability of puncture failure; During actual use, once the path planning is completed, real-time puncture guidance information can be provided through a display, visual interface or headphones. This guidance information includes the recommended puncture angle, puncture depth, and the best path along the blood vessel. The path guidance will be continuously updated and real-time feedback will be given to help doctors accurately execute in actual operations. Doctors will accurately locate the puncture site according to the system's guidance strategy. The system will monitor the puncture process in real time and provide fine-tuning suggestions based on actual conditions, such as adjusting the puncture angle, position or depth to avoid damaging surrounding tissues or causing puncture failure. At the same time, data will be recorded after a successful puncture, including information such as puncture position, depth and angle, for future reference and analysis. By continuously collecting patient data, the guidance strategy can be gradually optimized and the intelligent decision-making support capabilities can be enhanced.

[0017] Example 2, a real-time imaging-based venous puncture path guidance system, see Figure 1 As shown, including: The patient difference compensation module includes a data acquisition unit and a difference analysis unit; the data acquisition unit is used to acquire the patient's physiological parameter data set X and the puncture site sign data set Y; wherein the patient's physiological parameter data set X, X={X n |n=1,2,…,N},Xn represents the nth patient physiological parameter data, N is the total number of patient physiological parameter data; puncture site sign data set Y, Y={Y m |m=1,2,…,M},Y m represents the mth item of puncture site sign data, where M is the total number of puncture site sign data; the difference analysis unit is used to perform analysis based on the patient physiological parameter data set X, the puncture site sign data set Y and the patient individual difference analysis model to obtain the patient puncture difference auxiliary feature; the patient individual difference analysis model includes a physiological parameter feature analysis layer, a site sign data analysis layer, a simulated puncture site layer and a result output layer, and the patient physiological parameter feature is screened by fusing the MLP and random forest models, and the puncture site sign data features are extracted in combination with the adaptive convolution kernel technology and the patient puncture difference auxiliary feature analysis is performed to eliminate individual differences; The real-time image analysis module includes an image analysis unit; the image analysis unit is used to obtain a real-time image of the patient's puncture site; image processing is performed based on the real-time image of the patient's puncture site, the patient's puncture difference auxiliary features and the puncture image real-time analysis model to obtain the real-time image features of the puncture site; the puncture image real-time analysis model is used to use a composite convolution structure to speed up the image processing speed and ensure real-time performance; The venous puncture guidance module includes a puncture guidance unit; the puncture guidance unit is used to perform real-time path planning based on the patient's puncture difference auxiliary characteristics, the real-time image characteristics of the puncture site and the venous puncture guidance model to obtain the patient's real-time venous puncture path guidance strategy; and perform real-time venous puncture auxiliary guidance on the patient according to the patient's real-time venous puncture path guidance strategy.

[0018] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for guiding the path of venous puncture based on real-time images, characterized in that: The following steps are involved: Obtaining a patient physiological parameter data set X and a puncture site sign data set Y; performing analysis based on the patient physiological parameter data set X, the puncture site sign data set Y and a patient individual difference analysis model to obtain a patient puncture difference auxiliary feature; The patient individual difference analysis model includes a physiological parameter feature analysis layer, a site sign data analysis layer, a simulated puncture site layer, and a result output layer. It screens the patient's physiological parameter features by fusing the MLP and random forest models, and extracts the puncture site sign data features by combining the adaptive convolution kernel technology and performs auxiliary feature analysis of patient puncture differences to eliminate individual differences. Among them, the patient physiological parameter data set X, X = {X n |n=1,2,…,N},X n represents the nth patient physiological parameter data, N is the total number of patient physiological parameter data; puncture site sign data set Y, Y={Y m |m=1,2,…,M},Y m represents the mth item of puncture site sign data, where M is the total number of puncture site sign data; Acquire a real-time image of the patient's puncture site; perform image processing based on the real-time image of the patient's puncture site, the patient's puncture difference auxiliary features and the real-time analysis model of the puncture image to obtain the real-time image features of the puncture site; the real-time analysis model of the puncture image is used to use a composite convolution structure to speed up the image processing speed and ensure real-time performance; Real-time path planning is performed based on the patient's puncture difference auxiliary features, real-time image features of the puncture site and the venous puncture guidance model to obtain the patient's real-time venous puncture path guidance strategy; real-time venous puncture auxiliary guidance is performed on the patient based on the patient's real-time venous puncture path guidance strategy.

2. A method for guiding the venous puncture path based on real-time imaging according to claim 1, characterized in that: The physiological parameter feature analysis layer in the patient individual difference analysis model is used to perform feature screening analysis on the patient physiological parameter data set X to obtain the patient physiological parameter data feature set X'; Specific steps for feature screening analysis: The physiological parameter feature analysis layer is constructed based on the fusion of MLP model and random forest model to analyze the patient's physiological parameter data X n Perform feature screening and extraction to obtain the patient's physiological parameter data feature X n ' and feature selection label Q n ; Based on the patient's physiological parameter data feature X n ' and feature selection label Q n Construct patient physiological parameter data feature set X'; feature selection label Q n For patient physiological parameter data feature X n 'Screen and determine whether to keep; The site sign data analysis layer is used to perform feature analysis on the puncture site sign data set Y to obtain the puncture site sign data feature set Y'; Specific steps for feature analysis: Use the adaptive convolution kernel in the CNN model to analyze the puncture site sign data Y m Perform feature extraction to obtain the puncture site sign data feature Y m '; The simulated puncture site layer is used to perform simulation optimization based on the patient's physiological parameter data feature set X' and the puncture site sign data feature set Y' to obtain the patient's puncture difference auxiliary features; The result output layer is used to output auxiliary features of patient puncture differences.

3. A method for guiding the venous puncture path based on real-time imaging according to claim 2, characterized in that: The specific steps for simulation optimization in the simulated puncture site layer include: Based on the patient's physiological parameter data feature set X', the puncture site sign data feature set Y' and digital twin technology, a patient simulation puncture feedback model is constructed; Construct K patient simulation puncture parameter individuals L k , each patient simulates the puncture parameter individual L k represents the puncture parameters used by the patient simulation puncture feedback model; K patient simulation puncture parameter individuals L k Combination, get the iterative population of patient simulation puncture parameters; set the maximum number of iterations; patient simulation puncture parameter individual L k The fitness is S k ; Based on the sparrow search algorithm, the iterative population of patient simulation puncture parameters is divided into a finder population and a follower population; the finder search direction vector and the follower search direction vector are introduced; the finder search direction vector and the follower search direction vector are constructed based on the particle swarm algorithm; In the population iteration process, the finder search direction vector is used to guide the finder population to update its position, and the follower search direction vector is used to guide the follower population to update its position; at the same time, some individuals are randomly selected from the patient simulation puncture parameter iteration population as the disturbance item population; the disturbance item population updates its position based on the average position of the finder population; When the maximum number of iterations is reached, the patient simulation puncture parameter individual corresponding to the current maximum fitness is output, which is the optimal patient simulation puncture parameter individual; based on the optimal patient simulation puncture parameter individual and the patient simulation puncture feedback model, the patient puncture difference auxiliary features are output.

4. A method for guiding the venous puncture path based on real-time imaging according to claim 3, characterized in that: The real-time analysis model of puncture images includes an image preprocessing layer, an image feature analysis layer and an image feature output layer; The image preprocessing layer is used to preprocess the real-time image of the puncture site of the patient to obtain a preprocessed real-time image of the puncture site of the patient; The image feature analysis layer is used to extract image features based on the pre-processed real-time image of the puncture site of the patient and the auxiliary features of the patient's puncture difference, so as to obtain the real-time image features of the puncture site; The image feature output layer is used to output real-time image features of the puncture site.

5. A method for guiding the venous puncture path based on real-time imaging according to claim 4, characterized in that: The image feature analysis layer is constructed by a composite structure of three convolutional layers and a fully connected composite layer; In a convolutional layer composite structure, there are convolutional layers, normalization layers, activation function layers and maximum pooling layers; The real-time image of the patient's puncture site is preprocessed by a composite structure of three convolutional layers to obtain a real-time multi-dimensional feature map of the patient's puncture site; In the fully connected composite layer, feature integration is performed based on the real-time multi-dimensional feature map of the patient's puncture site and the auxiliary features of the patient's puncture difference to obtain the real-time image features of the puncture site; The fully connected composite layer consists of three fully connected layers.

6. A method for guiding the venous puncture path based on real-time imaging according to claim 5, characterized in that: The venipuncture guidance model includes a three-dimensional model optimization layer, a path planning layer, and a strategy output layer; The three-dimensional model optimization layer optimizes the patient simulation puncture feedback model based on the real-time image features of the puncture site and the auxiliary features of the patient puncture difference to obtain an optimized three-dimensional patient simulation puncture feedback model; The path planning layer is used to perform path planning based on the optimized three-dimensional patient simulation puncture feedback model and the improved A* algorithm to obtain the patient's real-time venous puncture path guidance strategy; The strategy output layer is used to output the patient's real-time venipuncture path guidance strategy.

7. A method for guiding the venous puncture path based on real-time imaging according to claim 6, characterized in that: The specific steps of path planning in the path planning layer include: The heuristic function of the A* algorithm is improved to obtain the improved heuristic function F(i)=g(i)+h1(i)*W1+h2(i)*W2; where g(i) represents the actual distance cost from the starting point of the venous puncture path to the end point of the venous puncture path, h1(i) represents the predicted distance cost from the end point of the venous puncture path to the optimal solution point of the reverse search, and h2(i) represents the predicted distance cost from the end point of the venous puncture path to the midpoint of the venous puncture path; W1 represents the dynamic weight value of h1(i); W2 represents the dynamic weight value of h2(i); i represents the current end point of the venous puncture path; The improved heuristic function F(i) and the optimized three-dimensional patient simulation puncture feedback model are used for dynamic path planning to obtain the patient's real-time intravenous puncture path guidance strategy.

8. A real-time image-based venous puncture path guidance system, characterized in that: The system uses a real-time image-based venous puncture path guidance method as described in any one of claims 1 to 7, including: The patient difference compensation module includes a data acquisition unit and a difference analysis unit; the data acquisition unit is used to acquire the patient's physiological parameter data set X and the puncture site sign data set Y; wherein the patient's physiological parameter data set X, X={X n |n=1,2,…,N},X n represents the nth patient physiological parameter data, N is the total number of patient physiological parameter data; puncture site sign data set Y, Y={Y m |m=1,2,…,M},Y m represents the mth item of puncture site sign data, where M is the total number of puncture site sign data; the difference analysis unit is used to perform analysis based on the patient physiological parameter data set X, the puncture site sign data set Y and the patient individual difference analysis model to obtain the patient puncture difference auxiliary feature; the patient individual difference analysis model includes a physiological parameter feature analysis layer, a site sign data analysis layer, a simulated puncture site layer and a result output layer, and the patient physiological parameter feature is screened by fusing the MLP and random forest models, and the puncture site sign data features are extracted in combination with the adaptive convolution kernel technology and the patient puncture difference auxiliary feature analysis is performed to eliminate individual differences; The real-time image analysis module includes an image analysis unit; the image analysis unit is used to obtain a real-time image of the patient's puncture site; image processing is performed based on the real-time image of the patient's puncture site, the patient's puncture difference auxiliary features and the puncture image real-time analysis model to obtain the real-time image features of the puncture site; the puncture image real-time analysis model is used to use a composite convolution structure to speed up the image processing speed and ensure real-time performance; The venous puncture guidance module includes a puncture guidance unit; the puncture guidance unit is used to perform real-time path planning based on the patient's puncture difference auxiliary characteristics, the real-time image characteristics of the puncture site and the venous puncture guidance model to obtain the patient's real-time venous puncture path guidance strategy; and perform real-time venous puncture auxiliary guidance on the patient according to the patient's real-time venous puncture path guidance strategy.

Images