Model training method, and blood vessel anastomosis part clogging identification method and device

Abstract

The invention provides a model training method and a blood vessel anastomosis part clogging identification method and device, belongs to the technical field of deep learning, and aims to solve the problem that early clogging of an anastomosis part cannot be timely and accurately judged at present. The method comprises the following steps: training a first one-dimensional convolutional neural network model by using gradient features of a sample blood vessel anastomosis clogging image to obtain a target blood vessel anastomosis part clogging recognition model, and performing clogging recognition by using the target blood vessel anastomosis part clogging recognition model according to a target blood vessel anastomosis part image. Gradient features and gradient change features in the image can be recognized, so that early clogging of vascular anastomosis can be recognized, and the accuracy and reliability of a vascular anastomosis early clogging recognition result are improved.

Description

Technical Field

[0001] This application relates to the field of deep learning technology, and in particular to a model training method, a method and apparatus for identifying blockages at vascular anastomosis sites. Background Technology

[0002] Vascular anastomosis is a delicate surgical technique that reconnects two separated blood vessels to restore or establish blood flow. It utilizes microsurgical techniques to precisely suture the severed ends of the vessels, re-establishing the interrupted blood supply. It is a crucial step commonly used in wound repair and tissue or organ transplantation. Its primary medical goals are to ensure the survival of grafts (such as skin flaps or limbs), improve blood circulation in ischemic areas, and replace the function of diseased blood vessels. However, postoperatively, blood clotting can easily lead to blockage at the anastomosis site, affecting the surgical outcome.

[0003] In related technologies, patients mainly rely on regular check-ups and receiving anticoagulation regimens. Physicians assess the degree of coagulation at the anastomosis site and adjust the patient's anticoagulation regimen in a timely manner to avoid excessive blockage at the vascular anastomosis site.

[0004] However, in the aforementioned methods, for large-scale comprehensive anastomosis surgeries, such as those involving vascular, tendon, and nerve anastomosis, swelling often occurs at the tendon anastomosis site after completion. This swelling can affect the assessment of vascular anastomosis blockage, making it difficult for surgeons to accurately and promptly identify early blockage at the anastomosis site. Currently, there is still a lack of an effective method for identifying early blockage in vascular anastomosis. Summary of the Invention

[0005] In view of the above problems, embodiments of this application provide a model training method, a method for identifying vascular anastomosis blockage, an apparatus, an electronic device, and a readable storage medium to overcome or at least partially solve the above problems.

[0006] In a first aspect, embodiments of this application provide a model training method, the method comprising: Acquire images of vascular anastomosis and blockage in a sample patient; wherein the image capture point is located outside the plane formed by the direction of blood flow and the direction of blockage and swelling; Determine the gradient sequence of the sample skin surface in the sample vascular anastomosis and occlusion image; The gradient sequence of the sample skin surface is input into the first one-dimensional convolutional neural network model to be trained to obtain the sample swelling prediction probability output by the first one-dimensional convolutional neural network model. Based on the predicted probability of swelling in the sample and the swelling label value of the vascular anastomosis blockage image of the sample, the model loss value of the first one-dimensional convolutional neural network model is determined. Based on the model loss value, the model parameters and hyperparameters of the first one-dimensional convolutional neural network model are adjusted to obtain a target vascular anastomosis site stasis identification model; the target vascular anastomosis site stasis identification model is used to determine the stasis identification result of the target patient based on the target vascular anastomosis site image of the target patient.

[0007] Optionally, determining the sample skin surface gradient sequence of the sample vascular anastomosis occlusion image includes: Along the anastomotic vessel corresponding to the sample blood vessel anastomosis blockage image, determine a first pixel sequence image with dimension 1×N in the sample blood vessel anastomosis blockage image; According to the HSI color model, the first pixel sequence image is decomposed to obtain the first brightness sequence image; The brightness difference between adjacent pixels in the first brightness sequence image is calculated item by item to obtain the gradient sequence of the sample skin surface.

[0008] Optionally, determining the sample skin surface gradient sequence of the sample vascular anastomosis occlusion image includes: Determine the skin surface contour of the sample in the vascular anastomosis and blockage image; The surface contour of the sampled skin was sampled at equal intervals to obtain multiple sampling points; The positional deviation between adjacent sampling points is calculated one by one to obtain the gradient sequence of the sample skin surface.

[0009] Optionally, the model loss value includes a binary cross-entropy loss value and a multi-class cross-entropy loss value. Determining the model loss value of the first one-dimensional convolutional neural network model based on the sample swelling prediction probability and the swelling labeling value of the sample vascular anastomosis blockage image includes: Based on the predicted probability of swelling in the sample and the swelling label value of the vascular anastomosis blockage image of the sample, the binary cross-entropy loss value and the multi-class cross-entropy loss value of the first one-dimensional convolutional neural network model are determined respectively.

[0010] Optionally, the step of adjusting the model parameters and hyperparameters of the first one-dimensional convolutional neural network model based on the model loss value to obtain the target vascular anastomosis site occlusion identification model includes: Based on the binary cross-entropy loss value and the multi-class cross-entropy loss value, the model parameters of the first one-dimensional convolutional neural network model are adjusted once to obtain the first one-dimensional convolutional neural network model with updated parameters. Re-execute the process of acquiring the sample vascular anastomosis and blockage images of the sample patient to begin the next model training; When the number of consecutive occurrences of the binary cross-entropy loss value being less than the first preset threshold is greater than or equal to the first occurrence, and the number of consecutive occurrences of the multi-class cross-entropy loss value being less than the second preset threshold is greater than or equal to the first occurrence, the adjustment of the model parameters and hyperparameters of the first one-dimensional convolutional neural network model is stopped, and the target vascular anastomosis site occlusion identification model is obtained.

[0011] Secondly, embodiments of this application provide a method for identifying vascular anastomosis site blockage, the method comprising: Acquire images of the target vascular anastomosis site in the target patient; wherein the image of the target vascular anastomosis blockage is taken from a point outside the plane formed by the direction of vascular flow and the direction of blockage and swelling; Determine the first skin surface gradient sequence of the image of the target vascular anastomosis site; The first skin surface gradient sequence is input into the target vascular anastomosis site stasis identification model to obtain the first swelling prediction probability output by the target vascular anastomosis site stasis identification model; wherein, the target vascular anastomosis site stasis identification model is obtained based on the model training method described in any of the above. Based on the first swelling prediction probability, the blockage identification result of the target patient is determined.

[0012] Optionally, determining the first skin surface gradient sequence of the target vascular anastomosis site image includes: Along the anastomotic vessel corresponding to the target blood vessel anastomosis site image, a first pixel sequence image with dimension 1×N is determined in the target blood vessel anastomosis site image; According to the HSI color model, the first pixel sequence image is decomposed to obtain the first brightness sequence image; The brightness difference between adjacent pixels in the first brightness sequence image is calculated item by item to obtain the first skin surface gradient sequence.

[0013] Optionally, determining the first skin surface gradient sequence of the target vascular anastomosis site image includes: Determine the first skin surface contour of the image of the target vascular anastomosis site; The contour of the first skin surface is sampled at equal intervals to obtain multiple sampling points; The positional deviation between adjacent sampling points is calculated one by one to obtain the first skin surface gradient sequence.

[0014] Thirdly, embodiments of this application provide a model training apparatus, the apparatus comprising: The first acquisition module is used to acquire an image of the sample blood vessel anastomosis blockage in the sample patient; wherein the image capture point of the sample blood vessel anastomosis blockage is located outside the plane formed by the direction of blood vessel flow and the direction of blockage and swelling; The first determining module is used to determine the gradient sequence of the sample skin surface in the sample vascular anastomosis blockage image; The first input-output module is used to input the gradient sequence of the sample skin surface into the first one-dimensional convolutional neural network model to be trained, and to obtain the sample swelling prediction probability output by the first one-dimensional convolutional neural network model. The second determining module is used to determine the model loss value of the first one-dimensional convolutional neural network model based on the sample swelling prediction probability and the swelling label value of the sample vascular anastomosis blockage image. An adjustment module is used to adjust the model parameters and hyperparameters of the first one-dimensional convolutional neural network model based on the model loss value to obtain a target vascular anastomosis site stasis identification model; the target vascular anastomosis site stasis identification model is used to determine the stasis identification result of the target patient based on the target vascular anastomosis site image of the target patient.

[0015] Optionally, the first determining module includes: The first determining submodule is used to determine a first pixel sequence image with dimension 1×N in the sample blood vessel anastomosis blockage image along the anastomotic blood vessel corresponding to the sample blood vessel anastomosis blockage image; The first decomposition submodule is used to decompose the first pixel sequence image according to the HSI color model to obtain the first brightness sequence image. The first calculation submodule is used to calculate the brightness difference between adjacent pixels in the first brightness sequence image item by item to obtain the gradient sequence of the sample skin surface.

[0016] Optionally, the first determining module includes: The second determining submodule is used to determine the skin surface contour of the sample in the sample vascular anastomosis blockage image; The first sampling submodule is used to perform equidistant sampling of the surface contour of the sample skin to obtain multiple sampling points; The second calculation submodule is used to calculate the positional deviation between adjacent sampling points one by one to obtain the gradient sequence of the sample skin surface.

[0017] Optionally, the model loss value includes a binary cross-entropy loss value and a multi-class cross-entropy loss value, and the second determining module includes: The third determining submodule is used to determine the binary cross-entropy loss value and the multi-class cross-entropy loss value of the first one-dimensional convolutional neural network model based on the sample swelling prediction probability and the swelling label value of the sample vascular anastomosis blockage image.

[0018] Optionally, the adjustment module includes: The adjustment submodule is used to adjust the model parameters of the first one-dimensional convolutional neural network model once based on the binary cross-entropy loss value and the multi-class cross-entropy loss value, so as to obtain the first one-dimensional convolutional neural network model with updated parameters. The execution submodule is used to re-execute the acquisition of the sample vascular anastomosis blockage images of the sample patient in order to start the next model training; The stop adjustment submodule is used to stop adjusting the model parameters and hyperparameters of the first one-dimensional convolutional neural network model when the number of consecutive occurrences of the binary cross-entropy loss value being less than the first preset threshold is greater than or equal to the first occurrence, and the number of consecutive occurrences of the multi-class cross-entropy loss value being less than the second preset threshold is greater than or equal to the first occurrence, so as to obtain the target vascular anastomosis site occlusion identification model.

[0019] Fourthly, embodiments of this application provide a device for identifying vascular anastomosis site blockage, the device comprising: The second acquisition module is used to acquire images of the target vascular anastomosis site of the target patient; wherein, the image capture point of the target vascular anastomosis blockage image is located outside the plane formed by the direction of vascular flow and the direction of blockage and swelling; The third determining module is used to determine the first skin surface gradient sequence of the image of the target blood vessel anastomosis site; The second input / output module is used to input the first skin surface gradient sequence into the target vascular anastomosis site blockage recognition model to obtain the first swelling prediction probability output by the target vascular anastomosis site blockage recognition model; wherein, the target vascular anastomosis site blockage recognition model is obtained based on the model training method described above. The fourth determining module is used to determine the blockage identification result of the target patient based on the first swelling prediction probability.

[0020] Optionally, the third determining module includes: The fourth determining submodule is used to determine a first pixel sequence image with dimension 1×N in the target blood vessel anastomosis site image along the anastomotic vessel corresponding to the target blood vessel anastomosis site image; The second decomposition submodule is used to decompose the first pixel sequence image according to the HSI color model to obtain the first brightness sequence image. The third calculation submodule is used to calculate the brightness difference between adjacent pixels in the first brightness sequence image item by item to obtain the first skin surface gradient sequence.

[0021] Optionally, the third determining module includes: The fourth determining submodule is used to determine the first skin surface contour of the target blood vessel anastomosis site image; The second sampling submodule is used to perform equidistant sampling on the contour of the first skin surface to obtain multiple sampling points; The fourth calculation submodule is used to calculate the positional deviation between adjacent sampling points one by one to obtain the first skin surface gradient sequence.

[0022] Fifthly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the model training method as described above, or the method for identifying vascular anastomosis blockage.

[0023] Sixthly, embodiments of this application provide a readable storage medium storing a program or instructions, which, when executed by a processor, implement the model training method described above, or the method for identifying vascular anastomosis blockage.

[0024] The specific beneficial effects are as follows: This application embodiment acquires a sample image of vascular anastomosis and blockage from a patient. The image capture point is located outside the plane formed by the blood vessel flow direction and the direction of blockage and swelling. The gradient sequence of the skin surface of the sample image is determined, and this gradient sequence is input into a first-dimensional convolutional neural network model to be trained. The predicted swelling probability output by the first-dimensional convolutional neural network model is obtained. Based on the predicted swelling probability and the swelling label value of the sample vascular anastomosis and blockage image, the model loss value of the first-dimensional convolutional neural network model is determined. Based on the model loss value, the model parameters of the first-dimensional convolutional neural network model are adjusted. By analyzing the number of parameters and hyperparameters, a target vascular anastomosis site clogging identification model is obtained. This model is used to determine the clogging identification result of a target patient based on the target vascular anastomosis site image. The gradient features of the sample vascular anastomosis clogging images can be used to train the first one-dimensional convolutional neural network model to obtain the target vascular anastomosis site clogging identification model. When using the target vascular anastomosis site clogging identification model to identify clogging based on the target vascular anastomosis site image, the gradient features and gradient change features in the image can be identified, thereby identifying early vascular anastomosis clogging and improving the accuracy and reliability of early vascular anastomosis clogging identification results. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A schematic flowchart illustrating a model training method provided in an embodiment of this application; Figure 2 A flowchart illustrating another model training method provided in an embodiment of this application; Figure 3 A flowchart illustrating a method for identifying vascular anastomosis blockage as provided in this application embodiment; Figure 4 A comparison image of grayscale and contour recognition of suspected vascular occlusion provided in the embodiments of this application; Figure 5 A logic block diagram of a model training device provided in an embodiment of this application; Figure 6 A logic block diagram of a vascular anastomosis site blockage identification device provided in this application embodiment; Figure 7 This is a schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0027] Exemplary embodiments of this application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of this application are shown in the drawings, it should be understood that this application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.

[0028] Reference Figure 1 , Figure 1 This is a flowchart illustrating a model training method provided in an embodiment of this application. The method may include: Step 101: Obtain a sample vascular anastomosis blockage image from the sample patient; wherein the image capture point is located outside the plane formed by the direction of vascular flow and the direction of blockage and swelling.

[0029] In the embodiments of this application, desensitized sample images of vascular anastomosis blockage can be collected from a database of hospitals where vascular anastomosis surgery was performed. These images are then classified using patient identifiers and labeled with blockage diagnosis results. This yields sample images of vascular anastomosis blockage from sample patients that can be used for model training. The sample images must meet preset conditions: the image capture point must be outside the plane formed by the direction of blood flow and the direction of blockage and swelling. This ensures that the captured images have richer local variation features and are easier to identify.

[0030] Step 102: Determine the gradient sequence of the sample skin surface of the sample vascular anastomosis blockage image.

[0031] In embodiments of this application, for a sample image of vascular anastomosis blockage containing vascular anastomoses, a gradient sequence of the sample skin surface in the image can be determined. The gradient can be either a pixel value gradient or a contour gradient. If it is a pixel value gradient, the pixel value gradient is calculated based on the pixel value difference between image pixels and their neighboring pixels. If it is a contour gradient, equidistant sampling can be performed on the contour, and the positional deviation between adjacent sampling points can be calculated to obtain the contour gradient.

[0032] Step 103: Input the gradient sequence of the sample skin surface into the first one-dimensional convolutional neural network model to be trained to obtain the sample swelling prediction probability output by the first one-dimensional convolutional neural network model.

[0033] In the embodiments of this application, a one-dimensional convolutional neural network model is used as the base model to be trained, and the activation function of the output layer of the one-dimensional convolutional neural network model is modified to a softmax function with adjustable hyperparameters, the expression of which is shown in Equation 1 below: (Equation 1) In equation 1 above, is a hyperparameter, and its value range is (0, 1). and All of these are input data for the output layer.

[0034] The gradient sequence of the sample skin surface can be input into the first one-dimensional convolutional neural network model to be trained, so as to obtain the sample swelling prediction probability output by the first one-dimensional convolutional neural network model.

[0035] Step 104: Based on the sample swelling prediction probability and the swelling label value of the sample vascular anastomosis blockage image, determine the model loss value of the first one-dimensional convolutional neural network model.

[0036] In the embodiments of this application, the model loss value of the first one-dimensional convolutional neural network model can be calculated based on the sample swelling prediction probability and the swelling label value of the sample vascular anastomosis blockage image. If the swelling label value only contains two values, 0 and 1, the model loss value can be a binary cross-entropy loss value; if the swelling label value contains multiple values ​​between 0 and 1, the model loss value can be a multi-class cross-entropy loss value; if the swelling label value not only uses "0" and "1" to represent whether there is blockage, but also uses values ​​between 0 and 1 to represent the degree of blockage, the model loss value can simultaneously include both binary cross-entropy loss value and multi-class cross-entropy loss value.

[0037] Step 105: Based on the model loss value, adjust the model parameters and hyperparameters of the first one-dimensional convolutional neural network model to obtain a target vascular anastomosis site stasis identification model; the target vascular anastomosis site stasis identification model is used to determine the stasis identification result of the target patient based on the target vascular anastomosis site image of the target patient.

[0038] In the embodiments of this application, the model parameters and hyperparameters of the first one-dimensional convolutional neural network model can be adjusted according to the model loss value, thereby training the first one-dimensional convolutional neural network model to obtain a target vascular anastomosis site blockage recognition model. When adjusting the model parameters and hyperparameters, adjustments can be made in the direction that reduces the model loss value. After each parameter adjustment, step 101 can be re-executed to start the next training. If the model loss value meets the preset convergence condition, or the model training times reach the preset number, the adjustment of model parameters and hyperparameters is stopped, thereby obtaining the trained target vascular anastomosis site blockage recognition model.

[0039] In the embodiments of this application, a sample vascular anastomosis and occlusion image of a patient is acquired. The image capture point is located outside the plane formed by the blood vessel flow direction and the direction of occlusion and swelling. The gradient sequence of the skin surface of the sample vascular anastomosis and occlusion image is determined. This gradient sequence is input into a first one-dimensional convolutional neural network model to be trained, resulting in a predicted swelling probability output by the first one-dimensional convolutional neural network model. Based on the predicted swelling probability and the swelling label value of the sample vascular anastomosis and occlusion image, the model loss value of the first one-dimensional convolutional neural network model is determined. Based on the model loss value, the model of the first one-dimensional convolutional neural network model is adjusted. The model parameters and hyperparameters are used to obtain the target vascular anastomosis site congestion identification model. This model is used to determine the congestion identification result of a target patient based on the target vascular anastomosis site image. The gradient features of the sample vascular anastomosis congestion images can be used to train the first one-dimensional convolutional neural network model to obtain the target vascular anastomosis site congestion identification model. When using this model to identify congestion based on the target vascular anastomosis site image, it can identify gradient features and gradient change features in the image, thereby identifying early vascular anastomosis congestion and improving the accuracy and reliability of early vascular anastomosis congestion identification results.

[0040] Reference Figure 2 , Figure 2 This is a flowchart illustrating another model training method provided in an embodiment of this application. The method may include: Step 201: Obtain a sample vascular anastomosis blockage image from the sample patient; wherein the image capture point is located outside the plane formed by the direction of vascular flow and the direction of blockage and swelling.

[0041] Step 202: Determine the gradient sequence of the sample skin surface of the sample vascular anastomosis blockage image.

[0042] Optionally, step 202 may include the following sub-steps: Sub-step 2021: Along the anastomotic vessel corresponding to the sample blood vessel anastomosis blockage image, determine the first pixel sequence image with dimension 1×N in the sample blood vessel anastomosis blockage image.

[0043] Sub-step 2022: Decompose the first pixel sequence image according to the HSI color model to obtain the first brightness sequence image.

[0044] Sub-step 2023: Calculate the brightness difference between adjacent pixels in the first brightness sequence image item by item to obtain the gradient sequence of the sample skin surface.

[0045] In the embodiments of this application, to simplify the input data, representative pixels can be selected from the sample vascular anastomosis blockage image to calculate the sample skin surface gradient sequence. First, pixels can be selected along the corresponding anastomotic vessels in the sample vascular anastomosis blockage image to determine a first pixel sequence image with dimension 1×N. Here, N can be greater than or equal to the number of pixels on the long side of the sample vascular anastomosis blockage image to ensure that the first pixel sequence image necessarily spans the vascular anastomosis. When the anastomosis is swollen, it affects the brightness of the surrounding skin tissue under illumination. Therefore, according to the HSI color model, the first pixel sequence image can be decomposed into three types of sequence images: hue, contrast, and luminance. In this embodiment, only the first luminance sequence image is selected. Then, the luminance difference between adjacent pixels in the first luminance sequence image is calculated to obtain the sample skin surface gradient sequence. Here, the gradient is the luminance gradient. Through sub-steps 2021 to 2023, the sample skin surface gradient sequence can be calculated from the luminance of a portion of the pixels in the image, which to some extent improves the accuracy and reliability of the sample skin surface gradient sequence as input data.

[0046] Sub-step 2024: Determine the skin surface contour of the sample vascular anastomosis blockage image.

[0047] Sub-step 2025: Perform equidistant sampling on the surface contour of the sample skin to obtain multiple sampling points.

[0048] Sub-step 2026: Calculate the positional deviation between adjacent sampling points one by one to obtain the gradient sequence of the sample skin surface.

[0049] In the embodiments of this application, the gradient sequence of the sample skin surface can also be calculated using contours. First, an edge detection algorithm is used to process the sample vascular anastomosis and blockage image to obtain the sample skin surface contour. Then, equidistant sampling is performed on the sample skin surface contour to obtain multiple sampling points. Next, the positional deviation between adjacent sampling points is calculated item by item. The ratio of the difference in the ordinate to the difference in the abscissa of each positional deviation is used as the gradient value, thus obtaining the sample skin surface gradient sequence. Through sub-steps 2024 to 2026, the sample skin surface gradient sequence can be calculated using contours, which reduces the amount of sample skin surface gradient sequence data while improving the accuracy of the sample skin surface gradient sequence.

[0050] Step 203: Input the gradient sequence of the sample skin surface into the first one-dimensional convolutional neural network model to be trained to obtain the sample swelling prediction probability output by the first one-dimensional convolutional neural network model.

[0051] Step 204: Based on the predicted probability of swelling in the sample and the swelling label value of the vascular anastomosis blockage image of the sample, determine the binary cross-entropy loss value and the multi-class cross-entropy loss value of the first one-dimensional convolutional neural network model, respectively.

[0052] In the embodiments of this application, the format of the swelling annotation value of the sample vascular anastomosis blockage image can be set as a regular expression, with the format [x, y], where the symbol "[ ]" is the identifier of the annotation value, the symbol "," is the separator, the "x" before the separator uses the values ​​"0" and "1" to indicate whether the vascular anastomosis is blocked, where "0" indicates no blockage, and the "y" after the separator indicates the degree of blockage, which is a percentage less than 1. The setting rule is as follows: if x is 0, then y must be 0; if x is not 0, then y must not be 0 and y must be less than 1. If the values ​​of x and y do not conform to the above rules, the swelling annotation value is determined to be illegal, and an error is reported. At this time, the model loss value can include the binary cross-entropy loss value and the multi-class cross-entropy loss value. The x in the swelling annotation value is used to calculate the binary cross-entropy loss value together with the sample swelling prediction probability, and the y in the swelling annotation value is used to calculate the multi-class cross-entropy loss value together with the sample swelling prediction probability. Step 204 can improve the accuracy and reliability of the model loss value to some extent.

[0053] Step 205: Based on the binary cross-entropy loss value and the multi-class cross-entropy loss value, adjust the model parameters of the first one-dimensional convolutional neural network model once to obtain the first one-dimensional convolutional neural network model with updated parameters.

[0054] Step 206: Re-execute the process of obtaining the sample vascular anastomosis blockage image of the sample patient to begin the next model training.

[0055] Step 207: If the number of consecutive occurrences of the binary cross-entropy loss value being less than the first preset threshold is greater than or equal to the first occurrence, and the number of consecutive occurrences of the multi-class cross-entropy loss value being less than the second preset threshold is greater than or equal to the first occurrence, stop adjusting the model parameters and hyperparameters of the first one-dimensional convolutional neural network model to obtain the target vascular anastomosis site blockage identification model.

[0056] In the embodiments of this application, the model parameters of the first one-dimensional convolutional neural network model can be adjusted once based on the calculated binary cross-entropy loss value and multi-class cross-entropy loss value, thereby obtaining the first one-dimensional convolutional neural network with updated parameters. Then, step 201 can be re-executed to begin the next model training. The model convergence condition is set as follows: the number of consecutive occurrences of the binary cross-entropy loss value being less than the first preset threshold is greater than or equal to the first occurrence, and the number of consecutive occurrences of the multi-class cross-entropy loss value being less than the second preset threshold is greater than or equal to the first occurrence. The second preset threshold and the first preset threshold can be the same or different. If the above model convergence condition is met, the adjustment of the model parameters and hyperparameters of the first one-dimensional convolutional neural network model can be stopped, thereby obtaining the target vascular anastomosis site blockage recognition model. Through steps 205 to 207, the model can be jointly trained using dual loss values, which can improve the recognition accuracy and reliability of the target vascular anastomosis site blockage recognition model to a certain extent.

[0057] Reference Figure 3 , Figure 3 This application provides a flowchart illustrating a method for identifying vascular anastomosis blockage, which may include: Step 301: Obtain an image of the target vascular anastomosis site of the target patient; wherein the image of the target vascular anastomosis blockage is captured at a point outside the plane formed by the direction of vascular flow and the direction of blockage and swelling.

[0058] In the embodiments of this application, a shooting point can be set outside the plane formed by the direction of blood vessel flow and the direction of stagnation and swelling, and an image of the target blood vessel anastomosis site of the target patient can be obtained by taking a picture with an optical camera.

[0059] Step 302: Determine the first skin surface gradient sequence of the target vascular anastomosis site image.

[0060] In the embodiments of this application, the first skin surface gradient sequence is similar to the sample skin surface gradient sequence. The implementation details of this step can be found in the embodiment of step 102, and will not be repeated here.

[0061] Optionally, step 302 may include the following sub-steps: Sub-step 3021: Along the anastomotic vessel corresponding to the target blood vessel anastomosis site image, determine a first pixel sequence image with dimension 1×N in the target blood vessel anastomosis site image.

[0062] Sub-step 3022: Decompose the first pixel sequence image according to the HSI color model to obtain the first brightness sequence image.

[0063] Sub-step 3023: Calculate the brightness difference between adjacent pixels in the first brightness sequence image item by item to obtain the first skin surface gradient sequence.

[0064] Sub-step 3024: Determine the first skin surface contour of the target blood vessel anastomosis site image.

[0065] Sub-step 3025: Perform equidistant sampling on the contour of the first skin surface to obtain multiple sampling points.

[0066] Sub-step 3026: Calculate the positional deviation between adjacent sampling points one by one to obtain the first skin surface gradient sequence.

[0067] In the embodiments of this application, the first skin surface gradient sequence can be calculated in two ways: the brightness gradient is calculated through sub-steps 3021 to 3023, and the contour gradient is calculated through sub-steps 3024 to 3026. For specific calculation methods, please refer to the embodiments of step 202 and its sub-steps. The choice of calculation method depends on the selection of the shooting point. If the shooting point is close to the plane formed by the direction of blood vessel flow and the direction of congestion and swelling, sub-steps 3021 to 3023 are selected for calculation; if the distance is far, sub-steps 3024 to 3026 are selected. The threshold for determining the distance can be set according to the actual situation, typically 0.2 meters.

[0068] Step 303: Input the first skin surface gradient sequence into the target vascular anastomosis site blockage recognition model to obtain the first swelling prediction probability output by the target vascular anastomosis site blockage recognition model; wherein, the target vascular anastomosis site blockage recognition model is obtained based on the model training method described in any of the above.

[0069] Step 304: Based on the first swelling prediction probability, determine the blockage identification result of the target patient.

[0070] In the embodiments of this application, the clogging identification result can be determined based on the value range of the first swelling prediction probability. If the clogging identification grading range is (0, 20%), (20%, 40%), (40%, 60%), (60%, 80%), and (80%, 100%), corresponding to slight swelling, minor swelling, moderate swelling, significant swelling, and extreme swelling, respectively, then when the first swelling prediction probability is within the range (0, 20%), the clogging identification result is slight swelling; when the first swelling prediction probability is within the range (40%, 60%), the clogging identification result is moderate swelling. Through steps 301 to 304, the clogging identification model of the target vascular anastomosis site can be used for clogging identification, which can improve the accuracy and reliability of the early clogging identification result of vascular anastomosis to a certain extent.

[0071] refer to Figure 4, Figure 4 This image shows a comparison of a grayscale image and a contour recognition image of a suspected vascular blockage, provided in an embodiment of this application. In the image, the left image is the grayscale image obtained by converting the original suspected vascular blockage image to grayscale, and the right image is the contour recognition image obtained by performing contour recognition on the grayscale image (the contour is represented by red lines). The project code is as follows: import cv2 import os import tkinter as tk from tkinter import filedialog, messagebox import numpy as np def select_image_file(): """ Allow users to select image files - Windows system optimization """ root = tk.Tk() root.withdraw() root.attributes('-topmost', True) # Keep Windows system windows on top file_types = [ ("image files", "*.jpg;*.jpeg;*.png;*.bmp;*.tiff;*.tif"), ("JPEG file", "*.jpg;*.jpeg"), ("PNG file", "*.png"), ("Bitmap file", "*.bmp"), ("All files", "*.*") ] file_path = filedialog.askopenfilename( title="Select Grayscale Image File", filetypes=file_types, initialdir=os.path.expanduser("~\\Pictures") # Windows default picture directory ) return file_path def windows_path_processing(file_path): """ Windows system path handling """ if not file_path: return None # Ensure you use Windows path separators file_path = file_path.replace(' / ', '\\') # Verify file existence if not os.path.exists(file_path): raise FileNotFoundError(f"File does not exist: {file_path}") return file_path def detect_contours_windows(image_path): """ Windows system optimized contour detection function """ try: # Read Image original_image = cv2.imread(image_path) if original_image is None: Raise a ValueError("Unable to read image file. Please check the file path and format"). # Convert to grayscale image if len(original_image.shape) == 3: gray_image = cv2.cvtColor(original_image, cv2.COLOR_BGR2GRAY) else: gray_image = original_image # Gaussian blur noise reduction blurred = cv2.GaussianBlur(gray_image, (5, 5), 0) # Canny Edge Detection edges = cv2.Canny(blurred, 50, 150) # Find Outlines contours,_=cv2.findContours(edges,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE) # Create outline image if len(original_image.shape) == 2: contour_image = cv2.cvtColor(original_image, cv2.COLOR_GRAY2BGR) else: contour_image = original_image.copy() # Draw the outline (red, line width 2 pixels) cv2.drawContours(contour_image, contours, -1, (0, 0, 255), 2) return contour_image, contours, len(contours) except Exception as e: raise Exception(f"Image processing error: {str(e)}") def save_contour_image_windows(original_path, contour_image): """ Windows system path saving function """ try: # Windows path handling original_path = windows_path_processing(original_path) # Path parsing component directory = os.path.dirname(original_path) filename = os.path.basename(original_path) name, ext = os.path.splitext(filename) # Generate output filename (Windows compatible) new_filename = f"{name}_contours{ext}" output_path = os.path.join(directory, new_filename) # Avoid filename conflicts counter = 1 while os.path.exists(output_path): new_filename = f"{name}_contours_{counter}{ext}" output_path = os.path.join(directory, new_filename) counter += 1 # Save image success = cv2.imwrite(output_path, contour_image) If not successful: raise ValueError("Image saving failed") return output_path except Exception as e: raise Exception(f"Saving error: {str(e)}") def display_windows_notification(title, message): """ Windows system notification """ messagebox.showinfo(title, message) def main_windows(): """ Windows system main function """ print("=== Windows Image Contour Detector===") print("Please select the image file to process...") try: # Select file image_path = select_image_file() if not image_path: print("No file selected, program exits.") return print(f"File selected: {image_path}") # Windows path verification image_path = windows_path_processing(image_path) # Detect contours print("Detecting image contours...") contour_image, contours, contour_count = detect_contours_windows(image_path) # Save results print("Saving contour image...") output_path = save_contour_image_windows(image_path, contour_image) # Display results print( ===Processing complete===") print(f"Input image: {image_path}") print(f"Output image: {output_path}") print(f"Number of detected contours: {contour_count}") # Windows System Notification display_windows_notification( "Contour detection complete" f"Successfully detected {contour_count} contours\n" Output file: {output_path} ) # Ask whether to display image response = input(" Should the processing result be displayed? (y / n): ").lower() if response in ['y', 'yes']: # Adjust the display window size cv2.namedWindow('Original Image', cv2.WINDOW_NORMAL) cv2.namedWindow('Contour Detection Result', cv2.WINDOW_NORMAL) original = cv2.imread(image_path) cv2.imshow('original image', original) cv2.imshow('Contour detection result', contour_image) print("Press any key to close the image window...") cv2.waitKey(0) cv2.destroyAllWindows() print(" Program execution complete! except Exception as e: error_msg = f"Error: {str(e)}" print(error_msg) display_windows_notification("error", error_msg) if __name__ == "__main__": main_windows() For blockage identification, test cases [10, 11, 11, 11, 11, 11, 10, 10, 10, 10] and [10, 11, 11, 15, 25, 14, 11, 11, 11, 10] were used for testing (the actual gradient sequence contains far more gradients than the test cases). The identification result of the first test case is as follows: "======Test Cases=====" No obvious signs of blood vessel blockage were detected. The identification results for the second test case are as follows: "======Test Cases=====" An abnormal change has been detected! Maximum gradient: 10.00, location: 5.

[0072] refer to Figure 5 , Figure 5 This application provides a logic block diagram of a model training device 500, which may include: The first acquisition module 501 is used to acquire a sample vascular anastomosis blockage image of a sample patient; wherein the acquisition point of the sample vascular anastomosis blockage image is located outside the plane formed by the direction of vascular flow and the direction of blockage and swelling; The first determining module 502 is used to determine the gradient sequence of the sample skin surface of the sample vascular anastomosis blockage image; The first input / output module 503 is used to input the gradient sequence of the sample skin surface into the first one-dimensional convolutional neural network model to be trained, and to obtain the sample swelling prediction probability output by the first one-dimensional convolutional neural network model. The second determining module 504 is used to determine the model loss value of the first one-dimensional convolutional neural network model based on the sample swelling prediction probability and the swelling label value of the sample vascular anastomosis blockage image. The adjustment module 505 is used to adjust the model parameters and hyperparameters of the first one-dimensional convolutional neural network model based on the model loss value to obtain a target vascular anastomosis site stasis identification model; the target vascular anastomosis site stasis identification model is used to determine the stasis identification result of the target patient based on the target vascular anastomosis site image of the target patient.

[0073] Optionally, the first determining module 502 includes: The first determining submodule is used to determine a first pixel sequence image with dimension 1×N in the sample blood vessel anastomosis blockage image along the anastomotic blood vessel corresponding to the sample blood vessel anastomosis blockage image; The first decomposition submodule is used to decompose the first pixel sequence image according to the HSI color model to obtain the first brightness sequence image. The first calculation submodule is used to calculate the brightness difference between adjacent pixels in the first brightness sequence image item by item to obtain the gradient sequence of the sample skin surface.

[0074] Optionally, the first determining module 502 includes: The second determining submodule is used to determine the skin surface contour of the sample in the sample vascular anastomosis blockage image; The first sampling submodule is used to perform equidistant sampling of the surface contour of the sample skin to obtain multiple sampling points; The second calculation submodule is used to calculate the positional deviation between adjacent sampling points one by one to obtain the gradient sequence of the sample skin surface.

[0075] Optionally, the model loss value includes a binary cross-entropy loss value and a multi-class cross-entropy loss value, and the second determining module 504 includes: The third determining submodule is used to determine the binary cross-entropy loss value and the multi-class cross-entropy loss value of the first one-dimensional convolutional neural network model based on the sample swelling prediction probability and the swelling label value of the sample vascular anastomosis blockage image.

[0076] Optionally, the adjustment module 505 includes: The adjustment submodule is used to adjust the model parameters of the first one-dimensional convolutional neural network model once based on the binary cross-entropy loss value and the multi-class cross-entropy loss value, so as to obtain the first one-dimensional convolutional neural network model with updated parameters. The execution submodule is used to re-execute the acquisition of the sample vascular anastomosis blockage images of the sample patient in order to start the next model training; The stop adjustment submodule is used to stop adjusting the model parameters and hyperparameters of the first one-dimensional convolutional neural network model when the number of consecutive occurrences of the binary cross-entropy loss value being less than the first preset threshold is greater than or equal to the first occurrence, and the number of consecutive occurrences of the multi-class cross-entropy loss value being less than the second preset threshold is greater than or equal to the first occurrence, so as to obtain the target vascular anastomosis site occlusion identification model.

[0077] The model training device in this application embodiment can be integrated into an electronic device, or a component in an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or other devices besides a terminal. For example, the electronic device can be a GPU box, a mobile phone, a robot, an ultra-mobile personal computer (UMPC), or a personal digital assistant (PDA), etc., and can also be a server, a personal computer (PC), etc. This application embodiment does not specifically limit the specific implementation.

[0078] The model training device in this application embodiment can be a device with an operating system. This operating system can be Android, Linux, Windows, or other possible operating systems; this application embodiment does not specifically limit it.

[0079] The model training device provided in this application embodiment can achieve... Figure 1 and Figure 2 The various processes implemented in the method embodiments will not be described again here to avoid repetition. Reference Figure 6 , Figure 6 This application provides a logic block diagram of a vascular anastomosis blockage identification device 600, which may include: The second acquisition module 601 is used to acquire an image of the target vascular anastomosis site of the target patient; wherein, the image capture point of the target vascular anastomosis blockage image is located outside the plane formed by the direction of vascular flow and the direction of blockage and swelling; The third determining module 602 is used to determine the first skin surface gradient sequence of the target blood vessel anastomosis site image; The second input / output module 603 is used to input the first skin surface gradient sequence into the target vascular anastomosis site blockage recognition model to obtain the first swelling prediction probability output by the target vascular anastomosis site blockage recognition model; wherein, the target vascular anastomosis site blockage recognition model is obtained based on the model training method described in any of the above. The fourth determining module 604 is used to determine the blockage identification result of the target patient based on the first swelling prediction probability.

[0080] Optionally, the third determining module 602 includes: The fourth determining submodule is used to determine a first pixel sequence image with dimension 1×N in the target blood vessel anastomosis site image along the anastomotic vessel corresponding to the target blood vessel anastomosis site image; The second decomposition submodule is used to decompose the first pixel sequence image according to the HSI color model to obtain the first brightness sequence image. The third calculation submodule is used to calculate the brightness difference between adjacent pixels in the first brightness sequence image item by item to obtain the first skin surface gradient sequence.

[0081] Optionally, the third determining module 602 includes: The fourth determining submodule is used to determine the first skin surface contour of the target blood vessel anastomosis site image; The second sampling submodule is used to perform equidistant sampling on the contour of the first skin surface to obtain multiple sampling points; The fourth calculation submodule is used to calculate the positional deviation between adjacent sampling points one by one to obtain the first skin surface gradient sequence.

[0082] The vascular anastomosis site blockage identification device in this application embodiment can be integrated into an electronic device, or a component of an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or other devices besides a terminal. For example, the electronic device can be a GPU BOX, mobile phone, tablet computer, laptop computer, PDA, robot, ultra-mobile personal computer (UMPC), or personal digital assistant (PDA), etc., and can also be a server, personal computer (PC), etc. This application embodiment does not specifically limit the specific implementation.

[0083] The vascular anastomosis site blockage identification device in this application embodiment can be a device with an operating system. This operating system can be Android, Linux, Windows, or other possible operating systems; this application embodiment does not specifically limit it.

[0084] The vascular anastomosis site occlusion identification device provided in this application embodiment can achieve... Figure 3 The various processes implemented in the method implementation examples will not be described again here to avoid repetition.

[0085] This application provides an electronic device, see [link to relevant documentation] Figure 7 The electronic device 70 includes: a processor 701, a memory 702, and a computer program 7021 stored in the memory 702 and executable on the processor 701. When the processor 701 executes the program, it implements the model training method of the foregoing embodiments, or the method for identifying vascular anastomosis site blockage.

[0086] This application also provides a computer-readable storage medium storing a computer program / instruction thereon, which, when executed by a processor, implements the steps in the model training method or the vascular anastomosis site blockage identification method disclosed in this application.

[0087] This application also provides a computer program product that, when run on an electronic device, enables the processor to implement the steps in the model training method or the vascular anastomosis site blockage identification method disclosed in this application.

[0088] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0089] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, electronic devices, and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0090] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0091] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0092] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0093] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0094] The above provides a detailed description of the model training method, the method and apparatus for identifying vascular anastomosis blockages provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and its core ideas. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A model training method, characterized in that, The method includes: Acquire images of vascular anastomosis and blockage in a sample patient; wherein the image capture point is located outside the plane formed by the direction of blood flow and the direction of blockage and swelling; Determine the gradient sequence of the sample skin surface in the sample vascular anastomosis and occlusion image; The gradient sequence of the sample skin surface is input into the first one-dimensional convolutional neural network model to be trained to obtain the sample swelling prediction probability output by the first one-dimensional convolutional neural network model. Based on the predicted probability of swelling in the sample and the swelling label value of the vascular anastomosis blockage image of the sample, the model loss value of the first one-dimensional convolutional neural network model is determined. Based on the model loss value, the model parameters and hyperparameters of the first one-dimensional convolutional neural network model are adjusted to obtain a target vascular anastomosis site stasis identification model; the target vascular anastomosis site stasis identification model is used to determine the stasis identification result of the target patient based on the target vascular anastomosis site image of the target patient.

2. The method according to claim 1, characterized in that, The step of determining the sample skin surface gradient sequence of the sample vascular anastomosis blockage image includes: Along the anastomotic vessel corresponding to the sample blood vessel anastomosis blockage image, determine a first pixel sequence image with dimension 1×N in the sample blood vessel anastomosis blockage image; According to the HSI color model, the first pixel sequence image is decomposed to obtain the first brightness sequence image; The brightness difference between adjacent pixels in the first brightness sequence image is calculated item by item to obtain the gradient sequence of the sample skin surface.

3. The method according to claim 1, characterized in that, The step of determining the sample skin surface gradient sequence of the sample vascular anastomosis blockage image includes: Determine the skin surface contour of the sample in the vascular anastomosis and blockage image; The surface contour of the sampled skin was sampled at equal intervals to obtain multiple sampling points; The positional deviation between adjacent sampling points is calculated one by one to obtain the gradient sequence of the sample skin surface.

4. The method according to claim 1, characterized in that, The model loss value includes a binary cross-entropy loss value and a multi-class cross-entropy loss value. The determination of the model loss value of the first one-dimensional convolutional neural network model based on the sample swelling prediction probability and the swelling annotation value of the sample vascular anastomosis blockage image includes: Based on the predicted probability of swelling in the sample and the swelling label value of the vascular anastomosis blockage image of the sample, the binary cross-entropy loss value and the multi-class cross-entropy loss value of the first one-dimensional convolutional neural network model are determined respectively.

5. The method according to claim 4, characterized in that, The step of adjusting the model parameters and hyperparameters of the first one-dimensional convolutional neural network model based on the model loss value to obtain the target vascular anastomosis site occlusion identification model includes: Based on the binary cross-entropy loss value and the multi-class cross-entropy loss value, the model parameters of the first one-dimensional convolutional neural network model are adjusted once to obtain the first one-dimensional convolutional neural network model with updated parameters. Re-execute the process of acquiring the sample vascular anastomosis and blockage images of the sample patient to begin the next model training; When the number of consecutive occurrences of the binary cross-entropy loss value being less than the first preset threshold is greater than or equal to the first occurrence, and the number of consecutive occurrences of the multi-class cross-entropy loss value being less than the second preset threshold is greater than or equal to the first occurrence, the adjustment of the model parameters and hyperparameters of the first one-dimensional convolutional neural network model is stopped, and the target vascular anastomosis site occlusion identification model is obtained.

6. A method for identifying vascular anastomosis blockage, characterized in that, The method includes: Acquire images of the target vascular anastomosis site in the target patient; wherein the image of the target vascular anastomosis blockage is taken from a point outside the plane formed by the direction of vascular flow and the direction of blockage and swelling; Determine the first skin surface gradient sequence of the image of the target vascular anastomosis site; The first skin surface gradient sequence is input into the target vascular anastomosis site stasis identification model to obtain the first swelling prediction probability output by the target vascular anastomosis site stasis identification model; wherein, the target vascular anastomosis site stasis identification model is obtained based on the model training method described in any one of claims 1 to 5; Based on the first swelling prediction probability, the blockage identification result of the target patient is determined.

7. The method according to claim 6, characterized in that, The first skin surface gradient sequence for determining the image of the target vascular anastomosis site includes: Along the anastomotic vessel corresponding to the target blood vessel anastomosis site image, a first pixel sequence image with dimension 1×N is determined in the target blood vessel anastomosis site image; According to the HSI color model, the first pixel sequence image is decomposed to obtain the first brightness sequence image; The brightness difference between adjacent pixels in the first brightness sequence image is calculated item by item to obtain the first skin surface gradient sequence.

8. The method according to claim 6, characterized in that, The first skin surface gradient sequence for determining the image of the target vascular anastomosis site includes: Determine the first skin surface contour of the image of the target vascular anastomosis site; The contour of the first skin surface is sampled at equal intervals to obtain multiple sampling points; The positional deviation between adjacent sampling points is calculated one by one to obtain the first skin surface gradient sequence.

9. A model training device, characterized in that, The device includes: The first acquisition module is used to acquire an image of the sample blood vessel anastomosis blockage in the sample patient; wherein the image capture point of the sample blood vessel anastomosis blockage is located outside the plane formed by the direction of blood vessel flow and the direction of blockage and swelling; The first determining module is used to determine the gradient sequence of the sample skin surface in the sample vascular anastomosis blockage image; The first input-output module is used to input the gradient sequence of the sample skin surface into the first one-dimensional convolutional neural network model to be trained, and to obtain the sample swelling prediction probability output by the first one-dimensional convolutional neural network model. The second determining module is used to determine the model loss value of the first one-dimensional convolutional neural network model based on the sample swelling prediction probability and the swelling label value of the sample vascular anastomosis blockage image. An adjustment module is used to adjust the model parameters and hyperparameters of the first one-dimensional convolutional neural network model based on the model loss value to obtain a target vascular anastomosis site stasis identification model; the target vascular anastomosis site stasis identification model is used to determine the stasis identification result of the target patient based on the target vascular anastomosis site image of the target patient.

10. A device for identifying vascular anastomosis blockage, characterized in that, The device includes: The second acquisition module is used to acquire images of the target vascular anastomosis site of the target patient; wherein, the image capture point of the target vascular anastomosis blockage image is located outside the plane formed by the direction of vascular flow and the direction of blockage and swelling; The third determining module is used to determine the first skin surface gradient sequence of the image of the target blood vessel anastomosis site; The second input / output module is used to input the first skin surface gradient sequence into the target vascular anastomosis site blockage recognition model to obtain the first swelling prediction probability output by the target vascular anastomosis site blockage recognition model; wherein, the target vascular anastomosis site blockage recognition model is obtained based on the model training method according to any one of claims 1 to 5; The fourth determining module is used to determine the blockage identification result of the target patient based on the first swelling prediction probability.

Images