Limb thrombus monitoring device and method based on arteriovenous recognition

Abstract

The invention provides a limb thrombus monitoring device and method based on arteriovenous recognition, and the device comprises a monitoring management middle platform, a multi-modal data collection module, a vein recognition module, a feature extraction module, and a risk assessment module. The monitoring management middle platform is connected with the multi-modal data acquisition module, the vein recognition module, the feature extraction module and the risk assessment module, and manages each unit; the multi-modal data acquisition module is used for acquiring image data of limb parts of a patient; and the vein recognition module is used for recognizing and positioning the limb vein position of the patient based on the near-infrared imaging data to obtain vein position information. The accuracy of thrombus monitoring, evaluation and judgment can be improved, and the problem that the thrombus monitoring process mainly depends on clinical experience of medical staff for judgment is avoided. Therefore, the thrombus monitoring and judging efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of arteriovenous recognition, and particularly to a limb thrombus monitoring device and method based on arteriovenous recognition. Background Art

[0002] In the field of modern intravenous therapy, for patients who need long-term infusion or infusion of hypertonic and irritating drugs, peripherally inserted central catheters (PICCs) or midline catheters (MCs) are often used clinically to establish vascular access. During the indwelling period of such catheters, the incidence of deep vein thrombosis (DVT) is as high as 15% - 30%, and the pulmonary embolism (PE) caused by it is a serious complication threatening the life safety of patients. Therefore, during the indwelling period, patients need to be regularly monitored every day for the formation of thrombus to avoid the occurrence of complications.

[0003] However, currently, during the process of medical staff's monitoring and judgment of limb thrombus in clinical practice, medical staff mainly rely on clinical manifestations such as Homan sign and limb swelling degree for preliminary screening, and then perform imaging examinations after the screening is qualified. As a result, the entire monitoring and judgment process mainly relies on the clinical experience of medical staff, resulting in low efficiency and unstable accuracy in the monitoring and judgment of the limb thrombus formation process. Summary of the Invention

[0004] The present invention provides a limb thrombus monitoring device and method based on arteriovenous recognition, aiming to improve the efficiency and accuracy of thrombus monitoring and judgment.

[0005] In a first aspect, the present invention provides a limb thrombus monitoring device based on arteriovenous recognition, including a monitoring and management middle platform, a multimodal data acquisition module, a venous blood vessel recognition module, a feature extraction module, and a risk assessment module. The monitoring and management middle platform is respectively connected to the multimodal data acquisition module, the venous blood vessel recognition module, the feature extraction module, and the risk assessment module to manage each unit;

[0006] The multimodal data acquisition module is used to acquire image data of the patient's limb part; the image data includes near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data;

[0007] The venous blood vessel recognition module is used to identify and locate the position of the patient's limb vein based on the near-infrared imaging data to obtain venous blood vessel position information;

[0008] The feature extraction module is used to extract features from the ultrasonic imaging data and the infrared thermal imaging data based on the venous blood vessel position information to respectively obtain blood flow velocity data and target temperature data at the venous blood vessel;

[0009] A risk assessment module, configured to input the blood flow velocity data and the target temperature data into a thrombus risk assessment model, obtain a risk assessment result output by the thrombus risk assessment model, and send the risk assessment result to a terminal device; the thrombus risk assessment model is trained based on sample data of blood flow velocity and temperature and risk assessment label results.

[0010] In a second aspect, the present invention further provides a method for monitoring limb thrombus based on arteriovenous identification, which is implemented based on the device for monitoring limb thrombus based on arteriovenous identification described in the first aspect. The method for monitoring limb thrombus based on arteriovenous identification includes:

[0011] Obtain image data of a patient's limb part; the image data includes near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data;

[0012] Based on the near-infrared imaging data, identify and locate the position of the patient's limb veins to obtain venous blood vessel position information;

[0013] Based on the venous blood vessel position information, extract features from the ultrasonic imaging data and the infrared thermal imaging data to respectively obtain blood flow velocity data and target temperature data at the venous blood vessels;

[0014] Input the blood flow velocity data and the target temperature data into a thrombus risk assessment model, obtain a risk assessment result output by the thrombus risk assessment model, and send the risk assessment result to a terminal device; the thrombus risk assessment model is trained based on sample data of blood flow velocity and temperature and risk assessment label results.

[0015] In a third aspect, the present invention further provides an electronic device, including: a memory for storing a computer software program; a processor for reading and executing the computer software program to implement the method for monitoring limb thrombus based on arteriovenous identification as described in any one of the above.

[0016] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium, in which a computer software program is stored, and when the computer software program is executed by a processor, it implements the method for monitoring limb thrombus based on arteriovenous identification as described in any one of the above.

[0017] In a fifth aspect, the present invention further provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the method for monitoring limb thrombus based on arteriovenous identification as described in any one of the above.

[0018] The limb thrombus monitoring device based on arteriovenous recognition provided by the embodiments of the present invention can obtain more accurate blood flow velocity data and target temperature data through the complementary fusion of various data such as near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data. Moreover, the entire process is implemented by functional modules, thus reducing human interference factors and improving the monitoring efficiency. Subsequently, using the more accurate blood flow velocity data and target temperature data, the thrombus risk is evaluated through a thrombus risk assessment model, which can improve the accuracy of thrombus monitoring and assessment judgment, and avoid the problem that the thrombus monitoring process mainly relies on the clinical experience of medical staff for judgment. Therefore, the embodiments of the present invention improve the efficiency and accuracy of thrombus monitoring and judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a schematic structural diagram of the limb thrombus monitoring device based on arteriovenous recognition provided by the present invention;

[0020] Figure 2 is a schematic flowchart of the limb thrombus monitoring method based on arteriovenous recognition provided by the present invention;

[0021] Figure 3 is an embodiment diagram of the electronic device provided by the embodiments of the present invention;

[0022] Figure 4 is an embodiment diagram of the computer-readable storage medium provided by the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0024] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0025] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to implement and use the present invention. In the following description, details are set forth for purposes of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but rather to be in line with the broadest scope consistent with the principles and features disclosed in the present invention.

[0026] Optionally, referring to Figure 1 as shown in Figure 1 is a schematic structural diagram of a limb thrombus monitoring device based on arteriovenous recognition provided by the present invention. The limb thrombus monitoring device based on arteriovenous recognition includes a monitoring and management middle platform, a multimodal data acquisition module, a venous blood vessel recognition module, a feature extraction module, and a risk assessment module. The monitoring and management middle platform is respectively connected to the multimodal data acquisition module, the venous blood vessel recognition module, the feature extraction module, and the risk assessment module to manage each unit.

[0027] Optionally, the monitoring and management middle platform has a large-capacity and highly reliable data storage function, which can classify and store near-infrared imaging data, ultrasonic imaging data, infrared thermal imaging data, etc. collected by the multimodal data acquisition module. At the same time, it can also perform cloud backup of the stored data through networking. In addition, the monitoring and management middle platform can intelligently schedule the multimodal data acquisition module to collect data at appropriate time nodes according to the preset monitoring plan and the actual situation of the patient. For example, for high-risk patients, increase the data acquisition frequency; for patients with stable conditions, appropriately reduce the acquisition frequency.

[0028] Optionally, the monitoring and management middle platform has a visual interface to provide intuitive monitoring data, and display the various monitoring data of the patient in the form of charts and images, which is convenient for medical staff to quickly understand the patient's condition and its changes.

[0029] Furthermore, the multimodal data acquisition module is used to obtain image data of the patient's limb part; the image data includes near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data. Among them, the multimodal data acquisition module can adopt an integrated design, integrating the near-infrared imaging sensor, ultrasonic imaging probe, and infrared thermal imaging sensor into one body. Moreover, multiple groups of near-infrared imaging sensors are set to be used for collaborative monitoring from multiple angles. The near-infrared imaging sensor can select models with high resolution and wide dynamic range. For example, the pixel of a certain model sensor can reach 5 million, and it can clearly capture the fine structure of blood vessels. The ultrasonic imaging probe adopts products with high frequency (such as 10 MHz) and high precision, and can accurately measure blood flow parameters. The infrared thermal imaging sensor has high sensitivity and can detect temperature changes of 0.1 °C. In actual operation, medical staff aim the monitoring device at the part of the patient's limb to be monitored. After turning it on, each sensor works synchronously to obtain near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data simultaneously. This reduces the human interference factors in the data acquisition process and improves work efficiency.

[0030] Furthermore, the venous vessel identification module is used to identify and locate the position of the patient's limb veins based on the near-infrared imaging data, and obtain the venous vessel position information. Among them, when monitoring a patient with an indwelling PICC catheter, the PICC catheter is set in the limb part and is connected to the venous vessels. Therefore, by accurately identifying and locating the venous vessel position information, the venous vessel identification module can provide accurate positioning assistance for judging whether there is a thrombus formation at the venous vessels in the subsequent process, so as to improve the accuracy of monitoring and also provide a reliable basis for the subsequent data fusion.

[0031] Furthermore, the feature extraction module is used to extract features from the ultrasonic imaging data and infrared thermal imaging data based on the venous vessel position information, and respectively obtain the blood flow velocity data and target temperature data at the venous vessels. When a thrombus forms in the venous vessels, it will obstruct and interfere with the blood flow velocity at the venous vessels, causing the blood flow velocity at the venous vessels to decrease significantly. At the same time, in the initial stage of thrombus formation, the temperature of the tissues around the venous vessels may have abnormal changes. Therefore, through the accurately positioned venous vessel position information, more accurate blood flow velocity data and target temperature data can be obtained, thereby improving the reliability of the feature data to facilitate the subsequent risk judgment of thrombus formation.

[0032] Further, the risk assessment module is used to input the blood flow velocity data and the target temperature data into a thrombus risk assessment model, obtain the risk assessment result output by the thrombus risk assessment model, and send the risk assessment result to the terminal device; the thrombus risk assessment model is trained based on the sample data of blood flow velocity and temperature and the risk assessment label results. Among them, the blood flow velocity data and the target temperature data obtained by the feature extraction module are used as inputs and input into a pre-trained thrombus risk assessment model. The thrombus risk assessment model can adopt model architectures such as deep neural networks, LSTM neural networks, and support vector machines (SVM), and is trained through a large amount of sample data of blood flow velocity and temperature and the corresponding risk assessment label results. Finally, based on accurate and reliable blood flow velocity data and target temperature data, a more accurate risk assessment result is obtained, and the problem of relying mainly on the clinical experience of medical staff for judgment in the past is avoided throughout the process.

[0033] Optionally, in one embodiment, if in the obtained risk assessment result, the blood flow velocity is lower than the normal range and the target temperature is higher than the normal threshold, the model will output a corresponding high-risk assessment result according to the trained weights and classification boundaries. Finally, the risk assessment result is sent to the terminal device (such as a tablet computer or mobile phone of medical staff) through a wireless communication module (such as Bluetooth, Wi-Fi), so as to remind medical staff to pay attention to the thrombus risk of patients and improve medical efficiency and patient safety.

[0034] In the embodiment of the present invention, through the mutual complementation and fusion of various data such as infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data, more accurate blood flow velocity data and target temperature data can be obtained, and the entire process is realized by functional modules, so the human interference factors are reduced and the monitoring efficiency is improved; then, using more accurate blood flow velocity data and target temperature data, the thrombus risk is evaluated by the thrombus risk assessment model, which can improve the accuracy of thrombus monitoring and assessment judgment, and avoid the problem of mainly relying on the clinical experience of medical staff for judgment during the thrombus monitoring process. Therefore, the embodiment of the present invention improves the efficiency and accuracy of thrombus monitoring and judgment.

[0035] Optionally, referring to Figure 2 , Figure 2 is a schematic flowchart of the limb thrombus monitoring method based on arteriovenous identification provided by the present invention. In the embodiment of the present invention, the execution subject of the limb thrombus monitoring method based on arteriovenous identification is a limb thrombus monitoring device. Therefore, the limb thrombus monitoring method based on arteriovenous identification includes:

[0036] Step 10, obtaining image data of the patient's limb part; the image data includes near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data.

[0037] Optionally, the limb thrombus monitoring device in the embodiments of the present invention, according to a preset monitoring instruction, monitors and acquires image data of a patient's limb in real time. Among them, the near-infrared imaging data is acquired by a near-infrared imaging sensor, and the near-infrared sensor is installed and set from multiple angles, so that when monitoring the patient's limb, near-infrared imaging data of the patient's limb can be acquired from multiple angles. The near-infrared imaging data includes near-infrared imaging images. The near-infrared imaging sensor can be selected from models with high resolution and wide dynamic range. For example, the pixel of a certain model sensor can reach 5 million, and it can clearly capture the fine structure of blood vessels. The ultrasonic imaging data is acquired by an ultrasonic imaging probe. The ultrasonic imaging probe uses a product with a high frequency (such as 10 MHz) and high precision, and can accurately measure blood flow parameters. And the actual working frequency can also be adjusted according to different conditions of the patient (such as obesity degree, blood vessel depth). The ultrasonic imaging data includes ultrasonic imaging images and corresponding ultrasonic imaging device parameters. The infrared thermal imaging data is acquired by an infrared thermal imaging sensor. The infrared thermal imaging sensor has high sensitivity and can detect a temperature change of 0.1 °C. By sensing the infrared radiation emitted from the human body surface, a temperature distribution image is generated. The infrared thermal imaging data includes infrared thermal imaging images.

[0038] Further, in one embodiment, when monitoring a patient with a mid-long catheter indwelling, the monitoring device is placed near the cubital fossa of the patient's upper limb (i.e., aligned with the indwelling mid-long catheter). The near-infrared imaging data can clearly present the general trend of the venous blood vessels, the ultrasonic imaging data can present the blood flow velocity state in the blood vessels in real time, and the infrared thermal imaging data can record the skin surface temperature distribution, so that comprehensive multi-modal image data can be obtained in one operation.

[0039] Step 20: Identify and locate the venous position of the patient's limb based on the near-infrared imaging data to obtain venous blood vessel position information.

[0040] Optionally, the limb thrombus monitoring device accurately identifies and locates the venous position of the patient's limb according to the acquired near-infrared imaging data, and cooperates with the three-dimensional gradient tensor field based on the near-infrared imaging data obtained from multiple perspectives to obtain venous blood vessel position information in three-dimensional space, so as to improve the accuracy of obtaining the position, as specifically described in Steps 201 - 205.

[0041] Step 30: Extract features from the ultrasonic imaging data and the infrared thermal imaging data based on the venous blood vessel position information to obtain blood flow velocity data and target temperature data at the venous blood vessels respectively.

[0042] Optionally, based on the determined venous blood vessel position information, on the one hand, when extracting features from the ultrasonic imaging data, the limb thrombus monitoring device can extract the area in the ultrasonic imaging data that is also located at the position of the venous blood vessel position information, and then analyze and process the extracted area to finally obtain more accurate blood flow velocity data at the venous blood vessel, as specifically described in steps 3011 - 3015. On the other hand, when extracting features from the infrared thermal imaging data, the device can also extract the area in the infrared thermal imaging data that is located at the position of the venous blood vessel position information, and then perform analysis and processing on the extracted area in the same way to finally obtain more accurate target temperature data at the venous blood vessel, as specifically described in steps 3021 - 3025.

[0043] Step 40: Input the blood flow velocity data and the target temperature data into the thrombus risk assessment model to obtain the risk assessment result output by the thrombus risk assessment model, and send the risk assessment result to the terminal device; the thrombus risk assessment model is trained based on the sample data of blood flow velocity and temperature and the risk assessment label results.

[0044] Optionally, based on the obtained blood flow velocity data and target temperature data at the venous blood vessel, the limb thrombus monitoring device uses the blood flow velocity data and the target temperature data as the input of the thrombus risk assessment model. Inside the thrombus risk assessment model, the model will output corresponding risk assessment results according to the trained weights and classification boundaries. These include no risk assessment result, low risk assessment result, medium risk assessment result, high risk assessment result, etc. For example, for patients with PICC catheters, if the detected blood flow abnormally decelerates to <5 cm / s and the temperature rises by 0.5 °C, a high risk assessment result is determined, triggering a high risk warning, and the high risk assessment result is sent to the terminal device. The risk assessment result is also displayed on the display interface of the limb thrombus monitoring device. At the same time, when the low risk assessment result, medium risk assessment result, and high risk assessment result are output, the risk assessment result is sent to the terminal device, such as the tablet computer or mobile phone of medical staff, via a high-speed wireless communication module such as Bluetooth 5.0 or Wi-Fi 6, so that medical staff can view the risk assessment result in a timely manner and be reminded to pay attention to the patient's thrombus risk.

[0045] Furthermore, the thrombus risk assessment model adopts the support vector machine (SVM) algorithm. When training with a large amount of sample data of blood flow velocity and temperature and the corresponding risk assessment label results, during the training process, the values of the weight vector and the bias term are determined by minimizing the loss function. In actual application, the real-time collected blood flow velocity and target temperature data are standardized to meet the model input requirements, and then input into the model for calculation to obtain the risk assessment result.

[0046] In the embodiment of the present invention, various data such as infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data are mutually complementary and fused, so as to obtain more accurate blood flow velocity data and target temperature data. Moreover, the whole process is realized by functional modules, thus reducing human interference factors and improving the monitoring efficiency. Then, using the more accurate blood flow velocity data and target temperature data, the thrombus risk is evaluated by the thrombus risk assessment model, which can improve the accuracy of thrombus monitoring and assessment judgment, and avoid the problem that the thrombus monitoring process mainly relies on the clinical experience of medical staff for judgment. Therefore, the embodiment of the present invention improves the efficiency and accuracy of thrombus monitoring and judgment.

[0047] In one embodiment, the descriptions of steps 201 - 205 are as follows:

[0048] Step 201: Perform image enhancement on the near-infrared imaging data to obtain a target enhanced image.

[0049] Optionally, the near-infrared imaging data consists of near-infrared imaging images. When the limb thrombus monitoring device determines the venous vessel position information according to the near-infrared imaging images, since when the near-infrared imaging images are obtained, the near-infrared imaging images contain venous vessel information, surrounding tissue information, and indwelling catheter end information, and when the monitoring clarity is also related to the patient's fat layer thickness, when the patient's fat layer is relatively thick, it will also interfere with the near-infrared imaging images. Therefore, in order to improve the clarity of the venous vessels in the near-infrared imaging images, the limb thrombus monitoring device needs to perform image enhancement processing on the near-infrared imaging data, specifically as described in steps 2011 - 2014.

[0050] Step 202: Perform spatial gradient processing on the target enhanced image to obtain a gradient vector, and construct a three-dimensional gradient tensor field based on the gradient vector and a preset tensor voting function.

[0051] Optionally, after the limb thrombus monitoring device performs image enhancement on the near-infrared imaging data to obtain a target enhanced image, it performs spatial gradient processing on the target enhanced image to obtain the direction and amplitude information of the pixel gray change in the image, and then determines the gradient vector according to the direction and amplitude information of the image pixels, so that the gradient vector after spatial gradient processing can highlight the edge information in the image. In the process of constructing the three-dimensional gradient tensor field, the tensor voting function is introduced as where represents the coordinate of the current pixel point, represents the coordinate of the neighborhood pixel point, d represents the distance between the current pixel point and the neighborhood pixel point, The kernel function represented as anisotropic usually takes the form of a Gaussian function. Tensor voting calculations are performed on each pixel point in the target enhanced image, and finally a three-dimensional gradient tensor field of the entire image is constructed. By comprehensively considering the gradient relationship and spatial position relationship between pixels, it can more comprehensively describe the structural characteristics of venous blood vessels in the image, providing a richer information basis for accurately extracting the vascular skeleton subsequently and enhancing the adaptability of the algorithm to complex images.

[0052] Further, in one embodiment, when performing spatial gradient processing, an improved Sobel operator is used for spatial gradient calculation. The templates of the traditional Sobel operator in the x and y directions are respectively and When calculating the gradient magnitude G of the improved Sobel operator, considering the neighborhood correlation of image pixels, a weighted summation method is adopted. Let the gradient of the image in the x direction be G x , and the gradient in the y direction be G y . The improved gradient magnitude calculation formula is where a0 and b0 are the weight coefficients determined by the neighborhood pixel correlation and are optimized through experimental or machine learning methods. According to the gradient of the image in the x direction being G x and the gradient in the y direction being G y , the gradient vector

[0053] Further, in one embodiment, to construct a three-dimensional gradient tensor field, assume that a pixel point in the image is p0, where the gradient necklace is For other surrounding pixel points q0, the voting tensor field of pixel point p0 for q0 is calculated according to the tensor voting function as The distance between pixel points p0 and q0 is d. Then, through tensor voting calculation operations on each pixel point in the image, a three-dimensional gradient tensor field is finally constructed.

[0054] Step 203, perform eigenvalue decomposition based on the three-dimensional gradient tensor field to obtain vascular skeleton information.

[0055] Optionally, the limb thrombus monitoring device performs eigenvalue decomposition on the constructed three-dimensional gradient tensor field T, where T sz is represented as Solve the characteristic equation |T - λI T | = 0 to obtain three eigenvalues λ1 ≥ λ2 ≥ λ3, where I T is the identity matrix. Then, judge the vascular skeleton points, and the vascular skeleton points are determined by the ratio of the eigenvalues. For example, when and When this is the case, it is determined that the point is considered to be located on the blood vessel skeleton. This is because on the blood vessel centerline, the eigenvalue of the gradient tensor in the blood vessel direction is larger, while the two eigenvalues in the direction perpendicular to the blood vessel are relatively small and close. Each point in the three-dimensional gradient tensor field is traversed, and all blood vessel skeleton points are filtered according to the above conditions, and then the blood vessel skeleton is obtained.

[0056] Step 204: Perform perspective parallax processing on multiple target enhanced images to obtain three-dimensional depth information values; the multiple target enhanced images are near-infrared images from different perspectives.

[0057] Optionally, the limb thrombus monitoring device classifies the target enhanced images from different perspectives. Then, according to the principle of triangulation and the pixel point information of the target enhanced images collected from different perspectives, such as by performing such calculations on all corresponding pixel points in the two images, the three-dimensional depth information value distribution of the entire image area is obtained. Specifically, as described in steps 2041 - 2045, it is possible to utilize the information complementarity of the images from different perspectives to restore the depth position of the venous blood vessels in the three-dimensional space, providing key depth dimension information for accurately positioning the venous blood vessels in the three-dimensional space, making up for the limitations of only analyzing two-dimensional images, and improving the spatial accuracy of venous blood vessel positioning.

[0058] Step 205: Integrate the blood vessel skeleton information with the three-dimensional depth information values to obtain the venous blood vessel position information in the three-dimensional space.

[0059] Optionally, the limb thrombus monitoring device integrates the obtained blood vessel skeleton information with the three-dimensional depth information values, and a fusion method based on coordinate transformation can be adopted. Combine the two-dimensional coordinates (x, y) of the blood vessel skeleton points with the corresponding three-dimensional depth information value Z, and convert it to a unified three-dimensional space coordinate system through coordinate transformation, that is, for each point on the blood vessel skeleton, assign the corresponding three-dimensional depth information value to this point, so as to determine the coordinates (x, y, Z) of this point in the three-dimensional space, and finally obtain the venous blood vessel position information in the three-dimensional space.

[0060] The method for identifying and positioning the venous position of a patient's limb based on near-infrared imaging data in the embodiments of the present invention, through the orderly processing of multiple steps, from image enhancement to improve image quality, to spatial gradient processing and eigenvalue analysis to extract the blood vessel skeleton, then to perspective parallax processing to obtain depth information, and finally to fusion to obtain the venous blood vessel position information in the three-dimensional space. Making each step cooperate with each other, it can comprehensively and accurately position the venous blood vessels, and compared with a single image analysis method, it improves the accuracy and reliability of venous blood vessel positioning.

[0061] In one embodiment, the descriptions of steps 2011 - 2014 are as follows:

[0062] Step 2011: Perform multi-modal feature extraction on the near-infrared imaging image to obtain vascular skeleton features and indwelling catheter features respectively.

[0063] Optionally, during the image enhancement process of the limb thrombus monitoring device, an improved deep learning model is used for multi-modal feature extraction. A multi-branch model based on a convolutional neural network (CNN) is constructed. The deep learning model has a powerful automatic feature learning ability. Through training with a large amount of near-infrared imaging image data, it can learn the unique feature patterns of blood vessels and indwelling catheters. The multi-branch model structure can optimize the feature extraction process for different targets respectively, improving the accuracy and pertinence of feature extraction. Therefore, when the near-infrared imaging image is input, one branch focuses on vascular skeleton feature extraction, and the other branch focuses on indwelling catheter feature extraction. For the vascular skeleton feature extraction branch, a series of convolutional layers and pooling layers are adopted. The convolutional layer performs a convolution operation on the image through a convolution kernel, such as the two-dimensional convolution formula: where O(x,y) represents the convolution output result, I(x,y) represents the input image, and K - (i,j) represents the convolution kernel, and k represents the convolution kernel size. Through multiple convolutional and pooling operations, the vascular skeleton features at different scales in the image are gradually extracted. For the indwelling catheter feature extraction branch, a similar structure is also adopted, but the convolution kernel parameters and network structure are optimized according to prior knowledge such as the shape and material of the indwelling catheter. For example, the indwelling catheter usually appears as a slender tubular structure in the near-infrared image. By adjusting the convolution kernel size and stride, it can better capture this structural feature. After training, the two branches respectively output the vascular skeleton feature map F xu and the indwelling catheter feature map F li .

[0064] Step 2012: Remove the part similar to the indwelling catheter feature in the infrared imaging image based on a preset suppression threshold to obtain a first enhanced image.

[0065] Optionally, in the near-infrared imaging image, the presence of the indwelling catheter may interfere with the observation and analysis of blood vessels. By removing the part similar to the indwelling catheter feature, this interference can be reduced and the blood vessel information can be highlighted. It is beneficial for subsequent accurate extraction of blood vessel-related features and localization of blood vessels. Therefore, after the limb thrombus monitoring device obtains the indwelling catheter feature map F li , by calculating the similarity between each pixel region in the original target enhanced image F yuan and the indwelling catheter feature, the calculation formula is: where p represents a pixel region in the image, p i represents the value of the i-th pixel in this region, and F li (i) represents the value of the indwelling catheter feature at the corresponding position. Then, for the similarity greater than the preset suppression threshold Tyu For the pixel region, set its gray value to a relatively low value (such as 0), so as to remove the part in the image that is similar to the indwelling catheter feature, and obtain the first enhanced image F one Suppression threshold T yu It can be determined by statistically analyzing a large number of near-infrared imaging images containing indwelling catheters, finding the range of characteristic values that can accurately distinguish indwelling catheters from other tissues, taking the median value as the initial threshold, and then fine-tuning through experiments.

[0066] Step 2013: Based on the vascular skeleton features, perform contrast enhancement and denoising on the infrared imaging image to obtain the second enhanced image.

[0067] Optionally, after obtaining the vascular skeleton features, the limb thrombus monitoring device uses the Contrast Limited Adaptive Histogram Equalization (CLAHE) method to perform contrast enhancement on the infrared imaging image. For example, set the gray level of the image to L dui , for a small region R in the following xiao , its gray histogram is k0 = 0, 1,..., L dui -1. Then the gray value g(x, y) of the region after CLAHE processing is calculated by the calculation formula as where k0 represents the gray value of the pixel point (x, y) in the original image. Through this formula, the gray distribution is adaptively adjusted according to the region where the vascular skeleton features are located, and the contrast is enhanced.

[0068] Furthermore, when the limb thrombus monitoring device denoises the infrared imaging image according to the vascular skeleton features, it uses a denoising algorithm based on Non-Local Means (NLM). For a pixel point p0 in the image, its denoised pixel value where Ω represents a neighborhood centered on p0 in the image, and ω(p0, q0) represents the weight between the pixel points p0 and q0, and the weight is calculated by comparing the similarity of the neighborhood blocks centered on p0 and q0. Combining these two methods, the image is processed based on the vascular skeleton features to obtain the second enhanced image F two .

[0069] Step 2014: Register and fuse the first enhanced image and the second enhanced image to obtain the target enhanced image.

[0070] Optionally, after determining the first enhanced image F one and the second enhanced image F two , the limb thrombus monitoring device uses an image registration method based on feature point matching. In the first enhanced image F one and the second enhanced image F twoAmong them, feature points are extracted by an improved version of the SIFT (Scale-Invariant Feature Transform) algorithm. The improvement lies in optimizing the description of feature points according to the characteristics of near-infrared imaging images, making them more adaptable to the characteristics of blood vessels and tissues. After finding the matching feature point pairs, the transformation matrix is calculated to register F one and F two . Let the transformation matrix be T b . For the pixel point (x1, y1) in F one , its coordinates in the registered image after transformation are . After registration, image fusion is performed. The weighted average fusion method is adopted. The pixel value F mu of the target enhanced image F mu (x, y) = α1F one (x, y) + (1 - α1)F two (x, y), where α1 represents the weight coefficient, which is determined through experiments according to the characteristics of the image and the processing effect, and generally takes values between 0.4 and 0.6. The weight is adjusted so that the fused image can not only retain the clear background after removing the indwelling catheter but also highlight the enhanced blood vessel features, and finally obtain the target enhanced image F mu .

[0071] In the embodiment of the present invention, through a series of steps such as multi-modal feature extraction, indwelling catheter interference removal, targeted enhancement of the blood vessel area, and image registration and fusion, the near-infrared imaging image is processed comprehensively and specifically. Compared with a single image enhancement method, it can more effectively highlight the features of venous blood vessels, remove interference factors, improve the image quality, provide a solid image basis for accurately identifying and locating venous blood vessels, and improve the accuracy and reliability of venous blood vessel positioning.

[0072] In one embodiment, the descriptions of steps 2041 - 2045 are as follows:

[0073] In step 2041, optical flow processing is performed on each pair of adjacent-view target enhanced images to obtain the optical flow vector of each pixel in the target enhanced image.

[0074] Optionally, when the limb thrombus monitoring device performs optical flow processing on each pair of adjacent-view target enhanced images, a dense optical flow algorithm based on the variational method, such as the Brox optical flow algorithm, is adopted. The dense optical flow algorithm based on the variational method can calculate the optical flow vector of each pixel in the image and provide more comprehensive image motion information. In near-infrared imaging images, the details of blood vessels and surrounding tissues are rich, and the dense optical flow algorithm can better capture the motion changes of these details and provide a more accurate data basis for subsequent matching and disparity calculation. This algorithm constructs an energy function based on the brightness constancy assumption and the spatial smoothness constraint. Let the adjacent-view target enhanced images be F1(x, y) and F2(x, y) respectively, and the optical flow vector denoted as the displacement of pixel (x, y) between two images, and the energy function consists of data terms and regularization terms , that is data terms F2(x + u(x, y), y + v gu (x, y))] 2 , which is used to measure the brightness change of pixels between two images and reflects the brightness constancy assumption. The regularization term is used to ensure the spatial smoothness of the optical flow field, and α g represents the weight coefficient for balancing the data term and the regularization term. Then, by minimizing the energy function the optical flow vector is solved using an iterative optimization algorithm (such as Gauss - Seidel iteration) to obtain the optical flow vector of each pixel.

[0075] Step 2042: Based on the optical flow vectors, match the target enhanced images of multiple viewpoints to obtain the optical flow trajectories of corresponding pixels under different viewpoints.

[0076] Optionally, after obtaining the optical flow vectors of each pair of adjacent viewpoint images, the limb thrombus monitoring device establishes the optical flow trajectories of corresponding pixels under different viewpoints by tracking the optical flow vectors. Taking the images F1, F2, and F3 of three viewpoints as an example, starting from a certain pixel (x1, y1) in image F1, according to the optical flow vector between F1 and F2, its corresponding pixel in F2 can be determined and then, according to the optical flow vector between (x2, y2) in F2 and F3, its corresponding pixel in F3 is determined to form the optical flow trajectory [(x1, y1), (x2, y2), (x3, y3)] from F1 to F3. For images of multiple viewpoints, by analogy, the optical flow trajectory spanning all viewpoints is established for each pixel, and finally the optical flow trajectories of corresponding pixels under different viewpoints are obtained.

[0077] Step 2043: Perform differential analysis on the optical flow trajectories under different viewpoints to determine the parallax.

[0078] Optionally, when determining the parallax, the limb thrombus monitoring device first, according to the fact that the parallax refers to the position difference of corresponding points of the same object in images of different viewpoints, sets the optical flow trajectories corresponding to the same object point under different viewpoints as (x1, y1), (x2, y2), …, (x n , y n )], and taking the first viewpoint as the reference, calculates the parallax between other viewpoints and the first viewpoint. For the i - th viewpoint and the first viewpoint, its parallax is Traverse the optical flow trajectories of all pixels to finally obtain the disparity map of the entire image. The disparity is determined by analyzing the differences in optical flow trajectories, intuitively utilizing the change information of pixel positions under different perspectives, and can accurately reflect the displacement differences of objects under different perspectives.

[0079] Step 2044: Optimize the disparity using multi-view geometric constraints to obtain the optimized disparity.

[0080] Optionally, after obtaining the disparity, the limb thrombus monitoring device adopts multi-view geometric constraints, which consider the imaging geometric relationship of the camera (or near-infrared sensor) and the geometric consistency between images of different perspectives. Assume the camera model is a pinhole camera model, and the internal parameter matrix of the camera is K juz , taking the images F1, F2, and F3 of three perspectives as examples, there is a epipolar geometric relationship. Assume the coordinate of pixel p0 in F1 is m1, in F2 is m2, and in F3 is m3, then the following epipolar constraints are satisfied: where F 12 and F 13 are the fundamental matrices between image F1 and F2, F1 and F3 respectively. Using multi-view geometric constraints, the previously calculated disparity is optimized. By minimizing an energy function, which includes a smooth term of the disparity (ensuring that the disparity changes of adjacent pixels are not large) and a term satisfying multi-view geometric constraints. Such as the energy function where represents the disparity of pixel p0, S pen represents the disparity of its adjacent pixel, and λ1 and λ2 represent weight coefficients respectively. By iteratively optimizing and solving this energy function, the optimized disparity is obtained. Optimizing the disparity using multi-view geometric constraints can utilize the geometric characteristics of camera imaging and the geometric relationship between different perspectives, eliminate the noise and unreasonable disparity changes in the initial disparity calculation, and improve the accuracy and stability of the disparity.

[0081] Step 2045: Based on the optimized disparity and combined with the parameters of the near-infrared image device, determine the depth value of each pixel point, and determine the three-dimensional depth information value according to the depth value.

[0082] Optionally, the limb thrombus monitoring device combines the optimized disparity S with the baseline distance B and focal length f of the camera jc , according to the principle of triangulation, the depth is used to calculate the depth value Z of each pixel point s , combine the depth values of all pixel points with the corresponding pixel coordinates (x, y) to form three-dimensional point cloud data, thereby determining the three-dimensional depth information value.

[0083] The embodiment of the present invention comprehensively and systematically extracts accurate three-dimensional depth information from multi-view images through a series of steps such as optical flow processing, image matching, disparity calculation, disparity optimization, and depth value calculation. Each step is interrelated, gradually improving the accuracy and reliability of the data. Compared with some simple three-dimensional reconstruction methods based on a single view or a single algorithm, this method makes full use of the information complementarity and geometric constraint relationship between multi-view images, and can more accurately restore the position and shape of veins in three-dimensional space.

[0084] In one embodiment, steps 3011 to 3015 are described as follows:

[0085] Step 3011: Position the ultrasound imaging data based on the venous blood vessel position information to obtain a first venous blood vessel region in the ultrasound imaging data.

[0086] Optionally, the limb thrombosis monitoring device uses image registration technology to accurately match the acquired venous position information (from the near-infrared imaging data processing results) with the ultrasound imaging data based on the venous position information. First, the venous position information is transformed to make it consistent with the coordinate system of the ultrasound imaging data. Assume that the transformation matrix from the near-infrared imaging data coordinate system to the ultrasound imaging data coordinate system is T zh , for the point (x nir ,y nir ), the transformed coordinates are The first venous region is determined by searching for regions with coordinates close to the converted ones in the ultrasound imaging data and combining the morphological characteristics of the blood vessels (such as the diameter range and continuity of the direction of the blood vessels). During the search, an algorithm based on region growing is used, with the converted coordinate points as seed points. According to the grayscale difference between the blood vessel region and the surrounding tissue in the ultrasound image, the region is gradually expanded until the blood vessel boundary conditions are reached. The search method based on the region growing algorithm can delineate the region according to the actual morphological characteristics of the blood vessels, which is more in line with the actual situation of the blood vessels and reduces the possibility of misjudgment.

[0087] Step 3012: Perform spectrum analysis on the ultrasonic echo signal in the first venous blood vessel region to obtain an ultrasonic echo spectrum.

[0088] Optionally, after determining the first venous blood vessel area in the ultrasound imaging data, the limb thrombosis monitoring device uses a fast Fourier transform (FFT) algorithm to perform spectrum analysis on the ultrasound echo signal in the first venous blood vessel area. Suppose the ultrasound echo signal is s(t), and after discrete sampling, s[n] is obtained, n=0, 1, ..., N-1, where N represents the number of sampling points. Perform FFT transformation on s[n], and the formula is: k s = 0, 1, …, N - 1, the obtained S[k s is the ultrasonic echo spectrum.

[0089] Step 3013, correct the ultrasonic echo spectrum based on the position depth where the first venous blood vessel region is located to obtain the corrected ultrasonic echo spectrum.

[0090] Optionally, when the ultrasonic signal propagates in human tissues, it will be attenuated due to factors such as tissue absorption and scattering, and the attenuation degree is related to the propagation depth. Therefore, the limb thrombus monitoring device corrects the ultrasonic echo spectrum according to the position depth where the first venous blood vessel region is located. Let the position depth where the first venous blood vessel region is located be d s , and the attenuation coefficient of the ultrasonic signal be α d (d), and its relationship with the depth d s can be determined by experimental measurement or theoretical model (such as the empirical formula α d (d s ) = α 00 + βd s , where α 00 represents the initial attenuation coefficient, and β is a coefficient related to tissue characteristics). For each frequency component s[k s in the ultrasonic echo spectrum S[k s , the corrected spectrum value

[0091] At the same time, considering the change in the propagation speed of the ultrasonic signal at different depths, let the propagation speed of the ultrasonic wave at the depth d s be v c (d s ). According to the Doppler effect, the frequency will shift. Let the original frequency be f yu , and the corrected frequency where v0 represents the propagation speed of the ultrasonic wave in the standard medium. By performing the above corrections on the amplitude and frequency of the spectrum, a corrected ultrasonic echo spectrum that more accurately reflects the blood flow in the blood vessel is obtained.

[0092] Step 3014, perform blood flow motion analysis on the corrected ultrasonic echo spectrum to obtain the characteristic frequencies related to blood flow motion.

[0093] Optionally, in the received corrected ultrasonic echo spectrum by the limb thrombus monitoring device, the frequency components related to blood flow motion are manifested as specific peaks or frequency bands. According to the Doppler effect, when the ultrasonic signal encounters the moving blood flow, a frequency shift will occur. Let the ultrasonic emission frequency be f yu , the blood flow velocity be v0, and the propagation speed of the ultrasonic wave in the standard medium be c0, then the Doppler frequency shift where θ represents the angle between the ultrasonic propagation direction and the blood flow direction. Then, by analyzing the corrected ultrasonic echo spectrum, the peak value of the frequency shift corresponding to the blood flow motion is found, and the frequency corresponding to this peak value is the characteristic frequency f cha 。

[0094] Step 3015: Analyze and process the characteristic frequency based on a preset blood flow velocity function to obtain the blood flow velocity data at the venous blood vessel.

[0095] Optionally, the limb thrombus monitoring device obtains the characteristic frequency f cha After that, according to the preset blood flow velocity function With the ultrasonic device parameter f obtained by measurement yu and θ, substituting them into the above formula can calculate the blood flow velocity data at the venous blood vessel.

[0096] In the embodiment of the present invention, through an orderly process of multiple steps, from accurate blood vessel area positioning, to ultrasonic echo signal spectrum analysis, spectrum correction, characteristic frequency extraction, and finally calculating the blood flow velocity. Each step is closely connected, gradually improving the accuracy and pertinence of the data. Compared with some technologies that use ultrasonic imaging or other single methods alone to obtain the blood flow velocity, this method fully combines the blood vessel position information determined by near-infrared imaging and the blood flow information of ultrasonic imaging, comprehensively considers the ultrasonic signal propagation characteristics and blood flow motion characteristics, and can measure the blood flow velocity at the venous blood vessel more accurately.

[0097] In one embodiment, the descriptions of steps 3021 - 3025 are as follows:

[0098] Step 3021: Construct an infrared thermal imaging sequence based on the infrared imaging data within a preset time interval.

[0099] Optionally, the limb thrombus monitoring device will, within a preset time interval Δt, at a fixed frame rate f gu For the infrared thermal imaging data collected from the target limb part, with the total number of frames collected being δ, the acquisition time Arrange each frame of the collected infrared thermal imaging image F to (x, y), t = 1, 2, …, δ, in chronological order to construct an infrared thermal imaging sequence {F to (x, y)}.

[0100] Furthermore, in order to ensure the image quality, during the acquisition process, the limb thrombus monitoring device preheats and calibrates the infrared thermal imaging device to ensure the accuracy of the device's temperature measurement. At the same time, preprocess the acquired images, including removing bad pixels, performing non-uniformity correction, etc. For bad pixel detection, use the median filtering method. Taking each pixel point as the center, take the median of the pixel values in its neighborhood. If the difference between the pixel value and the median exceeds a certain threshold, it is determined as a bad pixel and replaced.

[0101] Step 3022: Based on the venous blood vessel position information, locate the infrared thermal imaging sequence to obtain the second venous blood vessel region in the infrared thermal imaging data.

[0102] Optionally, the limb thrombus monitoring device uses the previously obtained venous blood vessel position information and converts its coordinates to the coordinate system of the infrared thermal imaging data. Assume the transformation matrix is T ir , for the point (x jir , y jir ) in the venous blood vessel position information, the converted coordinates are In each frame of the infrared thermal imaging sequence, taking the converted coordinates as the center, according to the morphological characteristics of the blood vessel (such as the approximate diameter range and shape of the blood vessel, etc.), use the template matching method to determine the second venous blood vessel region. Pre-build a template of the blood vessel shape, and the gray scale distribution of the template is similar to the characteristics of the blood vessel in the infrared thermal imaging. By calculating the similarity between the template and each region in the image, find the region with the highest matching degree with the template as the second venous blood vessel region. The similarity calculation uses the normalized cross-correlation algorithm. Let the template be T m (x, y), and the image region be I te (x, y), the formula for the normalized correlation coefficient C x is where and represent the means of the template and the image region respectively. Select the region with the largest C x value as the second venous blood vessel region, and perform such a positioning operation in each frame of the image to obtain the second venous blood vessel region in the entire infrared thermal imaging sequence.

[0103] Step 3023: Perform mean processing on all pixel points in the second venous blood vessel region to obtain the temperature average sequence.

[0104] Optionally, for the second venous blood vessel region of each frame of the infrared thermal imaging sequence, the limb thrombus monitoring device sets the temperature value of the pixel points in this region as T ij(to), i = 1, 2, ..., M, j = 1, 2, ..., N, t0 = 1, 2, ..., n, where M and N are the number of pixels in the region in the x and y directions, respectively. For each frame of the image, the average temperature of the second venous region is calculated by the formula Arrange the average temperature calculated for each frame of the image in chronological order to form a temperature average sequence

[0105] Step 3024, performing time series analysis on the temperature average series to obtain a time series analysis result.

[0106] Optionally, the limb thrombosis monitoring device averages the received temperature sequence The autoregressive moving average model (ARMA) is used for time series analysis. The expression of the ARMA model is: where φ i Expressed as the autoregressive coefficient, τ i φ is expressed as the moving average coefficient, ∈(to) is expressed as a white noise sequence, w and r are expressed as the autoregressive order and the moving average order, respectively. By minimizing the residual sum of squares, the maximum likelihood estimation algorithm is used to determine φ i , τ i , w and r, and by analyzing the autoregressive coefficient and moving average coefficient, as well as the residual information of the model, the trend, periodicity and other characteristics of the temperature average series are obtained to form the time series analysis results. For example, if the autoregressive coefficient φ1 is large and positive, it means that the current temperature has a strong positive correlation with the temperature at the previous moment, and the temperature has a certain continuity; if the moving average coefficient τ1 is large, it means that the recent noise has a greater impact on the current temperature.

[0107] Step 3025: Calibrate the average temperature sequence in the second venous blood vessel region at the current moment based on the time series analysis result to obtain target temperature data at the venous blood vessel.

[0108] Optionally, the limb thrombosis monitoring device analyzes the model parameters and features obtained from the time series analysis to da The average temperature series If the time series analysis results show a trend change, let the trend function be T tre (to), obtained by fitting, for example, the linear trend function T tre (to) = a + bto, where a and b are fitting coefficients. Then the calibrated temperature value is To eliminate the influence of trend factors on the current temperature. If there is a periodic change, let the period be T cuc, extract the periodic components through methods such as Fourier transform, and set the periodic function as T cuc (to), then the calibrated temperature value Meanwhile, considering factors such as measurement errors, and combining the residual information obtained from time series analysis, fine-tune the calibrated temperature value. Let the standard deviation of the residual be σ ∈ , according to the residual situation at the current moment, add or subtract a value related to σ ∈ , such as ±k e σ ∈ , k e represents a coefficient determined according to experience, generally taking values between 1 and 3, and finally obtain the target temperature data at the venous blood vessel

[0109] In the embodiment of the present invention, from constructing the infrared thermal imaging sequence, to positioning the blood vessel area based on the venous blood vessel position information, then to processing the pixel point temperature, analyzing the time series, and calibrating the temperature data, a complete and logically tight system is formed, reasonably processing the pixel point temperature and analyzing the time series, and performing temperature calibration based on the analysis results, improving the accuracy and reliability of the target temperature data.

[0110] In one embodiment, the limb thrombus monitoring method based on arteriovenous identification further includes: Step 401, determine the blood flow velocity standard value and the thrombus warning threshold based on the influence of the patient's personal data and the patient's indwelling time on blood flow; the patient's personal data includes the patient's age and the patient's BMI.

[0111] Input the blood flow velocity standard value, the thrombus warning threshold, the blood flow velocity data, and the target temperature data into the thrombus risk assessment model to obtain the risk assessment result output by the thrombus risk assessment model.

[0112] Optionally, when the limb thrombus monitoring device assesses the thrombus risk of a patient, as the patient ages, the vascular wall gradually undergoes structural and functional changes, such as reduced vascular elasticity, intimal thickening, and lumen stenosis. These changes lead to an increase in blood flow resistance and a slowdown in blood flow velocity. At the same time, being too high or too low in BMI (Body Mass Index) may affect blood flow. Obese patients (with a higher BMI) often have metabolic disorders such as hyperlipidemia and hyperglycemia. These factors cause an increase in blood viscosity. At the same time, obesity also causes fat deposition on the vascular wall, narrowing the vascular lumen, thereby increasing blood flow resistance and reducing blood flow velocity. On the contrary, patients with too low a BMI may have conditions such as malnutrition, resulting in a decrease in vascular wall elasticity or abnormal blood components, which also affects blood flow velocity. Finally, the length of time that medical devices such as indwelling catheters and stents are in the patient's body affects blood flow. On the one hand, indwelling devices may cause damage to vascular endothelial cells, activate the coagulation system, leading to platelet aggregation and thrombus formation; on the other hand, as the indwelling time prolongs, substances such as proteins and cells in the blood may be adsorbed on the surface of the device to form a biofilm, further affecting the blood flow state and slowing down the local blood flow velocity. Therefore, when the thrombus risk assessment model assesses the thrombus risk, it is necessary to make targeted adjustments to the standard value of blood flow velocity and the thrombus warning threshold for judging thrombus risk according to the patient's age, BMI status, and the patient's indwelling time, as specifically described in steps 4011 - 4015.

[0113] In one embodiment, the descriptions of steps 4011 - 4015 are as follows:

[0114] Step 4011, based on fluid mechanics and combined with the elastic characteristics of venous blood vessels, construct a blood flow model for blood flowing in venous blood vessels.

[0115] Optionally, the limb thrombus monitoring device uses the Navier - Stokes equation combined with a vascular elasticity model to construct a blood flow model. The Navier - Stokes equation describes the motion law of viscous incompressible fluids. Its expressions in the cylindrical coordinate system (considering the blood vessel approximated as a cylinder) include the continuity equation: Momentum equation (axial): Momentum equation (radial): Where u xu represents the axial velocity, v xu represents the radial velocity, ρ xu represents the blood density, p xu represents the pressure, μ xu represents the blood viscosity, x xu represents the axial coordinate, r xuDenoted as the radial coordinate, t xu Denoted as time. Considering the elastic properties of blood vessels, the Mooney-Rivlin model is adopted to describe the mechanical behavior of the vessel wall, and the stress-strain relationship of the vessel wall is where σ xu Denoted as stress, λ xu Denoted as the stretch ratio, and both C1 and C2 are denoted as material constants. The blood vessel elasticity model is coupled with the Navier-Stokes equation through boundary conditions. For example, at the vessel wall, the velocity satisfies the no-slip condition u xu = v xu = 0, and considering the influence of the deformation of the vessel wall on blood flow, a complete blood flow model of blood flowing in venous vessels is constructed.

[0116] Step 4012: Determine the elastic coefficient of the vessel wall based on the influence of the patient's age on the elasticity of the vessel wall and blood components.

[0117] Optionally, the limb thrombus monitoring device establishes a functional relationship between the patient's age and the elastic coefficient of the vessel wall through a large amount of clinical research data. The vessel wall will gradually lose its elasticity and become stiff. Generally, there is a certain functional relationship between the elastic modulus E of the vessel wall ta and the age A nian which can be obtained by fitting a large amount of clinical data. Through statistical analysis, it is found that the relationship between the elastic modulus E of the vessel wall ta and the age A nian is approximately: E ta = E cu + φ1A nian , where E cu Denoted as the initial elastic modulus (which can be determined according to the average data of healthy young people), and φ1 is denoted as the elastic change coefficient related to age (obtained through data analysis). The elastic coefficient α of the vessel wall ta is related to the elastic modulus E of the vessel wall ta . For example, there is a linear relationship between them α ta = φ2E ta , where φ2 is denoted as the proportionality coefficient, and then the relationship between the elastic coefficient α of the vessel wall ta and the age A nian can be obtained as: α ta = φ2(E cu + φ1A nian ).

[0118] Step 4013: Determine the current blood viscosity and the current vessel radius based on the influence of the patient's BMI on blood viscosity and vessel radius.

[0119] Optionally, after the limb thrombus monitoring device obtains the patient's BMI, according to the relationship between the patient's BMI and the blood viscosity μ xuand the blood vessel radius R xu There is a correlation. That is, patients with a higher BMI have relatively more fat and other components in their blood, resulting in an increase in blood viscosity. Therefore, through clinical research and data analysis, a functional relationship between blood viscosity μ xu and BMI can be established. For example, μ xu = μ cu + φ3BMI, where μ cu represents the reference value of blood viscosity under normal BMI, and φ3 represents the blood viscosity change coefficient related to BMI. At the same time, BMI also affects the blood vessel radius. Due to reasons such as fat accumulation in obese patients, the blood vessels may be compressed to a certain extent, resulting in a smaller blood vessel radius. Therefore, it is assumed that the relationship between the blood vessel radius R xu and BMI is: R xu = R cu - φ4BMI, where R cu represents the reference value of the blood vessel radius under normal BMI, and φ4 represents the blood vessel radius change coefficient related to BMI.

[0120] Step 4014, determine the current roughness based on the influence of the patient's indwelling time on the blood vessel wall roughness.

[0121] Optionally, the limb thrombus monitoring device determines the indwelling time according to the indwelling time of the patient input by the medical staff or the time from the initial use of the device to the current time. Since the situation of the patient indwelling the catheter, etc. will increase the roughness of the blood vessel wall over time. Let the initial roughness of the blood vessel wall be Q cu , as the indwelling time T cu increases, the change in roughness Q xu can be approximately expressed as: Q xu = Q cu + φ5T cu , where φ5 represents the roughness change coefficient related to the indwelling time, which can be obtained through clinical observation and data analysis.

[0122] Step 4015, input the current elastic coefficient, current blood viscosity, current blood vessel radius, and current roughness into the blood flow model for analysis and processing to determine the standard value of blood flow velocity and the thrombus warning threshold.

[0123] Optionally, the limb thrombus monitoring device is based on the current elastic coefficient α ta , current blood viscosity μ xu , current blood vessel radius R xu and current roughness Q xuSubstitute it into the blood flow model (i.e., the Navier-Stokes equation including the elastic characteristics of blood vessels). Solve the equation by numerical calculation methods (such as the finite element method) to obtain the blood flow velocity distribution in the venous vessels under different conditions. The standard value of blood flow velocity can be obtained through statistical analysis of the calculation results for healthy people (considering factors such as different ages, BMI, etc.). For example, take the average value of the blood flow velocity calculated under different conditions as the standard value of blood flow velocity v std , for the thrombus warning threshold, according to clinical research and statistical analysis, when the blood flow velocity is lower than a certain threshold, the risk of thrombus formation increases significantly. By analyzing a large number of clinical cases and combining the simulation results of the blood flow model, determine the thrombus warning threshold v thr . For example, statistically analyze the blood flow velocity data when thrombus occurs in different patients, and combine the blood flow velocity under different parameters simulated by the model to find a critical value as the thrombus warning threshold. When the actually measured blood flow velocity is lower than this threshold, it indicates that there may be a thrombus risk.

[0124] The embodiment of the present invention starts from constructing a blood flow model, comprehensively considers the influence of various factors such as patient age, BMI, indwelling time, etc. on blood flow-related parameters, and determines personalized blood flow velocity standard values and thrombus warning thresholds. Compared with traditional thrombus monitoring methods, it is more targeted at individuals, can more accurately evaluate the thrombus risk of patients, provides strong support for the early prevention and intervention of thrombus-related diseases in clinical practice, and improves the accuracy and effectiveness of limb thrombus monitoring.

[0125] Please refer to Figure 3 , Figure 3 , which is the embodiment diagram of the electronic device provided by the embodiment of the present invention. As Figure 3 shown, the embodiment of the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored on the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, the following steps are implemented:

[0126] Obtain image data of the patient's limb part; the image data includes near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data;

[0127] Based on the near-infrared imaging data, identify and locate the position of the patient's limb veins to obtain venous vessel position information;

[0128] Based on the venous vessel position information, extract features from the ultrasonic imaging data and the infrared thermal imaging data to obtain blood flow velocity data and target temperature data at the venous vessels respectively;

[0129] Input the blood flow velocity data and the target temperature data into the thrombus risk assessment model to obtain the risk assessment result output by the thrombus risk assessment model, and send the risk assessment result to the terminal device; the thrombus risk assessment model is trained based on the sample data of blood flow velocity and temperature and the risk assessment label result.

[0130] Please refer to Figure 4 , Figure 4 which is the embodiment diagram of the computer-readable storage medium provided by the embodiment of the present invention. As Figure 4 shown, this embodiment provides a computer-readable storage medium 400, on which a computer program 311 is stored. When the computer program 311 is executed by a processor, the following steps are implemented:

[0131] Obtain the image data of the patient's limb part; the image data includes near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data;

[0132] Based on the near-infrared imaging data, identify and locate the venous position of the patient's limb to obtain the venous blood vessel position information;

[0133] Based on the venous blood vessel position information, extract features from the ultrasonic imaging data and the infrared thermal imaging data to obtain the blood flow velocity data and the target temperature data at the venous blood vessel respectively;

[0134] Input the blood flow velocity data and the target temperature data into the thrombus risk assessment model to obtain the risk assessment result output by the thrombus risk assessment model, and send the risk assessment result to the terminal device; the thrombus risk assessment model is trained based on the sample data of blood flow velocity and temperature and the risk assessment label result.

[0135] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the limb thrombus monitoring method based on arteriovenous identification provided by the above-mentioned various methods. The limb thrombus monitoring method based on arteriovenous identification includes:

[0136] Obtain the image data of the patient's limb part; the image data includes near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data;

[0137] Based on the near-infrared imaging data, identify and locate the venous position of the patient's limb to obtain the venous blood vessel position information;

[0138] Based on the venous blood vessel position information, extract features from the ultrasonic imaging data and the infrared thermal imaging data to obtain the blood flow velocity data and the target temperature data at the venous blood vessel respectively;

[0139] Input the blood flow velocity data and the target temperature data into the thrombus risk assessment model, obtain the risk assessment result output by the thrombus risk assessment model, and send the risk assessment result to the terminal device; the thrombus risk assessment model is trained based on the sample data of blood flow velocity and temperature and the risk assessment label results.

[0140] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative work.

[0141] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A limb thrombus monitoring device based on arteriovenous recognition, characterized in that It includes a monitoring and management middle platform, a multi-modal data acquisition module, a venous blood vessel identification module, a feature extraction module, and a risk assessment module. The monitoring and management middle platform is respectively connected to the multi-modal data acquisition module, the venous blood vessel identification module, the feature extraction module, and the risk assessment module to manage each unit; The multi-modal data acquisition module is used to acquire image data of the patient's limb part; the image data includes near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data; The venous blood vessel identification module is used to identify and locate the position of the patient's limb veins based on the near-infrared imaging data to obtain venous blood vessel position information; The feature extraction module is used to extract features from the ultrasonic imaging data and the infrared thermal imaging data based on the venous blood vessel position information to respectively obtain blood flow velocity data and target temperature data at the venous blood vessels; The risk assessment module is used to input the blood flow velocity data and the target temperature data into a thrombus risk assessment model to obtain a risk assessment result output by the thrombus risk assessment model, and send the risk assessment result to a terminal device; The thrombus risk assessment model is trained based on sample data of blood flow velocity and temperature and risk assessment label results.

2. A limb thrombus monitoring method based on arteriovenous recognition, which is implemented based on the limb thrombus monitoring device based on arteriovenous recognition as described in claim 1, and is characterized in that, The limb thrombus monitoring method based on arteriovenous identification includes: Acquiring image data of the patient's limb part; the image data includes near-infrared imaging data, ultrasonic imaging data, and infrared thermal imaging data; Identifying and locating the position of the patient's limb veins based on the near-infrared imaging data to obtain venous blood vessel position information; Extracting features from the ultrasonic imaging data and the infrared thermal imaging data based on the venous blood vessel position information to respectively obtain blood flow velocity data and target temperature data at the venous blood vessels; Inputting the blood flow velocity data and the target temperature data into a thrombus risk assessment model to obtain a risk assessment result output by the thrombus risk assessment model, and sending the risk assessment result to a terminal device; the thrombus risk assessment model is trained based on sample data of blood flow velocity and temperature and risk assessment label results.

3. The method for monitoring limb thrombosis based on arteriovenous recognition according to claim 2, wherein The identifying and locating the position of the patient's limb veins based on the near-infrared imaging data to obtain venous blood vessel position information includes: Performing image enhancement on the near-infrared imaging data to obtain a target enhanced image; Performing spatial gradient processing on the target enhanced image to obtain a gradient vector, and constructing a three-dimensional gradient tensor field based on the gradient vector and a preset tensor voting function; Performing eigenvalue decomposition based on the three-dimensional gradient tensor field to obtain blood vessel skeleton information; Performing perspective parallax processing on multiple target enhanced images to obtain three-dimensional depth information values; the multiple target enhanced images are respectively near-infrared images from different perspectives; Fusing the blood vessel skeleton information and the three-dimensional depth information values to obtain venous blood vessel position information in three-dimensional space.

4. The method for monitoring limb thrombosis based on arteriovenous recognition according to claim 3, characterized in that The performing image enhancement on the near-infrared imaging data to obtain a target enhanced image includes: Performing multi-modal feature extraction on the near-infrared imaging image to respectively obtain blood vessel skeleton features and indwelling catheter features; Remove the part in the infrared imaging image that is similar to the indwelling catheter feature based on a preset suppression threshold to obtain a first enhanced image; Enhance the contrast and remove noise from the infrared imaging image based on the blood vessel skeleton feature to obtain a second enhanced image Register and fuse the first enhanced image and the second enhanced image to obtain a target enhanced image.

5. The method for limb thrombus monitoring based on arteriovenous recognition according to claim 3, characterized in that Performing perspective parallax processing on the multiple target enhanced images to obtain three-dimensional depth information values, including: Performing optical flow processing on the target enhanced images of each pair of adjacent perspectives to obtain an optical flow vector for each pixel in the target enhanced image; Matching the target enhanced images of multiple perspectives based on the optical flow vector to obtain an optical flow trajectory of corresponding pixels under different perspectives; Performing difference analysis on the optical flow trajectories under different perspectives to determine the parallax; Optimizing the parallax using multi-perspective geometric constraints to obtain an optimized parallax; Based on the optimized parallax and combined with the near-infrared image device parameters, determine the depth value of each pixel point, and determine the three-dimensional depth information value according to the depth value.

6. The method for limb thrombosis monitoring based on arteriovenous recognition according to claim 2, characterized in that, Obtaining blood flow velocity data at the venous blood vessels based on the venous blood vessel position information and the ultrasonic imaging data, including: Position the ultrasonic imaging data based on the venous blood vessel position information to obtain a first venous blood vessel region in the ultrasonic imaging data; Perform spectral analysis on the ultrasonic echo signals in the first venous blood vessel region to obtain an ultrasonic echo spectrum; Correct the ultrasonic echo spectrum based on the position depth where the first venous blood vessel region is located to obtain a corrected ultrasonic echo spectrum; Perform blood flow motion analysis on the corrected ultrasonic echo spectrum to obtain a characteristic frequency related to blood flow motion; Analyze and process the characteristic frequency based on a preset blood flow velocity function to obtain blood flow velocity data at the venous blood vessels.

7. The method for monitoring limb thrombosis based on arteriovenous recognition according to claim 2, wherein Obtaining target temperature data at the venous blood vessels based on the venous blood vessel position information and the infrared thermal imaging data, including: Construct an infrared thermal imaging sequence based on the infrared imaging data at preset time intervals; Position the infrared thermal imaging sequence based on the venous blood vessel position information to obtain a second venous blood vessel region in the infrared thermal imaging data; Perform mean processing on all pixel points in the second venous blood vessel region to obtain a temperature average sequence; Perform time series analysis on the temperature average sequence to obtain a time series analysis result; Calibrate the temperature average sequence in the second venous blood vessel region at the current moment based on the time series analysis result to obtain target temperature data at the venous blood vessels.

8. The method for limb thrombosis monitoring based on arteriovenous recognition according to any one of claims 2 to 7, characterized in that, The limb thrombus monitoring method based on arteriovenous identification further includes: Determine a blood flow velocity standard value and a thrombus warning threshold based on the influence of patient personal data and the patient's indwelling time on blood flow; the patient personal data includes the patient's age and the patient's BMI; Input the blood flow velocity standard value, the thrombus warning threshold, the blood flow velocity data, and the target temperature data into the thrombus risk assessment model to obtain the risk assessment result output by the thrombus risk assessment model; Correspondingly, determining the standard value of blood flow velocity and the thrombus warning threshold based on the influence of the patient's personal data and the patient's indwelling time on blood flow includes: Constructing a blood flow model of blood flowing in the venous blood vessels based on fluid mechanics and the elastic characteristics of the venous blood vessels; Determining the elastic coefficient of the blood vessel wall based on the influence of the patient's age on the elasticity of the blood vessel wall and blood components; Determining the current blood viscosity and the current blood vessel radius based on the influence of the patient's BMI on blood viscosity and blood vessel radius; Determining the current roughness based on the influence of the patient's indwelling time on the roughness of the blood vessel wall; Inputting the elastic coefficient, the current blood viscosity, the current blood vessel radius, and the current roughness into the blood flow model for analysis and processing to determine the standard value of blood flow velocity and the thrombus warning threshold.

9. An electronic device, comprising: A memory and a processor, characterized in that a computer software program is stored on the memory, and when the processor reads and executes the computer software program, the limb thrombus monitoring method based on arteriovenous identification according to any one of claims 2 to 8 is implemented.

10. A non-transitory computer-readable storage medium, characterized in that, A computer software program is stored in the storage medium, and when the computer software program is executed by the processor, the limb thrombus monitoring method based on arteriovenous identification according to any one of claims 2 to 8 is implemented.

Images