Umbilical artery automatic detection method, device and equipment, storage medium and program product

Abstract

The invention relates to the technical field of intelligent medical treatment, in particular to an automatic umbilical artery detection method, device, equipment and medium, and the method comprises the steps: obtaining a color Doppler ultrasound image of an abdominal wall passing through a bladder horizontal section; the color Doppler ultrasound image is input into an image classification model, and the image classification model outputs the umbilical artery category to which the bladder target belongs in the color Doppler ultrasound image; inputting the color Doppler ultrasound image into a target detection model, outputting a detection candidate frame of the bladder target and the umbilical artery category by the target detection model, and determining the number of the umbilical arteries of the bladder target according to the detection candidate frame; and performing consistency verification according to the type of the umbilical artery and the number of the umbilical artery, if the consistency verification is passed, generating a labeled image and diagnosis information of the bladder target, otherwise, generating a manual auditing mark of the color Doppler ultrasound image. Therefore, the problems of low identification precision, weak anti-interference capability, high possibility of false detection, missing detection and the like in the case of fine structure difference in the prior art are solved.

Description

Technical Field

[0001] The present application relates to the field of intelligent medical technology, and in particular to an umbilical artery automatic detection method, device, equipment, storage medium and program product. Background Art

[0002] The normal fetal umbilical cord consists of two umbilical arteries and one umbilical vein. A single umbilical artery may be associated with risks such as developmental delay and chromosomal abnormalities, making early identification crucial for prenatal screening. Currently, screening relies heavily on manual analysis of ultrasound images by physicians, which is susceptible to image quality and operator experience, and carries the risk of false positives and missed detections. With the advancement of deep learning, object detection models have been widely used for automated umbilical artery identification, effectively improving both efficiency and accuracy.

[0003] However, related technologies still have limitations. Current deep learning-based detection methods have low accuracy for small, slender umbilical structures and are insufficiently robust to image noise, blur, and occlusion. Furthermore, due to subtle differences in the number and structure of umbilical vessels, traditional algorithms are prone to blurred boundaries, false detections, and missed detections, impacting the accuracy and stability of practical applications. Summary of the Invention

[0004] The present application provides an umbilical artery automatic detection method, apparatus, equipment, storage medium and program product to solve the problems of low recognition accuracy, weak anti-interference ability, and easy false detection and missed detection when facing subtle structural differences in related technologies.

[0005] A first aspect of the present application provides an automatic umbilical artery detection method, comprising the following steps: acquiring a color Doppler ultrasound image of a horizontal cross-section of the abdominal wall through the bladder; inputting the color Doppler ultrasound image into an image classification model, the image classification model outputting the umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs; inputting the color Doppler ultrasound image into a target detection model, the target detection model outputting detection candidate frames for the bladder target and the umbilical artery category, and determining the number of umbilical arteries of the bladder target based on the detection candidate frames; performing a consistency check based on the umbilical artery category and the number of umbilical arteries, and if the consistency check passes, generating a labeled image and diagnostic information of the bladder target; otherwise, generating a manual review mark for the color Doppler ultrasound image.

[0006] Optionally, in one embodiment of the present application, the image classification model includes a Stem module, a four-level Stage module and a classification head, wherein the Stem module extracts global features of the entire image through global average pooling, the four-level Stage module obtains the category prediction probability after Softmax normalization, and the classification head determines the umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs based on the category prediction probability.

[0007] Optionally, in one embodiment of the present application, the target detection model includes a lightweight backbone network and a multi-scale feature fusion module, wherein the lightweight backbone network extracts spatial features of targets of different scales in color Doppler ultrasound images, and the multi-scale feature fusion module fuses the spatial features of targets of different scales and outputs detection candidate boxes for bladder targets and umbilical artery categories.

[0008] Optionally, in one embodiment of the present application, the detected candidate box includes a category label, a confidence level, and bounding box coordinate information.

[0009] Optionally, in one embodiment of the present application, determining the number of umbilical arteries of a bladder target based on a detection candidate box includes: determining the bladder target and the umbilical artery based on a category label; if the size of the bladder target is larger than a preset size and the confidence of the bladder target is greater than a preset confidence, determining that the bladder target exists in the color Doppler ultrasound image; calculating the distance between the candidate boxes of all umbilical arteries and the candidate box of the bladder target with the coordinate information of the bounding box of the bladder target as the center, and calculating a comprehensive score for each umbilical artery based on the confidence and the distance; and determining the number of umbilical arteries of the bladder target from all umbilical arteries based on the comprehensive score.

[0010] Optionally, in one embodiment of the present application, if there is no bladder target in the color Doppler ultrasound image, the number of umbilical arteries of the bladder target is determined according to the confidence levels of all umbilical arteries.

[0011] A second aspect of the present application provides an automatic umbilical artery detection device, comprising: an acquisition module for acquiring a color Doppler ultrasound image of a horizontal cross-section of the abdominal wall through the bladder; a classification module for inputting the color Doppler ultrasound image into an image classification model, the image classification model outputting the umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs; a determination module for inputting the color Doppler ultrasound image into a target detection model, the target detection model outputting detection candidate frames for the bladder target and the umbilical artery category, and determining the number of umbilical arteries of the bladder target based on the detection candidate frames; and a verification module for performing a consistency check based on the umbilical artery category and the number of umbilical arteries. If the consistency check passes, a labeled image and diagnostic information of the bladder target are generated; otherwise, a manual review mark of the color Doppler ultrasound image is generated.

[0012] Optionally, in one embodiment of the present application, the image classification model includes a Stem module, a four-level Stage module and a classification head, wherein the Stem module extracts global features of the entire image through global average pooling, the four-level Stage module obtains the category prediction probability after Softmax normalization, and the classification head determines the umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs based on the category prediction probability.

[0013] Optionally, in one embodiment of the present application, the target detection model includes a lightweight backbone network and a multi-scale feature fusion module, wherein the lightweight backbone network extracts spatial features of targets of different scales in color Doppler ultrasound images, and the multi-scale feature fusion module fuses the spatial features of targets of different scales and outputs detection candidate boxes for bladder targets and umbilical artery categories.

[0014] Optionally, in one embodiment of the present application, the detected candidate box includes a category label, a confidence level, and bounding box coordinate information.

[0015] Optionally, in one embodiment of the present application, the determination module is further used to determine the bladder target and the umbilical artery based on the category label; if the size of the bladder target is larger than a preset size and the confidence of the bladder target is greater than a preset confidence, then it is determined that the bladder target exists in the color Doppler ultrasound image; with the bounding box coordinate information of the bladder target as the center, the distance between the candidate boxes of all umbilical arteries and the candidate box of the bladder target is calculated, and a comprehensive score for each umbilical artery is calculated based on the confidence and distance; and the number of umbilical arteries of the bladder target is determined from all umbilical arteries based on the comprehensive score.

[0016] Optionally, in one embodiment of the present application, there is no bladder target in the color Doppler ultrasound image, and the number of umbilical arteries of the bladder target is determined according to the confidence levels of all umbilical arteries.

[0017] A third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the automatic umbilical artery detection method of the above embodiment.

[0018] A fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the automatic umbilical artery detection method as described in the above embodiment.

[0019] A fifth embodiment of the present application provides a computer program, which, when executed, is used to implement the automatic umbilical artery detection method as described in the above embodiment.

[0020] Therefore, this application has the following beneficial effects:

[0021] First, a color Doppler ultrasound image of the fetal abdominal wall, taken through the bladder, is acquired as the baseline data for subsequent analysis. This image is then fed into an image classification model, which determines the type of umbilical artery corresponding to the bladder target in the image. The same image is then fed into a target detection model, which outputs candidate detection boxes for the bladder target and its corresponding umbilical artery in the image. These candidate boxes can be used to determine the actual number of umbilical arteries. A consistency check is then performed based on the consistency between the classification and detection results. If the check passes, meaning the classification and detection results are consistent, an image with annotations of the bladder target and umbilical artery structures is generated, and the corresponding diagnostic information is output. If the check fails, a manual review mark is added to the image, indicating that further confirmation by a physician is required. This approach addresses the problems of low recognition accuracy, weak anti-interference capabilities, and the tendency to false or missed detections when encountering subtle structural differences, often associated with related technologies.

[0022] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0024] Figure 1 is a flow chart of a method for automatic umbilical artery detection according to an embodiment of the present application;

[0025] Figure 2 Schematic diagram of the overall structure according to an embodiment of the present application;

[0026] Figure 3 Schematic diagram of the reasoning process according to an embodiment of the present application;

[0027] Figure 4 This is a schematic diagram of structure-guided umbilical artery screening according to an embodiment of the present application;

[0028] Figure 5 Schematic diagram of classification, detection, fusion and determination according to an embodiment of the present application;

[0029] Figure 6 is an exemplary diagram of an automatic umbilical artery detection device according to an embodiment of the present application;

[0030] Figure 7 Schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] The following describes in detail embodiments of the present application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0032] The following describes, with reference to the accompanying drawings, the automatic umbilical artery detection method, apparatus, device, storage medium, and program product according to embodiments of the present application. To address the issues mentioned in the background art, the present application provides an automatic umbilical artery detection method. In this method, a color Doppler ultrasound image of a transvesical section of the fetal abdominal wall is first acquired as the basis for subsequent analysis. This image is then input into an image classification model, which determines the umbilical artery category corresponding to the bladder target in the image. The same image is then input into a target detection model, which outputs candidate detection boxes for the bladder target and its corresponding umbilical artery in the image. These candidate boxes can be used to determine the actual number of umbilical arteries. A consistency check is then performed based on the consistency between the classification and detection results. If the check passes, meaning the classification and detection results are consistent, an image with annotations of the bladder target and umbilical artery structures is generated, and corresponding diagnostic information is output. If the check fails, a manual review mark is added to the image, indicating that further confirmation of the diagnosis by a physician is required. This method overcomes the problems of low recognition accuracy, weak anti-interference capabilities, and the tendency to false or missed detections when dealing with subtle structural differences in related technologies.

[0033] Specifically, Figure 1 A schematic flow chart of an automatic umbilical artery detection method provided in an embodiment of the present application.

[0034] like Figure 1 As shown, the umbilical artery automatic detection method includes the following steps:

[0035] In step S101 , a color Doppler ultrasound image of a transbladder horizontal section of the abdominal wall is acquired.

[0036] The abdominal wall is the tissue layer on the front of the abdomen of a fetus or pregnant woman, including skin, fat, muscle, and other structures. During ultrasound examinations, the abdominal wall is one of the pathways for image penetration and formation. Color Doppler is an ultrasound imaging technique that detects the speed and direction of tissue or blood flow, displaying them as different colors to observe blood flow.

[0037] It is understandable that color Doppler imaging technology can clearly present the direction and velocity of blood flow, so that the umbilical artery and umbilical vein have good visibility and resolution in the image.

[0038] After the color Doppler ultrasound image is collected, the embodiment of the present application performs standardization preprocessing on it, including operations such as uniform resolution, normalized pixel values, and format conversion, to ensure the consistency and effectiveness of the input of subsequent image classification and target detection models. Figure 2 As shown, the image input module 10 implemented in the present application is used to acquire color Doppler ultrasound images of the abdominal wall through the bladder horizontal section. The acquired images have standard ultrasound resolution and grayscale or color flow information. After the acquisition is completed, the image will be standardized, including: scaling the image size to a fixed resolution (such as 640×640 pixels), normalizing the pixel values ​​of the three RGB channels (using mean values ​​of 0.485, 0.456, and 0.406 and standard deviations of 0.229, 0.224, and 0.225, respectively), and uniformly converting to RGB format. The preprocessed images will serve as input for subsequent classification reasoning and detection reasoning to ensure the consistency and standardization of the input data.

[0039] In step S102, the color Doppler ultrasound image is input into an image classification model, and the image classification model outputs the umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs.

[0040] The image classification model uses a lightweight convolutional neural network optimized based on the ConvNeXt Tiny architecture. This lightweight convolutional neural network, optimized based on the ConvNeXt Tiny architecture, is a highly efficient and high-performance deep learning model. By simplifying the network structure and reducing the number of parameters, it achieves faster computing speed and lower resource consumption while maintaining strong feature extraction capabilities, making it suitable for fast and accurate classification of color Doppler ultrasound images.

[0041] It is understandable that automatically identifying the umbilical artery category corresponding to the bladder area in the color Doppler ultrasound image through the image classification model can greatly improve the efficiency and accuracy of diagnosis and reduce human judgment errors. Figure 3 As shown, the pre-processed image is input to the image classification module 20 for reasoning. The image classification module uses a ConvNeXt Tiny-based network structure to analyze the input image and preliminarily determine whether it belongs to the single umbilical artery or double umbilical artery category, and finally outputs the image classification result.

[0042] In one embodiment of the present application, the image classification model includes a Stem module, a four-level Stage module and a classification head, wherein the Stem module extracts global features of the entire image through global average pooling, the four-level Stage module obtains the category prediction probability after Softmax normalization, and the classification head determines the umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs based on the category prediction probability.

[0043] Among them, global average pooling is a pooling operation that averages the values ​​of each channel of the entire feature map to obtain a global feature representation. The four-level Stage module refers to a deep structure composed of four stages in the model. Each stage is responsible for extracting higher-level features and gradually enriching the expression of the image. Softmax normalization is a method of converting the raw scores output by the model into a probability distribution, so that the predicted probabilities of all categories add up to 1, which facilitates category judgment. Category prediction probability refers to the probability value that the model predicts that the image belongs to each category. The higher the value, the more certain the model is that the image belongs to that category. The classification head is the final decision-making part of the model. It determines the final classification result based on the predicted category probability and outputs the umbilical artery category of the bladder target in the color Doppler ultrasound image.

[0044] It's clear that the structured image classification model achieves accurate identification of the bladder target umbilical artery in color Doppler ultrasound images. The Stem module extracts global features, ensuring the model captures the overall image information. The four-level Stage module refines features in stages, improving classification detail and accuracy. The classification head makes the final judgment based on probabilistic output. This hierarchical processing and probabilistic assessment reduces the possibility of false positives or missed detections.

[0045] Taking one embodiment of the present application as an example, the image classification model consists of a Stem module, a four-stage Stage module, and a classification head. The input image resolution is 224×224, and the output is one of two categories: single umbilical artery or double umbilical artery. The model extracts the overall image features through global average pooling, obtains the category prediction probability after softmax normalization, and finally uses the category with the highest probability as the preliminary judgment of the umbilical artery type in the image. This classification process is independent of the subsequent object detection reasoning, ensuring the stability and reliability of the reasoning process.

[0046] In step S103, the color Doppler ultrasound image is input into the target detection model, which outputs detection candidate boxes for the bladder target and the umbilical artery category, and determines the number of umbilical arteries of the bladder target based on the detection candidate boxes.

[0047] In one embodiment of the present application, the target detection model uses an improved YOLO11 structure network for inference. YOLO11 typically includes 11 convolutional layers, which are used to extract multi-level features of images, quickly and accurately detect targets in images in an end-to-end manner, and achieve real-time target recognition.

[0048] It can be understood that by accurately locating and classifying the bladder target and its umbilical artery in the color Doppler ultrasound image through the target detection model, the position and category information of each umbilical artery can be accurately identified. Figure 3As shown, in this embodiment of the present application, a standardized image is input to the object detection module 30, which uses a modified YOLO11 network for inference. The input image is first processed by two different modules. The object detection module uses the YOLO11 model to detect the bladder and umbilical artery in the image, generate corresponding candidate boxes and their confidence information, and output the detected locations and confidence levels of the bladder and umbilical artery.

[0049] In one embodiment of the present application, the target detection model includes a lightweight backbone network and a multi-scale feature fusion module, wherein the lightweight backbone network extracts the spatial features of targets of different scales in color Doppler ultrasound images, and the multi-scale feature fusion module fuses the spatial features of targets of different scales and outputs detection candidate boxes for bladder targets and umbilical artery categories.

[0050] The lightweight backbone network is a neural network with a simplified structure and high computational efficiency. In this embodiment of the application, it is used to extract spatial features of objects of different sizes from color Doppler ultrasound images. These features include information such as the target's location, shape, and texture. The multi-scale feature fusion module integrates spatial features from different scales to enhance the model's ability to recognize objects of various sizes.

[0051] It can be understood that efficient and accurate feature extraction is achieved through the lightweight backbone network, and multi-scale feature fusion enhances the model's ability to recognize umbilical artery targets of different sizes and shapes, accurately locates the bladder and its umbilical artery, and effectively distinguishes between single and double umbilical arteries.

[0052] In one embodiment of the present application, the detection candidate box includes a category label, a confidence level, and bounding box coordinate information.

[0053] The category label indicates the category of the detected object, which in this example is the bladder or umbilical artery. The confidence level indicates the model's confidence or probability in the detection result. The bounding box coordinates are used to locate the specific position of the object in the image, typically expressed as the coordinates of the upper left corner of the rectangular box and its width and height.

[0054] It can be understood that the detection candidate box provides accurate target category and location information, which facilitates the subsequent accurate identification and positioning of the bladder target and umbilical artery.

[0055] In one embodiment of the present application, determining the number of umbilical arteries of a bladder target based on a detection candidate box includes: determining the bladder target and the umbilical artery based on a category label; determining that the bladder target exists in the color Doppler ultrasound image if the size of the bladder target is larger than a preset size and the confidence level of the bladder target is larger than a preset confidence level; calculating the distance between the candidate boxes of all umbilical arteries and the candidate box of the bladder target with the coordinate information of the bounding box of the bladder target as the center, calculating a comprehensive score for each umbilical artery based on the confidence level and the distance; and determining the number of umbilical arteries of the bladder target from all umbilical arteries based on the comprehensive score.

[0056] The preset size and preset confidence level are fixed thresholds used to screen valid bladder targets, ensuring that the targets are large enough and that identification is reliable. For example, the preset size is set to no less than 5000 pixels2, or approximately 0.8% of the total image area, to filter out invalid targets such as artifacts and small blood vessels. The preset confidence level is set to 0.3, eliminating low-confidence targets while retaining true targets.

[0057] It can be understood that by combining the category label, confidence, and bounding box coordinate information output by the target detection model, bladder candidate frames with a size greater than a preset threshold and a confidence higher than a set lower limit are first screened out to ensure that the recognition results have sufficient credibility. Then, based on the position coordinates of the bladder target, the geometric distances between all umbilical artery candidate frames and it are calculated, and combined with the confidence of the umbilical artery candidate frame, a comprehensive score for each umbilical artery is obtained through a weighted scoring method. Finally, the final number of associations is determined from the umbilical arteries with the highest scores, thereby effectively eliminating false detection targets that may be far away from the bladder or have low credibility in the image.

[0058] The specific process is as follows: Figure 4 As shown, according to Figure 2 The output result of the target detection module 30 shown in the figure first determines whether a bladder target exists. If a detection frame of the bladder category exists and the confidence score is higher than the set threshold (such as 0.3), and the detection frame size is consistent with the conventional size range of ultrasound anatomy, then the bladder is considered to be present; otherwise, the bladder is considered to be absent. The result of the bladder existence determination will directly affect the subsequent umbilical artery candidate frame screening strategy. Figure 4 As shown in the figure, if the bladder is determined to be present, the center point of the bladder detection frame is used as a reference. For each umbilical artery candidate frame, the detection confidence score and the normalized Euclidean distance from the center point to the bladder center are calculated. The normalized distance is the actual distance from the center of the umbilical artery frame to the center of the bladder frame divided by the diagonal length of the entire image. The detection confidence score and the normalized Euclidean distance from the center point to the bladder center are then combined to calculate the comprehensive score of each umbilical artery candidate frame according to the weighted rule. The comprehensive score calculation formula is:

[0059] Comprehensive score = α × confidence score - β × normalized center distance

[0060] The weight parameters α and β can be determined based on experience or experiments. It is preferred to set α to be greater than β to ensure that confidence dominates the sorting. Sorting is done from high to low based on the comprehensive score, and the umbilical artery detection frame with the highest score is selected first.

[0061] In one embodiment of the present application, if no bladder target exists in the color Doppler ultrasound image, the number of umbilical arteries of the bladder target is determined according to the confidence levels of all umbilical arteries.

[0062] It is understandable that if the bladder is determined to be absent, no distance calculation is performed, and the umbilical artery confidence scores are directly sorted from high to low for screening. By determining the number of umbilical arteries based on the confidence scores of all umbilical artery candidate frames, it is possible to ensure that the umbilical artery can still be effectively identified and evaluated even when bladder information is absent, thereby reducing dependence on bladder detection results and improving the robustness of the embodiment of the present application.

[0063] In step S104, a consistency check is performed based on the umbilical artery type and the number of umbilical arteries. If the consistency check passes, a labeled image and diagnostic information of the bladder target are generated; otherwise, a manual review mark of the color Doppler ultrasound image is generated.

[0064] Consistency checking is the process of comparing and verifying two or more related results to determine whether they match each other or conform to expected rules. In this application, consistency checking involves comparing the umbilical artery category (such as single or double umbilical arteries) output by the image classification model with the number of umbilical arteries determined by the object detection model to check whether the two are consistent.

[0065] It is understandable that by performing consistency checks on the umbilical artery type and number, anomalies where the classification and test results do not match can be automatically screened and filtered out, reducing false positives and missed detections caused by the model's independent judgment. If the consistency check passes, a clearly labeled image and diagnostic report are automatically generated, improving diagnostic efficiency. If the check fails, a manual review process is triggered to ensure that complex or abnormal cases receive further manual confirmation, reducing the risk of errors in the automated process and enhancing the security and stability of the overall diagnostic process.

[0066] like Figure 5, compare the umbilical artery type output by the image classification module 20 with the number of umbilical artery detections. If the classification result is consistent with the number of umbilical artery detections (for example, it is classified as double umbilical arteries and two umbilical arteries are detected), the reasoning is deemed consistent and a diagnostic conclusion is generated directly; if the classification result is inconsistent with the number of detections, the image is marked as "pending review" and a manual secondary review is prompted. After completing the structural guidance and consistency verification, the diagnosis result output module 50 generates a color ultrasound image with an umbilical artery detection frame, a bladder detection frame, and an umbilical artery category annotation based on the fusion judgment result, and marks the inferred umbilical artery category (single umbilical artery or double umbilical artery) at the top center of the image. Synchronously generate an inference log file to record the image number, classification result, number of detections, consistency status, and whether it has entered manual review, to facilitate subsequent tracing and review.

[0067] According to the automatic umbilical artery detection method proposed in an embodiment of the present application, a color Doppler ultrasound image of a transbladder-level section of the fetal abdominal wall is first acquired as the basis for subsequent analysis. This image is then input into an image classification model, which determines the umbilical artery category corresponding to the bladder target in the image. The same image is then input into a target detection model, which outputs candidate detection boxes for the bladder target and its corresponding umbilical artery in the image. These candidate boxes can be used to determine the actual number of umbilical arteries. A consistency check is then performed based on the consistency between the classification and detection results. If the check passes, meaning the classification and detection results are consistent, an image with annotations of the bladder target and umbilical artery structures is generated, and the corresponding diagnostic information is output. If the check fails, a manual review mark is added to the image, indicating that further confirmation by a physician is required. This solves the problems of related technologies, such as low recognition accuracy, weak anti-interference capabilities, and the tendency to false detection and missed detection when dealing with subtle structural differences.

[0068] Next, the automatic umbilical artery detection device according to the embodiment of the present application will be described with reference to the accompanying drawings.

[0069] Figure 6 2 is a block diagram of an automatic umbilical artery detection device according to an embodiment of the present application.

[0070] like Figure 6 As shown, the umbilical artery automatic detection device 60 includes: an acquisition module 601 , a classification module 602 , a determination module 603 , and a verification module 604 .

[0071] Among them, the acquisition module 601 is used to obtain a color Doppler ultrasound image of a horizontal section of the abdominal wall through the bladder; the classification module 602 is used to input the color Doppler ultrasound image into an image classification model, and the image classification model outputs the umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs; the determination module 603 is used to input the color Doppler ultrasound image into a target detection model, and the target detection model outputs a detection candidate box for the bladder target and the umbilical artery category, and determines the number of umbilical arteries of the bladder target based on the detection candidate box; the verification module 604 is used to perform consistency verification based on the umbilical artery category and the number of umbilical arteries. If the consistency verification passes, a labeled image and diagnostic information of the bladder target are generated; otherwise, a manual review mark of the color Doppler ultrasound image is generated.

[0072] In one embodiment of the present application, the image classification model includes a Stem module, a four-level Stage module and a classification head, wherein the Stem module extracts global features of the entire image through global average pooling, the four-level Stage module obtains the category prediction probability after Softmax normalization, and the classification head determines the umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs based on the category prediction probability.

[0073] In one embodiment of the present application, the target detection model includes a lightweight backbone network and a multi-scale feature fusion module, wherein the lightweight backbone network extracts the spatial features of targets of different scales in color Doppler ultrasound images, and the multi-scale feature fusion module fuses the spatial features of targets of different scales and outputs detection candidate boxes for bladder targets and umbilical artery categories.

[0074] In one embodiment of the present application, the detection candidate box includes a category label, a confidence level, and bounding box coordinate information.

[0075] In one embodiment of the present application, the determination module 603 is further configured to determine a bladder target and an umbilical artery based on the category label; if the size of the bladder target is greater than a preset size and the confidence level of the bladder target is greater than a preset confidence level, then determining that a bladder target exists in the color Doppler ultrasound image; calculating the distance between the candidate boxes of all umbilical arteries and the candidate box of the bladder target with the bounding box coordinate information of the bladder target as the center, and calculating a comprehensive score for each umbilical artery based on the confidence level and the distance; and determining the number of umbilical arteries that are bladder targets from all umbilical arteries based on the comprehensive score.

[0076] In one embodiment of the present application, there is no bladder target in the color Doppler ultrasound image, and the number of umbilical arteries of the bladder target is determined according to the confidence levels of all umbilical arteries.

[0077] It should be noted that the aforementioned explanation of the embodiment of the automatic umbilical artery detection method is also applicable to the automatic umbilical artery detection device of this embodiment, and will not be repeated here.

[0078] According to the automatic umbilical artery detection device proposed in an embodiment of the present application, a color Doppler ultrasound image of a transbladder-level section of the fetal abdominal wall is first acquired as the basis for subsequent analysis. This image is then input into an image classification model, which determines the umbilical artery category corresponding to the bladder target in the image. The same image is then input into a target detection model, which outputs candidate detection boxes for the bladder target and its corresponding umbilical artery in the image. These candidate boxes can be used to determine the actual number of umbilical arteries. A consistency check is then performed based on the consistency between the classification and detection results. If the check passes, meaning the classification and detection results are consistent, an image with annotations of the bladder target and umbilical artery structures is generated, and the corresponding diagnostic information is output. If the check fails, a manual review mark is added to the image, indicating that further confirmation of the diagnosis by a physician is required. This solves the problems of related technologies, such as low recognition accuracy, weak anti-interference capabilities, and the tendency to false detection and missed detection when dealing with subtle structural differences.

[0079] Figure 7 This is a schematic diagram of the structure of an electronic product provided in an embodiment of the present application. The electronic product may include:

[0080] Memory 701 , processor 702 , and computer programs stored in the memory 701 and executable on the processor 702 .

[0081] When the processor 702 executes the program, the automatic umbilical artery detection method provided in the above embodiment is implemented.

[0082] Furthermore, the electronic product also includes:

[0083] The communication interface 703 is used for communication between the memory 701 and the processor 702 .

[0084] The memory 701 is used to store computer programs that can be run on the processor 702 .

[0085] The memory 701 may include a high-speed RAM (Random Access Memory) memory, and may also include a non-volatile memory, such as at least one disk memory.

[0086] If the memory 701, processor 702, and communication interface 703 are implemented independently, the communication interface 703, memory 701, and processor 702 can be connected to each other via a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0087] Optionally, in a specific implementation, if the memory 701, the processor 702 and the communication interface 703 are integrated on a chip, the memory 701, the processor 702 and the communication interface 703 can communicate with each other through an internal interface.

[0088] The processor 702 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0089] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the above-mentioned automatic umbilical artery detection method is implemented.

[0090] An embodiment of the present application further provides a computer program product, including a computer program or instructions, which, when executed, implements the above-mentioned automatic umbilical artery detection method.

[0091] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0092] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of this application, "N" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0093] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing a custom logical function or process step, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed in a different order than shown or discussed, including performing functions in a substantially simultaneous manner or in a reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application pertain.

[0094] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, the steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement the method: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array, a field programmable gate array, etc.

[0095] A person skilled in the art may understand that all or part of the steps carried out in the method for implementing the above-mentioned embodiment may be completed by instructing the relevant hardware through a program, and the above-mentioned program may be stored in a computer-readable storage medium, which, when executed, includes one of the steps of the method embodiment or a combination thereof.

[0096] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for automatic detection of umbilical artery, characterized in that: The following steps are involved: Color Doppler ultrasound images of the abdominal wall at the transvesical level were obtained; Inputting the color Doppler ultrasound image into an image classification model, wherein the image classification model outputs an umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs; Inputting the color Doppler ultrasound image into a target detection model, the target detection model outputting detection candidate boxes of bladder targets and umbilical artery categories, and determining the number of umbilical arteries of the bladder target according to the detection candidate boxes; A consistency check is performed based on the umbilical artery type and the number of umbilical arteries. If the consistency check passes, a labeled image and diagnostic information of the bladder target are generated; otherwise, a manual review mark of the color Doppler ultrasound image is generated.

2. The automatic umbilical artery detection method according to claim 1, characterized in that: The image classification model includes a Stem module, a four-level Stage module and a classification head, wherein the Stem module extracts global features of the entire image through global average pooling, the four-level Stage module obtains category prediction probability after Softmax normalization, and the classification head determines the umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs based on the category prediction probability.

3. The automatic umbilical artery detection method according to claim 1, characterized in that: The target detection model includes a lightweight backbone network and a multi-scale feature fusion module, wherein the lightweight backbone network extracts the spatial features of targets of different scales in the color Doppler ultrasound image, and the multi-scale feature fusion module fuses the spatial features of targets of different scales and outputs detection candidate boxes for bladder targets and umbilical artery categories.

4. The automatic umbilical artery detection method according to claim 1 or 3, characterized in that: The detection candidate box includes category label, confidence and bounding box coordinate information.

5. The automatic umbilical artery detection method according to claim 4, characterized in that: The determining the number of umbilical arteries of the bladder target according to the detected candidate frame includes: identifying a bladder target and an umbilical artery according to the class labels; If the size of the bladder target is larger than a preset size and the confidence level of the bladder target is larger than a preset confidence level, determining that the bladder target exists in the color Doppler ultrasound image; Taking the bounding box coordinate information of the bladder target as the center, calculating the distance between the candidate boxes of all umbilical arteries and the candidate box of the bladder target, and calculating the comprehensive score of each umbilical artery according to the confidence and the distance; The number of umbilical arteries of the bladder target is determined from all umbilical arteries according to the comprehensive score.

6. The automatic umbilical artery detection method according to claim 5, characterized in that: If the bladder target does not exist in the color Doppler ultrasound image, the number of umbilical arteries of the bladder target is determined according to the confidence levels of all umbilical arteries.

7. An automatic umbilical artery detection device, characterized in that: include: an acquisition module, used for acquiring a color Doppler ultrasound image of a transbladder horizontal section of the abdominal wall; a classification module, configured to input the color Doppler ultrasound image into an image classification model, wherein the image classification model outputs an umbilical artery category to which the bladder target in the color Doppler ultrasound image belongs; a determination module, configured to input the color Doppler ultrasound image into a target detection model, the target detection model outputting detection candidate boxes for bladder targets and umbilical artery categories, and determining the number of umbilical arteries of the bladder target based on the detection candidate boxes; A verification module is configured to perform consistency verification based on the umbilical artery type and the number of umbilical arteries, and if the consistency verification passes, generate a labeled image and diagnostic information of the bladder target; otherwise, generate a manual review mark of the color Doppler ultrasound image.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for automatic detection of the umbilical artery according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed, the automatic umbilical artery detection method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed, the automatic umbilical artery detection method according to any one of claims 1 to 6 is implemented.

Images