Femoral artery position determination method and device

Abstract

The invention discloses a femoral artery position determining method and device, and relates to the technical field of femoral artery puncture. A femoral artery position determination device comprises a femoral artery data acquisition module, a femoral artery signal detection module and a femoral artery position determination module. According to the invention, image analysis and patient sign data are combined, so that accurate perception and judgment of the femoral artery position are realized; preprocessing and analyzing the input femoral artery image and the physical sign data of the patient through a femoral artery image reflection light intensity analysis model to obtain femoral artery pulse signal data; enhancing the signal by using a pulse wave signal sensing model, and preliminarily determining the sensing range of the femoral artery; based on a femoral artery position determination model, fusing the physical sign data and the image features, and further screening the accurate position of the femoral artery; through a multi-level analysis model, efficient recognition and positioning of the femoral artery position are achieved, the accuracy and reliability of detection are improved, and the method has good application prospects.

Description

Technical Field

[0001] The present invention relates to the technical field of femoral artery puncture, and in particular to a method and device for determining the position of a femoral artery. Background Art

[0002] Femoral artery puncture is mainly used to obtain blood samples, perform diagnostic tests or implement interventional treatments. It is a key diagnostic and treatment method. During femoral artery puncture, if the position of the femoral artery is uncertain, serious complications may occur, including but not limited to local hematoma, retroperitoneal hematoma, pseudoaneurysm and arteriovenous fistula, etc. These complications will not only increase the patient's pain, but may also cause long-term effects on the patient's health and even endanger his life. Therefore, accurate femoral artery positioning is crucial to ensure the safety and effectiveness of the puncture operation.

[0003] Methods for determining the location of the femoral artery include surface positioning, stethoscope measurement, X-ray examination, ultrasound examination, CT angiography and magnetic resonance imaging examination; however, most of these methods rely on the professional skills of medical staff to accurately locate the femoral artery, and different medical staff or different conditions of patients may lead to inaccurate determination of the femoral artery location. Summary of the invention

[0004] The present invention aims to provide a method and device for determining the position of a femoral artery, so as to improve the accuracy of determining the position of the femoral artery.

[0005] A method for determining a femoral artery position comprises the following steps:

[0006] Step S1: Femoral artery data acquisition

[0007] Obtain the target femoral artery image set P to be analyzed, P = {P 1 , P 2 ,…,P n ,…,P N}; Among them, P n Represented as the target femoral artery image P to be analyzed in the target femoral artery image set P to be analyzed n , N is the total number of target femoral artery images in a set of target femoral artery images to be analyzed P; obtain target patient vital sign data; input the target femoral artery image set P to be analyzed and the target patient vital sign data into the femoral artery image reflected light intensity analysis model for analysis, and obtain the femoral artery pulse signal data set B and all preprocessed target femoral artery images to be analyzed P n ', B = {B 1 , B 2 , …, B n , …, B NThe femoral artery image reflected light intensity analysis model includes an image preprocessing layer, a regional analysis layer, a signal extraction layer and a result output layer, which is constructed based on a CNN model and a multi-layer perceptron, and is used to analyze the femoral artery pulse signal of the target femoral artery image set P to be analyzed in combination with the target patient's vital sign data;

[0008] Step S2: Femoral artery signal detection

[0009] The femoral artery pulse signal dataset B, all pre-processed target femoral artery images P n 'The signal is input into the femoral artery pulse wave signal perception model for analysis to obtain the femoral artery position perception range; the femoral artery pulse wave signal perception model includes a signal processing layer, a signal feature matching layer, an image overlap layer and a signal range output layer. The filter is constructed based on the improved optimization algorithm to further enhance the femoral artery pulse signal and preliminarily divide the femoral artery range;

[0010] Step S3: Femoral artery position determination

[0011] The femoral artery range is determined according to the femoral artery position perception range and the femoral artery position determination model to obtain the femoral artery position determination result; the next step is performed based on the femoral artery position determination result; the femoral artery position determination model includes a vital sign data acquisition layer, a feature fusion layer, a position determination layer and a position result output layer, which is constructed based on the CNN model and is used to further screen the precise range of the femoral artery.

[0012] As a preferred technical solution of the present invention, the femoral artery image reflected light intensity analysis model in step S1 includes an image preprocessing layer, a regional analysis layer, a signal extraction layer and a result output layer;

[0013] The image preprocessing layer is used to process the target femoral artery image P to be analyzed according to the target patient's vital sign data and the target femoral artery image P to be analyzed in the target femoral artery image set P n Perform feature analysis to obtain the preprocessed target femoral artery image P n ';

[0014] The regional analysis layer is used to pre-process the target femoral artery image P to be analyzed. n 'Identify the signal extraction area and obtain the femoral artery signal extraction area R n ;

[0015] The signal extraction layer is used to extract region R based on the femoral artery signal n Perform signal recognition to obtain femoral artery pulse signal data B n ;

[0016] The resulting output layer is used to convert all femoral artery pulse signal data B nThe femoral artery pulse signal data set B is obtained by combining, and the femoral artery pulse signal data set B is output.

[0017] As a preferred technical solution of the present invention, the specific steps of constructing the image preprocessing layer include:

[0018] In the image preprocessing layer, four image preprocessing adjustment and analysis layers are constructed;

[0019] In the first image preprocessing adjustment and analysis layer, a multi-layer perceptron is used to construct several fully connected layers for extracting high-dimensional representations of patient vital sign data to obtain patient vital sign data features; in the second image preprocessing adjustment and analysis layer, a CNN network is used to construct several convolutional layers for extracting image data features of patient image data to obtain patient image data features; in the third image preprocessing adjustment and analysis layer, a neural network structure is used to construct several embedding layers for splicing patient vital sign data features and patient image data features to obtain patient fusion feature vectors; the patient fusion feature vectors are input into the trained contrast loss optimization function for calculation to obtain optimized patient fusion feature vectors; in the fourth image preprocessing adjustment and analysis layer, a multi-layer perceptron is used to generate image adjustment parameters according to the optimized patient fusion feature vectors, and the input patient image data is adjusted using the image adjustment parameters to obtain preprocessed patient image data;

[0020] The specific steps for training the image preprocessing layer include:

[0021] Collecting several groups of image preprocessing training samples; each group of image preprocessing training samples contains sample input values ​​and sample output values; wherein the sample input values ​​are patient vital sign data and femoral artery images of patients to be processed, and the sample output values ​​are preprocessed femoral artery images of patients; combining several groups of image preprocessing training samples to obtain an image preprocessing training set;

[0022] The image preprocessing training set is input into the image preprocessing layer to perform model training with the sample output value as the target, and the image preprocessing layer to be evaluated is obtained; image evaluation is performed on the image preprocessing layer to be evaluated to obtain the image preprocessing layer model evaluation result; if the image preprocessing layer model evaluation result is passed, the image preprocessing layer to be evaluated is used as the image preprocessing layer in the femoral artery image reflected light intensity analysis model; otherwise, the image preprocessing training set is used to continue model training.

[0023] As a preferred technical solution of the present invention, the femoral artery pulse wave signal perception model includes a signal processing layer, a signal feature matching layer, an image overlapping layer and a signal range output layer;

[0024] The signal processing layer is used to process the femoral artery pulse signal data B in the femoral artery pulse signal data set B. nAdaptive hybrid filter is used to filter the femoral artery pulse signal data B n Perform denoising to obtain pre-processed femoral artery pulse signal data B n ';

[0025] The signal feature matching layer is used to pre-process the femoral artery pulse signal data B n ' and preprocess the target femoral artery image P to be analyzed n 'Perform feature matching to obtain the femoral artery enhanced marked image Z n ;

[0026] The image overlap layer is used to enhance all femoral artery images Z n Perform image overlap to obtain the femoral artery position sensing range;

[0027] The signal range output layer is used to output the femoral artery position perception range.

[0028] As a preferred technical solution of the present invention, the specific steps of constructing an adaptive hybrid filter include:

[0029] Construct a randomly generated K filter configuration parameter individual H k , k = 1, 2, ..., K; each filter configuration parameter individual H k contains a set of filter configuration parameters for constructing an adaptive hybrid filter; the K filter configuration parameters H k Combining into a filter configuration parameter iteration population; constructing a signal processing simulation training set;

[0030] Dividing the filter configuration parameter iteration population iteration phase into a filter configuration parameter first iteration phase, a filter configuration parameter second iteration phase and a filter configuration parameter third iteration phase;

[0031] The filter configuration parameters are individual H k Perform simulation calculation to obtain the fitness S k ; The specific steps of simulation calculation are:

[0032] Using the filter configuration parameter individual H k Construct a simulated adaptive hybrid filter; use the signal processing simulation training set to evaluate the model of the simulated adaptive hybrid filter, and obtain the evaluation result of the simulated adaptive hybrid filter model; use the evaluation result of the simulated adaptive hybrid filter model as the filter configuration parameter individual H k The fitness S k ;

[0033] In the first iteration of the filter configuration parameters, the formula H is used k ′=H k +W*(H k-rand(J)) for the filter configuration parameters individual H k Perform population iteration, H k 'Indicates updating the filter configuration parameter individual H k ', W represents the adaptive adjustment coefficient, rand(J) represents the random solution; and the filter configuration parameter individual H is updated k 'Recalculate fitness;

[0034] In the second iteration of filter configuration parameters, the formula The filter configuration parameters are individual H k To update, G k Denotes the filter configuration parameter individual H k The mutation probability, S k Denotes the filter configuration parameter individual H k The fitness S k , β represents the adaptive mutation parameter; using G k The filter configuration parameters are individual H k Update to get the new filter configuration parameter individual H k , and configure the new filter parameters for the individual H k Recalculate fitness;

[0035] In the third iteration phase of the filter configuration parameters, a new filter configuration parameter individual pair H is randomly generated. k Replace it and get the new filter configuration parameter individual H k , and configure the new filter parameters for the individual H k Recalculate fitness;

[0036] When the population iteration ends, the filter configuration parameter individual H corresponding to the maximum output fitness k , which is the optimal filter configuration parameter individual, and the filter configuration parameters in the optimal filter configuration parameter individual are used to construct an adaptive hybrid filter.

[0037] As a preferred technical solution of the present invention, the femoral artery position determination model includes a vital sign data acquisition layer, a feature fusion layer, a position determination layer and a position result output layer;

[0038] The vital sign data acquisition layer is used to reacquire the vital sign data of the target patient to obtain updated vital sign data of the target patient; perform feature extraction on the updated vital sign data of the target patient to obtain features of the updated vital sign data of the target patient;

[0039] The feature fusion layer is used to extract the features of the femoral artery position perception range to obtain the femoral artery position perception range features; the updated target patient's vital sign data features and the femoral artery position perception range features are feature fused to obtain the fused femoral artery position perception range features;

[0040] The position determination layer is used to further determine the position according to the fused femoral artery position sensing range feature and the femoral artery position sensing range, and obtain the femoral artery position determination result;

[0041] The position result output layer is used to output the femoral artery position determination results.

[0042] As a preferred technical solution of the present invention, the specific steps of training the position determination layer include:

[0043] Collecting several groups of femoral artery position determination training samples, each group of femoral artery position determination training samples includes training patient physical sign data, an initial femoral artery determination range and a corresponding final femoral artery determination range; using the final femoral artery determination range as a target value of the femoral artery position determination training sample; combining several groups of femoral artery position determination training samples to obtain a femoral artery position determination training set;

[0044] The femoral artery position determination training set is input into the femoral artery position determination model to train the position determination layer with the target value as the target to obtain the initial position determination layer; the initial position determination layer is evaluated, and if the initial position determination layer passes the model evaluation, the initial position determination layer is used as the position determination layer in the femoral artery position determination model; otherwise, the femoral artery position determination training set is used to continue model training.

[0045] A femoral artery position determination device, comprising:

[0046] The femoral artery data acquisition module includes a physical sign data acquisition unit and a femoral artery image data acquisition unit; the physical sign data acquisition unit is used to acquire the physical sign data of the target patient; the femoral artery image data acquisition unit is used to acquire the target femoral artery image set P to be analyzed, P = {P 1 , P 2 ,…,P n ,…,P N}; Among them, P n Represented as the target femoral artery image P to be analyzed in the target femoral artery image set P to be analyzed n , N is the total number of target femoral artery images in a set of target femoral artery images to be analyzed P;

[0047] The femoral artery signal detection module includes a femoral artery image processing unit and a signal sensing unit; the femoral artery image processing unit is used to input the target femoral artery image set P to be analyzed and the target patient's vital signs data into the femoral artery image reflected light intensity analysis model for analysis, and obtain the femoral artery pulse signal data set B and all pre-processed target femoral artery images P to be analyzed. n ', B = {B 1 , B 2 , …, B n , …, B N The femoral artery image reflected light intensity analysis model includes an image preprocessing layer, a regional analysis layer, a signal extraction layer and a result output layer, which is constructed based on a CNN model and a multi-layer perceptron and is used to analyze the femoral artery pulse signal of the target femoral artery image set P to be analyzed in combination with the target patient's vital sign data; the signal perception unit is used to combine the femoral artery pulse signal data set B, all preprocessed target femoral artery images P to be analyzed n 'The signal is input into the femoral artery pulse wave signal perception model for analysis to obtain the femoral artery position perception range; the femoral artery pulse wave signal perception model includes a signal processing layer, a signal feature matching layer, an image overlap layer and a signal range output layer. The filter is constructed based on the improved optimization algorithm to further enhance the femoral artery pulse signal and preliminarily divide the femoral artery range;

[0048] The femoral artery position determination module includes a position determination unit; the position determination unit is used to determine the femoral artery range according to the femoral artery position perception range and the femoral artery position determination model to obtain the femoral artery position determination result; the next step is performed based on the femoral artery position determination result; the femoral artery position determination model includes a vital sign data acquisition layer, a feature fusion layer, a position determination layer and a position result output layer, which is constructed based on the CNN model and is used to further screen the precise range of the femoral artery.

[0049] The present invention has the following advantages:

[0050] 1. The present invention realizes accurate perception and judgment of the femoral artery position by combining image analysis with patient vital sign data; pre-processes and analyzes the input femoral artery image and patient vital sign data through the femoral artery image reflected light intensity analysis model to obtain femoral artery pulse signal data; enhances the signal by using the pulse wave signal perception model to preliminarily determine the perception range of the femoral artery; based on the femoral artery position determination model, the vital sign data and image features are integrated to further screen the precise position of the femoral artery; through a multi-level analysis model, efficient recognition and positioning of the femoral artery position is realized, the accuracy and reliability of detection are improved, and it has good application prospects.

[0051] 2. The present invention improves the accuracy and effect of image preprocessing by constructing a multi-level image preprocessing layer, effectively combining patient vital sign data and femoral artery image data; uses a multi-layer perceptron and a convolutional neural network to extract high-dimensional features of patient vital signs and image data, and fuses the two through an embedding layer; uses a contrast loss optimization function to further optimize the fused feature vector, generates image adjustment parameters, and accurately adjusts the input image data, which helps to ensure high-quality preprocessing of the femoral artery image, provides more reliable data support for subsequent signal extraction and femoral artery position determination, and significantly improves the accuracy and robustness of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A schematic structural diagram of a femoral artery position determination device used in an embodiment of the present invention.

[0053] Figure 2 The figure is a flow chart of a method for determining the position of a femoral artery adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] 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.

[0055] Example 1, a method for determining the position of a femoral artery, see Figure 2 As shown, the following steps are included:

[0056] Step S1: Femoral artery data acquisition

[0057] Obtain the target femoral artery image set P to be analyzed, P = {P 1 , P 2 ,…,P n ,…,P N}; Among them, P n Represented as the target femoral artery image P to be analyzed in the target femoral artery image set P to be analyzed n , N is the total number of target femoral artery images in a set of target femoral artery images to be analyzed P; obtain target patient vital sign data; input the target femoral artery image set P to be analyzed and the target patient vital sign data into the femoral artery image reflected light intensity analysis model for analysis, and obtain the femoral artery pulse signal data set B and all preprocessed target femoral artery images to be analyzed P n ', B = {B 1 , B 2 , …, B n , …, B NThe femoral artery image reflected light intensity analysis model includes an image preprocessing layer, a regional analysis layer, a signal extraction layer and a result output layer, which is constructed based on a CNN model and a multi-layer perceptron, and is used to analyze the femoral artery pulse signal of the target femoral artery image set P to be analyzed in combination with the target patient's vital sign data;

[0058] A medical optical imaging method can be used to obtain a target femoral artery image set to be analyzed, and a number of sensors can be used to obtain target patient vital sign data;

[0059] The femoral artery image reflected light intensity analysis model in step S1 includes an image preprocessing layer, a regional analysis layer, a signal extraction layer and a result output layer;

[0060] The image preprocessing layer is used to process the target femoral artery image P to be analyzed according to the target patient's vital sign data and the target femoral artery image P to be analyzed in the target femoral artery image set P n Perform feature analysis to obtain the preprocessed target femoral artery image P n ';

[0061] The regional analysis layer is used to pre-process the target femoral artery image P to be analyzed. n 'Identify the signal extraction area and obtain the femoral artery signal extraction area R n ;

[0062] The signal extraction layer is used to extract region R based on the femoral artery signal n Perform signal recognition to obtain femoral artery pulse signal data B n ; The signal recognition method is to extract the signal using the gray value method;

[0063] The resulting output layer is used to convert all femoral artery pulse signal data B n Combining to obtain a femoral artery pulse signal data set B, and outputting the femoral artery pulse signal data set B;

[0064] By constructing a femoral artery image reflected light intensity analysis model, pulse signal data can be efficiently extracted from femoral artery images, which has significant beneficial effects; the image preprocessing layer combines patient vital sign data with femoral artery images for feature analysis, ensuring effective image preprocessing; the regional analysis layer accurately identifies the signal extraction area, improving the accuracy of signal extraction; the signal extraction layer further identifies the pulse signal in a specific area, ensuring the integrity and accuracy of the signal; the result output layer combines all extracted signal data to generate a complete femoral artery pulse signal data set; this layered processing method effectively improves the accuracy and reliability of signal extraction, providing a solid foundation for subsequent femoral artery analysis;

[0065] The specific steps to build the image preprocessing layer include:

[0066] In the image preprocessing layer, four image preprocessing adjustment and analysis layers are constructed;

[0067] In the first image preprocessing adjustment and analysis layer, a multi-layer perceptron is used to construct several fully connected layers for extracting high-dimensional representations of patient vital sign data to obtain patient vital sign data features; in the second image preprocessing adjustment and analysis layer, a CNN network is used to construct several convolutional layers for extracting image data features of patient image data to obtain patient image data features; in the third image preprocessing adjustment and analysis layer, a neural network structure is used to construct several embedding layers for splicing patient vital sign data features and patient image data features to obtain patient fusion feature vectors; the patient fusion feature vectors are input into the trained contrast loss optimization function for calculation to obtain optimized patient fusion feature vectors; in the fourth image preprocessing adjustment and analysis layer, a multi-layer perceptron is used to generate image adjustment parameters according to the optimized patient fusion feature vectors, and the input patient image data is adjusted using the image adjustment parameters to obtain preprocessed patient image data;

[0068] The specific steps for training the image preprocessing layer include:

[0069] Collecting several groups of image preprocessing training samples; each group of image preprocessing training samples contains sample input values ​​and sample output values; wherein the sample input values ​​are patient vital sign data and femoral artery images of patients to be processed, and the sample output values ​​are preprocessed femoral artery images of patients; combining several groups of image preprocessing training samples to obtain an image preprocessing training set;

[0070] The image preprocessing training set is input into the image preprocessing layer to perform model training with the sample output value as the target, and the image preprocessing layer to be evaluated is obtained; the image preprocessing layer to be evaluated is evaluated to obtain the image preprocessing layer model evaluation result; if the image preprocessing layer model evaluation result is passed, the image preprocessing layer to be evaluated is used as the image preprocessing layer in the femoral artery image reflected light intensity analysis model; otherwise, the model training is continued using the image preprocessing training set;

[0071] The multi-layer perceptron extracts the high-dimensional features of the patient's vital sign data, and deeply analyzes the patient's vital sign information through the fully connected layer, which ensures that richer and more accurate individual features of the patient are extracted, providing effective data support for subsequent processing. This step makes full use of the patient's physiological data to make the model more adaptable to individual differences; the CNN network extracts the features of the femoral artery image data, which can capture the subtle features in the image, such as light intensity changes and texture information, which is crucial for the subsequent femoral artery signal extraction. The application of the convolutional layer enables the model to automatically discover the key features in the image, avoiding the limitations of traditional manual feature selection; the vital sign data features and image data features are spliced ​​to form a fused feature vector. The design of the neural network embedding layer ensures that the two different dimensional data can be effectively integrated to avoid information loss and conflict, which can improve the performance of the model in processing multimodal data and make the fused feature vector more representative; the multi-layer perceptron is used to generate image adjustment parameters based on the optimized fused feature vector, and the image is accurately adjusted. The pertinence and accuracy of the image adjustment are further enhanced by calculating the contrast loss optimization function. This step ensures that the final output preprocessed image has the best quality, is suitable for subsequent analysis and processing, and optimizes the overall performance of the model;

[0072] Through the above steps, the image preprocessing layer can effectively combine the patient's physical signs and image features, improve the accuracy of image preprocessing, and ensure the accuracy of subsequent femoral artery pulse signal extraction. The multiple iterations and evaluation mechanism of the model also ensure its adaptability to different patients and improve the overall performance of the femoral artery image analysis model.

[0073] Step S2: Femoral artery signal detection

[0074] The femoral artery pulse signal dataset B, all pre-processed target femoral artery images P n 'The signal is input into the femoral artery pulse wave signal perception model for analysis to obtain the femoral artery position perception range; the femoral artery pulse wave signal perception model includes a signal processing layer, a signal feature matching layer, an image overlap layer and a signal range output layer. The filter is constructed based on the improved optimization algorithm to further enhance the femoral artery pulse signal and preliminarily divide the femoral artery range;

[0075] The femoral artery pulse wave signal perception model includes a signal processing layer, a signal feature matching layer, an image overlap layer and a signal range output layer;

[0076] The signal processing layer is used to process the femoral artery pulse signal data B in the femoral artery pulse signal data set B. n Adaptive hybrid filter is used to filter the femoral artery pulse signal data B n Perform denoising to obtain pre-processed femoral artery pulse signal data B n';

[0077] The signal feature matching layer is used to pre-process the femoral artery pulse signal data B n ' and preprocess the target femoral artery image P to be analyzed n 'Perform feature matching to obtain the femoral artery enhanced marked image Z n ;

[0078] The image overlap layer is used to enhance all femoral artery images Z n Perform image overlap to obtain the femoral artery position sensing range;

[0079] The signal range output layer is used to output the femoral artery position sensing range;

[0080] The femoral artery pulse signal is denoised using an adaptive hybrid filter to effectively remove noise from the signal, improve the clarity and accuracy of the signal, and make the pre-processed pulse signal more stable; by matching the denoised pulse signal data with the pre-processed femoral artery image for features, the relevant information of the image is enhanced, and a femoral artery enhanced marked image is obtained, which can improve the correlation between the pulse signal and the image, so that the system can accurately identify the specific location of the femoral artery signal and improve the ability to judge the signal area; all femoral artery enhanced marked images are overlapped to form a perception range of the femoral artery position. This layer integrates information through image superposition, which enhances the positioning effect of the femoral artery position. By overlapping the data of multiple images, accidental errors can be eliminated and the accurate perception of the target area can be improved;

[0081] The specific steps of constructing an adaptive hybrid filter include:

[0082] Construct a randomly generated K filter configuration parameter individual H k , k = 1, 2, ..., K; each filter configuration parameter individual H k contains a set of filter configuration parameters for constructing an adaptive hybrid filter; the K filter configuration parameters H k Combining into a filter configuration parameter iteration population; constructing a signal processing simulation training set;

[0083] Dividing the filter configuration parameter iteration population iteration phase into a filter configuration parameter first iteration phase, a filter configuration parameter second iteration phase and a filter configuration parameter third iteration phase;

[0084] The filter configuration parameters are individual H k Perform simulation calculation to obtain the fitness S k ; The specific steps of simulation calculation are:

[0085] Using the filter configuration parameter individual H kConstruct a simulated adaptive hybrid filter; use the signal processing simulation training set to evaluate the model of the simulated adaptive hybrid filter, and obtain the evaluation result of the simulated adaptive hybrid filter model; use the evaluation result of the simulated adaptive hybrid filter model as the filter configuration parameter individual H k The fitness S k ;

[0086] In the first iteration of the filter configuration parameters, the formula H is used k ′=H k +W*(H k -rand(J)) for the filter configuration parameters individual H k Perform population iteration, H k 'Indicates updating the filter configuration parameter individual H k ', W represents the adaptive adjustment coefficient, rand(J) represents the random solution; and the filter configuration parameter individual H is updated k 'Recalculate the fitness; the specific value of W is set by professional technicians according to actual conditions;

[0087] In the second iteration of the filter configuration parameters, the formula The filter configuration parameters are individual H k To update, G k Denotes the filter configuration parameter individual H k The mutation probability, S k Denotes the filter configuration parameter individual H k The fitness S k , β represents the adaptive mutation parameter; using G k The filter configuration parameters are individual H k Update to get the new filter configuration parameter individual H k , and configure the new filter parameters for the individual H k Recalculate the fitness; the specific value of β is set by professional technicians according to actual conditions;

[0088] In the third iteration phase of the filter configuration parameters, a new filter configuration parameter individual pair H is randomly generated. k Replace it and get the new filter configuration parameter individual H k , and configure the new filter parameters for the individual H k Recalculate fitness;

[0089] When the population iteration ends, the filter configuration parameter individual H corresponding to the maximum output fitness k , that is, the optimal filter configuration parameter individual, and the filter configuration parameters in the optimal filter configuration parameter individual are used to construct an adaptive hybrid filter;

[0090] K filter configuration parameter individuals are randomly generated to ensure the diversity of the initial values ​​of the filter parameters. By constructing an iterative population, the occurrence of local optimal problems is avoided, which helps to improve the global exploration ability in the optimization process and make the final filter configuration parameters more adaptable and robust. This method divides the iterative process into three stages and optimizes it by combining adaptive adjustment, mutation update and random replacement. In the first stage, the coefficient W is adaptively adjusted to make the parameter individuals more random in the early exploration and enhance the global search ability. In the second stage, the mutation probability and adaptive mutation parameter β are introduced to make the individuals gradually converge to the optimal area. In the third stage, new individuals are randomly generated to further increase the diversity of search and optimization effect. After the population iteration is completed, the filter configuration parameter individual with the largest fitness is selected to ensure the optimal performance of the filter. The adaptive hybrid filter finally constructed can dynamically adjust parameters according to signal characteristics, effectively improve the denoising ability and filtering effect of signal processing, and is particularly suitable for processing noise and interference in femoral artery pulse signals.

[0091] Step S3: Femoral artery position determination

[0092] The femoral artery range is determined according to the femoral artery position sensing range and the femoral artery position determination model to obtain the femoral artery position determination result; the next step is performed based on the femoral artery position determination result; the femoral artery position determination model includes a physical sign data acquisition layer, a feature fusion layer, a position determination layer and a position result output layer, which is constructed based on the CNN model and is used to further screen the precise range of the femoral artery;

[0093] The femoral artery location determination model includes a vital sign data acquisition layer, a feature fusion layer, a location determination layer and a location result output layer;

[0094] The vital sign data acquisition layer is used to reacquire the vital sign data of the target patient to obtain updated vital sign data of the target patient; extract features from the updated vital sign data of the target patient to obtain updated vital sign data features of the target patient; through the vital sign data acquisition layer, the latest vital sign data of the target patient can be dynamically reacquired, and feature extraction and fusion are performed in combination with the femoral artery image data. This method ensures that the current individual differences of the patient are always taken into account during the position determination process, increases the accuracy and personalization of femoral artery position judgment, and is suitable for the diverse needs of different patients;

[0095] The feature fusion layer is used to extract the features of the femoral artery position perception range to obtain the femoral artery position perception range features; the updated target patient's vital sign data features and the femoral artery position perception range features are feature fused to obtain the fused femoral artery position perception range features; the feature fusion layer fuses the patient's vital sign data with the femoral artery position perception range features to obtain a more comprehensive fusion feature; by combining vital sign and image information, the model can perform position screening in multiple dimensions, further improving the accuracy of femoral artery position determination and avoiding errors caused by single feature data;

[0096] The position determination layer is used to further determine the position according to the fused femoral artery position sensing range feature and the femoral artery position sensing range, and obtain the femoral artery position determination result;

[0097] The position result output layer is used to output the femoral artery position determination result;

[0098] The specific steps for training the location determination layer include:

[0099] Collecting several groups of femoral artery position determination training samples, each group of femoral artery position determination training samples includes training patient physical sign data, an initial femoral artery determination range and a corresponding final femoral artery determination range; using the final femoral artery determination range as a target value of the femoral artery position determination training sample; combining several groups of femoral artery position determination training samples to obtain a femoral artery position determination training set;

[0100] The femoral artery position determination training set is input into the femoral artery position determination model to train the position determination layer with the target value as the target, so as to obtain an initial position determination layer; the initial position determination layer is model evaluated, and if the initial position determination layer passes the model evaluation, the initial position determination layer is used as the position determination layer in the femoral artery position determination model; otherwise, the model training is continued using the femoral artery position determination training set;

[0101] By constructing a training set for femoral artery location determination and conducting iterative training, the model can be continuously optimized to ensure its robustness in different situations; the model evaluation mechanism ensures that the trained location determination layer has sufficient accuracy and stability, and models that fail the evaluation can continue to be trained to ensure that the final results meet the expected standards; through multi-level feature extraction, fusion, and CNN-based screening, the specific location of the femoral artery can be accurately determined; the dynamic update of vital sign data combined with multi-dimensional feature fusion ensures the personalization and accuracy of the model. Through continuous training and evaluation mechanisms, the model can provide efficient and accurate femoral artery location determination results in practical applications, which helps to facilitate the smooth progress of subsequent medical or testing operations.

[0102] Example 2, a femoral artery position determination device, see Figure 1 As shown, including:

[0103] The femoral artery data acquisition module includes a physical sign data acquisition unit and a femoral artery image data acquisition unit; the physical sign data acquisition unit is used to acquire the physical sign data of the target patient; the femoral artery image data acquisition unit is used to acquire the target femoral artery image set P to be analyzed, P = {P 1 , P 2 ,…,P n ,…,P N}; Among them, P n Represented as the target femoral artery image P to be analyzed in the target femoral artery image set P to be analyzed n , N is the total number of target femoral artery images in a set of target femoral artery images to be analyzed P;

[0104] The femoral artery signal detection module includes a femoral artery image processing unit and a signal sensing unit; the femoral artery image processing unit is used to input the target femoral artery image set P to be analyzed and the target patient's vital signs data into the femoral artery image reflected light intensity analysis model for analysis, and obtain the femoral artery pulse signal data set B and all pre-processed target femoral artery images P to be analyzed. n ', B = {B 1 , B 2 , …, B n , …, B N The femoral artery image reflected light intensity analysis model includes an image preprocessing layer, a regional analysis layer, a signal extraction layer and a result output layer, which is constructed based on a CNN model and a multi-layer perceptron and is used to analyze the femoral artery pulse signal of the target femoral artery image set P to be analyzed in combination with the target patient's vital sign data; the signal perception unit is used to combine the femoral artery pulse signal data set B, all preprocessed target femoral artery images P to be analyzed n 'The signal is input into the femoral artery pulse wave signal perception model for analysis to obtain the femoral artery position perception range; the femoral artery pulse wave signal perception model includes a signal processing layer, a signal feature matching layer, an image overlap layer and a signal range output layer. The filter is constructed based on the improved optimization algorithm to further enhance the femoral artery pulse signal and preliminarily divide the femoral artery range;

[0105] The femoral artery position determination module includes a position determination unit; the position determination unit is used to determine the femoral artery range according to the femoral artery position perception range and the femoral artery position determination model to obtain the femoral artery position determination result; the next step is performed based on the femoral artery position determination result; the femoral artery position determination model includes a vital sign data acquisition layer, a feature fusion layer, a position determination layer and a position result output layer, which is constructed based on the CNN model and is used to further screen the precise range of the femoral artery.

[0106] 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 determining the position of a femoral artery, characterized in that: The following steps are involved: Step S1: Femoral artery data acquisition Obtain the target femoral artery image set P to be analyzed, P = {P1, P2, ..., P n ,…,P N }; Among them, P n Represented as the target femoral artery image P to be analyzed in the target femoral artery image set P to be analyzed n , N is the total number of target femoral artery images in a set of target femoral artery images to be analyzed P; obtain target patient vital sign data; input the target femoral artery image set P to be analyzed and the target patient vital sign data into the femoral artery image reflected light intensity analysis model for analysis, and obtain the femoral artery pulse signal data set B and all preprocessed target femoral artery images to be analyzed P n ', B = {B1, B2, ..., B n , …, B N The femoral artery image reflected light intensity analysis model includes an image preprocessing layer, a regional analysis layer, a signal extraction layer and a result output layer, which is constructed based on a CNN model and a multi-layer perceptron, and is used to analyze the femoral artery pulse signal of the target femoral artery image set P to be analyzed in combination with the target patient's vital sign data; Step S2: Femoral artery signal detection The femoral artery pulse signal dataset B, all pre-processed target femoral artery images P n 'The signal is input into the femoral artery pulse wave signal perception model for analysis to obtain the femoral artery position perception range; the femoral artery pulse wave signal perception model includes a signal processing layer, a signal feature matching layer, an image overlap layer and a signal range output layer. The filter is constructed based on the improved optimization algorithm to further enhance the femoral artery pulse signal and preliminarily divide the femoral artery range; Step S3: Femoral artery position determination The femoral artery range is determined according to the femoral artery position perception range and the femoral artery position determination model to obtain the femoral artery position determination result; the next step is performed based on the femoral artery position determination result; the femoral artery position determination model includes a vital sign data acquisition layer, a feature fusion layer, a position determination layer and a position result output layer, which is constructed based on the CNN model and is used to further screen the precise range of the femoral artery.

2. A method for determining the position of a femoral artery according to claim 1, characterized in that: The femoral artery image reflected light intensity analysis model in step S1 includes an image preprocessing layer, a regional analysis layer, a signal extraction layer and a result output layer; The image preprocessing layer is used to process the target femoral artery image P to be analyzed according to the target patient's vital sign data and the target femoral artery image P to be analyzed in the target femoral artery image set P n Perform feature analysis to obtain the preprocessed target femoral artery image P n '; The regional analysis layer is used to pre-process the target femoral artery image P to be analyzed. n 'Identify the signal extraction area and obtain the femoral artery signal extraction area R n ; The signal extraction layer is used to extract region R based on the femoral artery signal n Perform signal recognition to obtain femoral artery pulse signal data B n ; The resulting output layer is used to convert all femoral artery pulse signal data B n The femoral artery pulse signal data set B is obtained by combining, and the femoral artery pulse signal data set B is output.

3. A method for determining the position of a femoral artery according to claim 2, characterized in that: The specific steps to build the image preprocessing layer include: In the image preprocessing layer, four image preprocessing adjustment and analysis layers are constructed; In the first image preprocessing adjustment and analysis layer, a multi-layer perceptron is used to construct several fully connected layers for extracting high-dimensional representations of patient vital sign data to obtain patient vital sign data features; in the second image preprocessing adjustment and analysis layer, a CNN network is used to construct several convolutional layers for extracting image data features of patient image data to obtain patient image data features; in the third image preprocessing adjustment and analysis layer, a neural network structure is used to construct several embedding layers for splicing patient vital sign data features and patient image data features to obtain patient fusion feature vectors; the patient fusion feature vectors are input into the trained contrast loss optimization function for calculation to obtain optimized patient fusion feature vectors; in the fourth image preprocessing adjustment and analysis layer, a multi-layer perceptron is used to generate image adjustment parameters according to the optimized patient fusion feature vectors, and the input patient image data is adjusted using the image adjustment parameters to obtain preprocessed patient image data; The specific steps for training the image preprocessing layer include: Collecting several groups of image preprocessing training samples; each group of image preprocessing training samples contains sample input values ​​and sample output values; wherein the sample input values ​​are patient vital sign data and femoral artery images of patients to be processed, and the sample output values ​​are preprocessed femoral artery images of patients; combining several groups of image preprocessing training samples to obtain an image preprocessing training set; The image preprocessing training set is input into the image preprocessing layer to perform model training with the sample output value as the target, and the image preprocessing layer to be evaluated is obtained; image evaluation is performed on the image preprocessing layer to be evaluated to obtain the image preprocessing layer model evaluation result; if the image preprocessing layer model evaluation result is passed, the image preprocessing layer to be evaluated is used as the image preprocessing layer in the femoral artery image reflected light intensity analysis model; otherwise, the image preprocessing training set is used to continue model training.

4. A method for determining the position of a femoral artery according to claim 3, characterized in that: The femoral artery pulse wave signal perception model includes a signal processing layer, a signal feature matching layer, an image overlap layer and a signal range output layer; The signal processing layer is used to process the femoral artery pulse signal data B in the femoral artery pulse signal data set B. n Adaptive hybrid filter is used to filter the femoral artery pulse signal data B n Perform denoising to obtain pre-processed femoral artery pulse signal data B n '; The signal feature matching layer is used to pre-process the femoral artery pulse signal data B n ' and preprocess the target femoral artery image P to be analyzed n 'Perform feature matching to obtain the femoral artery enhanced marked image Z n ; The image overlap layer is used to enhance all femoral artery images Z n Perform image overlap to obtain the femoral artery position sensing range; The signal range output layer is used to output the femoral artery position perception range.

5. A method for determining the position of a femoral artery according to claim 4, characterized in that: The specific steps of constructing an adaptive hybrid filter include: Construct a randomly generated K filter configuration parameter individual H k , k = 1, 2, ..., K; each filter configuration parameter individual H k contains a set of filter configuration parameters for constructing an adaptive hybrid filter; the K filter configuration parameters H k Combining into a filter configuration parameter iteration population; constructing a signal processing simulation training set; Dividing the filter configuration parameter iteration population iteration phase into a filter configuration parameter first iteration phase, a filter configuration parameter second iteration phase and a filter configuration parameter third iteration phase; The filter configuration parameters are individual H k Perform simulation calculation to obtain the fitness S k ; The specific steps of simulation calculation are: Using the filter configuration parameter individual H k Construct a simulated adaptive hybrid filter; use the signal processing simulation training set to evaluate the model of the simulated adaptive hybrid filter, and obtain the evaluation result of the simulated adaptive hybrid filter model; use the evaluation result of the simulated adaptive hybrid filter model as the filter configuration parameter individual H k The fitness S k ; In the first iteration of the filter configuration parameters, the formula H is used k ′=H k +W*(H k -rand(J)) for the filter configuration parameters individual H k Perform population iteration, H k 'Indicates updating the filter configuration parameter individual H k ', W represents the adaptive adjustment coefficient, rand(J) represents the random solution; and the filter configuration parameter individual H is updated k 'Recalculate fitness; In the second iteration of filter configuration parameters, the formula The filter configuration parameters are individual H k To update, G k Denotes the filter configuration parameter individual H k The mutation probability, S k Denotes the filter configuration parameter individual H k The fitness S k , β represents the adaptive mutation parameter; using G k The filter configuration parameters are individual H k Update to get the new filter configuration parameter individual H k , and configure the new filter parameters for the individual H k Recalculate fitness; In the third iteration phase of the filter configuration parameters, a new filter configuration parameter individual pair H is randomly generated. k Replace it and get the new filter configuration parameter individual H k , and configure the new filter parameters for the individual H k Recalculate fitness; When the population iteration ends, the filter configuration parameter individual H corresponding to the maximum output fitness k , which is the optimal filter configuration parameter individual, and the filter configuration parameters in the optimal filter configuration parameter individual are used to construct an adaptive hybrid filter.

6. A method for determining the position of a femoral artery according to claim 5, characterized in that: The femoral artery location determination model includes a vital sign data acquisition layer, a feature fusion layer, a location determination layer and a location result output layer; The vital sign data acquisition layer is used to reacquire the vital sign data of the target patient to obtain updated vital sign data of the target patient; Extracting features of the updated target patient's vital sign data to obtain updated target patient's vital sign data features; The feature fusion layer is used to extract the features of the femoral artery position perception range to obtain the femoral artery position perception range features; The updated target patient's vital sign data features and the femoral artery position sensing range features are fused to obtain a fused femoral artery position sensing range feature; The position determination layer is used to further determine the position according to the fused femoral artery position sensing range feature and the femoral artery position sensing range, and obtain the femoral artery position determination result; The position result output layer is used to output the femoral artery position determination results.

7. A method for determining the position of a femoral artery according to claim 6, characterized in that: The specific steps for training the location determination layer include: Collecting several groups of femoral artery position determination training samples, each group of femoral artery position determination training samples includes training patient physical sign data, an initial femoral artery determination range and a corresponding final femoral artery determination range; using the final femoral artery determination range as a target value of the femoral artery position determination training sample; combining several groups of femoral artery position determination training samples to obtain a femoral artery position determination training set; The femoral artery position determination training set is input into the femoral artery position determination model to train the position determination layer with the target value as the target to obtain the initial position determination layer; the initial position determination layer is evaluated, and if the initial position determination layer passes the model evaluation, the initial position determination layer is used as the position determination layer in the femoral artery position determination model; otherwise, the femoral artery position determination training set is used to continue model training.

8. A femoral artery position determination device, characterized in that: The device uses a femoral artery position determination method according to any one of claims 1 to 7, comprising: The femoral artery data acquisition module includes a physical sign data acquisition unit and a femoral artery image data acquisition unit; the physical sign data acquisition unit is used to acquire the physical sign data of the target patient; the femoral artery image data acquisition unit is used to acquire the target femoral artery image set P to be analyzed, P = {P1, P2, ..., P n ,…,P N }; Among them, P n Represented as the target femoral artery image P to be analyzed in the target femoral artery image set P to be analyzed n , N is the total number of target femoral artery images in a set of target femoral artery images to be analyzed P; The femoral artery signal detection module includes a femoral artery image processing unit and a signal sensing unit; the femoral artery image processing unit is used to input the target femoral artery image set P to be analyzed and the target patient's vital signs data into the femoral artery image reflected light intensity analysis model for analysis, and obtain the femoral artery pulse signal data set B and all pre-processed target femoral artery images P to be analyzed. n ', B = {B1, B2, ..., B n , …, B N The femoral artery image reflected light intensity analysis model includes an image preprocessing layer, a regional analysis layer, a signal extraction layer and a result output layer, which is constructed based on a CNN model and a multi-layer perceptron and is used to analyze the femoral artery pulse signal of the target femoral artery image set P to be analyzed in combination with the target patient's vital sign data; the signal perception unit is used to combine the femoral artery pulse signal data set B, all preprocessed target femoral artery images P to be analyzed n 'The signal is input into the femoral artery pulse wave signal perception model for analysis to obtain the femoral artery position perception range; the femoral artery pulse wave signal perception model includes a signal processing layer, a signal feature matching layer, an image overlap layer and a signal range output layer. The filter is constructed based on the improved optimization algorithm to further enhance the femoral artery pulse signal and preliminarily divide the femoral artery range; The femoral artery position determination module includes a position determination unit; the position determination unit is used to determine the femoral artery range according to the femoral artery position perception range and the femoral artery position determination model to obtain the femoral artery position determination result; the next step is performed based on the femoral artery position determination result; the femoral artery position determination model includes a vital sign data acquisition layer, a feature fusion layer, a position determination layer and a position result output layer, which is constructed based on the CNN model and is used to further screen the precise range of the femoral artery.

Images