Standard plane identification and labor process parameter measurement method and device based on multi-task network

Abstract

The invention belongs to the technical field of fusion medicine and computer vision, and relates to a standard plane identification and labor process parameter measurement method and device based on a multi-task network. The method comprises the following steps: S1, acquiring a perineal ultrasound image; s2, using a preset multi-branch convolutional neural network MT-UNet to classify a standard plane image with a clear fetal head and pubic symphysis structure of the transperineal ultrasound image, and segmenting a fetal head and pubic symphysis area on the standard plane image; s3, performing ellipse fitting on the segmented fetal head region contour to obtain a fitted fetal head ellipse equation; and S4, calculating a birth process progress angle and a fetal head-pubic symphysis distance based on the fetal head elliptic equation. According to the method, the burden of doctors can be relieved, the working efficiency of the doctors can be improved, and meanwhile, the subjective defects of different results of different doctors in the birth progress angle and fetal head-pubic symphysis distance measurement of the same pregnant woman due to different professional levels and experiences are overcome.

Description

Technical Field

[0001] This application relates to the technical fields of fusion medicine and computer vision, and particularly relates to a method and device for standard plane recognition and intrapartum parameter measurement based on a multi-task network. Background Art

[0002] Clinically, the assessment of intrapartum progress mainly relies on vaginal digital examination. However, this method highly depends on the experience of the examiner, has strong subjectivity and relatively large deviations. Moreover, since vaginal digital examination is an invasive operation, it will bring great discomfort and huge psychological pressure to pregnant women, and even cause fear and refusal of examination, delaying clinical diagnosis and treatment. Repeated examinations will also bring the risk of intrauterine infection, which is not conducive to intrapartum management.

[0003] With the wide application of ultrasound imaging technology in medicine, some studies have attempted to apply ultrasound imaging technology to intrapartum assessment. Ultrasound imaging technology uses the reflected signals of sound waves to obtain organ images in the human body. Compared with vaginal digital examination, ultrasound examination belongs to imaging, has no radiation, lower cost, is more objective and accurate, and has good repeatability. Using ultrasound technology can reduce the number of vaginal digital examinations and the risk of intrauterine infection. In addition, ultrasound examination is non-invasive and is more acceptable to pregnant women.

[0004] Among intrapartum ultrasound parameters, the International Society of Ultrasound in Obstetrics and Gynecology Guidelines and the experts of the Obstetrics Group of the Chinese Medical Association's Society of Obstetrics and Gynecology have reached a consensus on the clinical management of the new intrapartum process, recommending measuring the Angle of Progression (AoP) and the head-symphysis distance (HSD) based on transperineal ultrasound (TPU) images to evaluate the fetal head position. The definition of AoP is the angle formed by the long axis of the symphysis pubis and the tangent line passing through the lower edge of the symphysis pubis and the farthest point of the fetal skull. The HSD is defined as the minimum distance between the lower edge of the symphysis pubis and the fetal head along the direction perpendicular to the long axis of the symphysis pubis. Figure 2 It is a schematic diagram for measuring AoP and HSD on the obtained ultrasound image. For the measurement of AoP and HSD, it is necessary to simultaneously observe the longitudinal sagittal plane of the symphysis pubis and the fetal head structure in the ultrasound image, that is, the standard plane image. According to the standard plane image, the contours of the fetal head and the symphysis pubis and the left and right endpoint coordinates of the symphysis pubis are determined, and then AoP and HSD are measured.

[0005] However, currently in clinical practice, the identification of the standard plane and the measurement of AoP and HSD are generally carried out manually. Usually, ultrasound physicians or attending physicians perform image analysis, which takes a long time and requires rich clinical experience and professional knowledge. Moreover, there is a high similarity between ultrasound images, making it difficult for ultrasound doctors to determine whether it is a standard plane, and there are also certain subjective differences in the measurement results of AoP and HSD by different doctors. Therefore, the present invention discloses an automatic measurement method for AoP and HSD based on TPU images, which can simply and accurately automatically measure AoP and HSD, lower the technical threshold for using intrapartum ultrasound equipment, assist doctors in objectively judging the progress of labor, and make scientific decisions. Summary of the Invention

[0006] The purpose of the embodiments of the present application is to propose a method and device for standard plane recognition and labor process parameter measurement based on a multi-task network, and its main purpose is to improve the efficiency and accuracy of standard plane recognition and labor process parameter measurement based on a multi-task network.

[0007] To solve the above technical problems, the embodiments of the present application provide a method for standard plane recognition and labor process parameter measurement based on a multi-task network, and adopt the following technical solutions:

[0008] The method for standard plane recognition and labor process parameter measurement based on a multi-task network includes the following steps:

[0009] S1. Collect transperineal ultrasound images;

[0010] S2. Use a pre-set multi-branch convolutional neural network MT-UNet to classify the standard plane image with clear fetal head and symphysis pubis structures in the transperineal ultrasound image, and segment the fetal head and symphysis pubis regions on the standard plane image;

[0011] S3. Perform elliptical fitting on the contour of the segmented fetal head region to obtain the fitted fetal head elliptical equation;

[0012] S4. Calculate the labor progress angle and the fetal head-symphysis pubis distance based on the fetal head elliptical equation.

[0013] Further, the step S1 further includes:

[0014] Divide the transperineal ultrasound images into two data sets for iterative optimization and measurement.

[0015] Further, the step S2 specifically includes the following steps:

[0016] S21. Design a multi-branch convolutional neural network MT-UNet. This network MT-UNet has a shared encoding part and a decoding part. The decoding part includes an image classification part, an image segmentation part, and a feature point localization part. The shared encoding part extracts feature information from intrapartum ultrasound images. The image classification part determines whether the ultrasound image belongs to a standard plane image. The image segmentation part extracts the fetal head and symphysis pubis regions. The feature point localization part locates the two endpoints of the symphysis pubis in the form of a Gaussian heatmap.

[0017] S22. Use the backpropagation algorithm to iteratively optimize the multi-branch convolutional neural network MT-UNet according to the dataset.

[0018] S23. Use the iteratively optimized multi-branch convolutional neural network MT-UNet to perform standard plane recognition and image segmentation on ultrasound images.

[0019] Furthermore, the shared encoding part in step S21 includes five convolutional modules, and each convolutional module includes two convolutional layers and one downsampling layer.

[0020] Furthermore, the image classification part in step S21 contains two residual connection convolutional blocks. Each residual connection convolutional block includes a convolutional layer, a regularization layer, and an activation layer. The two residual connection convolutional blocks are connected to an average pooling operation. The average pooling is then connected to a fully connected layer and an activation layer. The classification feature map has two channels, respectively indicating whether it belongs to the standard plane.

[0021] Furthermore, the image segmentation part in step S21 contains 4 upsampling convolutional modules and an activation layer. Each upsampling convolutional module includes an upsampling layer and two convolutional layers. Here, bilinear sampling is used to enlarge the image size and halve the number of image channels. Subsequently, skip connections are used to splice the feature maps with the same resolution in the contraction stage and the feature maps in the current expansion stage.

[0022] Furthermore, step S3 specifically includes the following steps:

[0023] S31. Screen the fetal head and symphysis pubis regions, and retain the connected domain with the largest pixel area through connected component processing.

[0024] S32. Use the Sobel edge detection algorithm to extract the contour of the fetal head region.

[0025] S33. Use the least squares method to perform elliptical fitting on the contour of the fetal head region.

[0026] To solve the above technical problems, the embodiment of the present application also provides a standard plane recognition and intrapartum parameter measurement device based on a multi-task network, and adopts the following technical solutions:

[0027] A standard plane recognition and labor process parameter measurement device based on a multi-task network, comprising:

[0028] An acquisition module for acquiring transperineal ultrasound images;

[0029] A classification and segmentation module for classifying a standard plane image with clear fetal head and symphysis pubis structures from the transperineal ultrasound image by using a preset multi-branch convolutional neural network MT-UNet, and segmenting the fetal head and symphysis pubis regions on the standard plane image;

[0030] A fitting module for performing elliptical fitting on the contour of the segmented fetal head region to obtain a fitted fetal head ellipse equation;

[0031] A calculation module for calculating the labor progress angle and the fetal head-symphysis pubis distance based on the fetal head ellipse equation.

[0032] To solve the above technical problems, an embodiment of the present application further provides a computer device, adopting the following technical solution:

[0033] A computer device, comprising a memory and a processor, wherein computer-readable instructions are stored in the memory, and when the processor executes the computer-readable instructions, the steps of the above-mentioned standard plane recognition and labor process parameter measurement method based on a multi-task network are implemented.

[0034] To solve the above technical problems, an embodiment of the present application further provides a computer-readable storage medium, adopting the following technical solution:

[0035] A computer-readable storage medium, characterized in that computer-readable instructions are stored on the computer-readable storage medium, and when the computer-readable instructions are executed by a processor, the steps of the above-mentioned standard plane recognition and labor process parameter measurement method based on a multi-task network are implemented.

[0036] Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects: The present application designs a multi-task deep learning network that integrates standard plane recognition and intrapartum parameter measurement, and uses image classification, image segmentation, and feature point localization to automatically identify the standard plane and realize the automatic measurement of the intrapartum progress angle and the fetal head-pubic symphysis distance. Compared with a single-task single model, sharing a single model for multiple tasks can reduce memory usage and speed up the inference speed. At the same time, multiple related tasks complement each other by sharing information, which can improve the network performance. Experimental results show that the multi-task algorithm of the present application can effectively identify the standard plane, segment the fetal head-pubic symphysis region, and locate the endpoints of the pubic symphysis. The present application can reduce the burden on doctors, help improve the work efficiency of doctors, and at the same time eliminate the subjectivity drawbacks of different measurement results of the intrapartum progress angle and the fetal head-pubic symphysis distance of the same pregnant woman caused by different professional levels and experiences. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the solutions in the present application, the following will briefly introduce the drawings required for the description of the embodiments of the present application. Obviously, the following described drawings are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0038] Figure 1 It is an exemplary system architecture diagram to which the present application can be applied;

[0039] Figure 2 A flowchart of an embodiment of the method for standard plane recognition and intrapartum parameter measurement based on a multi-task network according to the present application;

[0040] Figure 3 It is a schematic diagram of an embodiment of the device for standard plane recognition and intrapartum parameter measurement based on a multi-task network according to the present application;

[0041] Figure 4 It is a schematic diagram of the AoP and HSD measurement methods in the embodiments of the present application;

[0042] Figure 5 It is the MT-UNet network structure diagram in the embodiments of the present application;

[0043] Figure 6 It is a statistical chart of the comparison of the results of measuring AoP and HSD using the method and manual measurement in the embodiments of the present application;

[0044] Figure 7 It is a schematic diagram of the device for standard plane recognition and intrapartum parameter measurement based on a multi-task network according to the present application;

[0045] Figure 8It is a schematic structural diagram of an embodiment of a computer device according to the present application. Detailed implementation manners

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs; the terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above drawings are intended to cover non-exclusive inclusion. The terms "first", "second", etc. in the specification and claims of this application or the above drawings are used to distinguish different objects and not to describe a specific order.

[0047] Reference to "embodiment" herein means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of this application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0048] In order to enable those skilled in the technical field to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0049] As Figure 1 shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0050] Users can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0051] The terminal devices 101, 102, and 103 can be various electronic devices with a display screen and supporting web browsing, including but not limited to smartphones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, and so on.

[0052] The server 105 can be a server that provides various services, such as a background server that supports the pages displayed on the terminal devices 101, 102, and 103.

[0053] It should be noted that the method for standard plane recognition and intrapartum parameter measurement based on a multi-task network provided in the embodiments of the present application is generally executed by the server. Correspondingly, the device for standard plane recognition and intrapartum parameter measurement based on a multi-task network is generally set in the terminal device.

[0054] It should be understood that Figure 1 the numbers of the terminal devices, network, and server in are merely illustrative. According to the implementation requirements, there can be any number of terminal devices, network, and server.

[0055] Continuing to refer to Figure 2 、 Figure 3 , a flowchart of an embodiment of the method for standard plane recognition and intrapartum parameter measurement based on a multi-task network according to the present application is shown. The method for standard plane recognition and intrapartum parameter measurement based on a multi-task network includes the following steps:

[0056] Step S1: Collect transperineal ultrasound images.

[0057] In this embodiment, the device used is produced by Philips Company. A total of 2445 ultrasound images are collected, including 1045 standard plane images and 1400 non-standard plane images. The resolution of each image is 512*416, and they are divided into two data sets, namely the iterative optimization set and the test set, according to a ratio of 7:3.

[0058] Step S2: Use a pre-set multi-branch convolutional neural network MT-UNet to classify the standard plane images with clear fetal head and symphysis pubis structures from the transperineal ultrasound images, and segment the fetal head and symphysis pubis regions on the standard plane images.

[0059] In this embodiment, step S2 specifically includes the following steps:

[0060] Step S21: Design a multi-branch convolutional neural network MT-UNet, which has a shared encoding part and a decoding part.

[0061] The decoding part includes three branches: an image classification part, an image segmentation part, and a feature point localization part, as Figure 5 shown.

[0062] The shared encoding part contains five convolutional modules, and each module includes two convolutional layers and a downsampling layer. The shared encoding part extracts feature information from intrapartum ultrasound images. The image classification part determines whether the ultrasound image belongs to a standard plane image. The image segmentation part extracts the fetal head and symphysis pubis regions. The feature point localization part locates the two endpoints of the symphysis pubis in the form of a Gaussian heatmap.

[0063] The image classification part contains two residual connection convolutional blocks. Each residual connection convolutional block contains a convolutional layer, a regularization layer, and an activation layer, followed by an average pooling operation, and finally a fully connected layer and an activation layer. The classification feature map has two channels, respectively indicating whether it belongs to the standard plane.

[0064] The image segmentation part contains 4 upsampling convolutional modules and an activation layer. Each upsampling convolutional module contains an upsampling layer and two convolutional layers. Here, bilinear sampling is used to enlarge the image size and halve the number of image channels. Subsequently, skip connections are used to splice the feature maps with the same resolution in the contraction stage and the feature maps in the current expansion stage.

[0065] The feature point localization part also adopts a similar structure. In addition, Gaussian heatmaps are generated for the coordinates of the two endpoints of the symphysis pubis respectively, and they are spliced together to form a two-channel Gaussian heatmap as the label of the endpoint localization part.

[0066] Step S22: Use the backpropagation algorithm to iteratively optimize the multi-branch convolutional neural network MT-UNet according to the dataset. The adam optimizer is used to solve the optimal network method, and the goal is to minimize the loss function.

[0067] Step S23: Use the multi-branch convolutional neural network MT-UNet optimized by iteration to perform standard plane recognition, image segmentation, and feature point localization on ultrasound images. The input is an intrapartum ultrasound image, and the output is the classification of the image and the segmentation of the fetal head and symphysis pubis regions.

[0068] Step S3: Perform elliptical fitting on the contour of the segmented fetal head region to obtain the fitted fetal head elliptical equation.

[0069] In this embodiment, the step S3 specifically includes the following steps:

[0070] Step S31: Screen the fetal head and pubic symphysis region, and retain the connected domain with the largest pixel area through connected component processing;

[0071] Step S32: Use the Sobel edge detection algorithm to extract the contour of the fetal head region. The input is the segmentation result map of MT-UNet, and the output is the contour point set of the fetal head region.

[0072] Step S33: Use the least squares method to perform elliptical fitting on the contour of the fetal head region. The input is the contour point set of the fetal head region, and the output is the fitted elliptical equation.

[0073] Step S4: Calculate the labor progress angle and the fetal head-pubic symphysis distance based on the fetal head elliptical equation.

[0074] As Figure 2 shown, according to the definition of the labor progress angle (AoP), find the tangent point where the right endpoint of the pubic symphysis is tangent to the fetal head, connect the left and right endpoints of the pubic symphysis and the tangent point, and calculate the labor progress angle (AoP) according to the trigonometric formula; according to the definition of the fetal head-pubic symphysis distance (HSD), connect the left and right endpoints of the pubic symphysis, draw a vertical line through the right endpoint, find the intersection point with the elliptical fit of the fetal head, and calculate the Euclidean distance from the right endpoint of the pubic symphysis to the intersection point to obtain the fetal head-pubic symphysis distance (HSD).

[0075] This application is implemented based on deep learning methods and image processing technologies, and can quickly and accurately select standard planes with clear fetal head and pubic symphysis structures and perform segmentation, and calculate the labor progress angle and the fetal head-pubic symphysis distance.

[0076] In Figure 6 it, the Bland-Altman statistical analysis chart was used to evaluate the consistency of the measurement results of AoP and HSD. The limits of consistency were defined as the mean ± 1.96 standard deviations. Among them, the consistency analysis result of AoP was: the 95% consistency range was from -7.9° to 7.1°, and 4.49% of the points were outside the consistency limits. The consistency analysis result of HSD was: the 95% consistency range was from -7.2 mm to 2.8 mm, and 4.28% of the points were outside the consistency limits. More than 95% of the data points of these two indicators were within the consistency range, and the significance P values were all less than 0.001, indicating that the automatic measurement methods of AoP and HSD proposed in this chapter were consistent with the manual measurement results. The differences remained relatively constant, indicating that there was a systematic error rather than a proportional error between these two measurement values, and there was no significant difference between the measurement results of the method disclosed in this embodiment and the measurement results of clinicians.

[0077] It should be understood that although the steps in the flowchart of the accompanying drawings are shown sequentially in the direction of the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless specifically stated herein, there is no strict order restriction for the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0078] Further reference is made to Figure 7 , as an implementation of the method shown above Figure 2 , this application provides an embodiment of a standard plane recognition and labor process parameter measurement device based on a multi-task network. This device embodiment corresponds to Figure 2 the method embodiment shown, and this device can be specifically applied to various electronic devices.

[0079] As shown in Figure 7 , the standard plane recognition and labor process parameter measurement device 300 based on a multi-task network described in this embodiment includes: an acquisition module 701, a classification and segmentation module 702, a fitting module 703, and a calculation module 704. Among them:

[0080] The acquisition module 701 is used to acquire transperineal ultrasound images;

[0081] The classification and segmentation module 702 is used to classify, using a pre-set multi-branch convolutional neural network MT-UNet, a standard plane image with clear fetal head and symphysis pubis structures from the transperineal ultrasound image, and segment the fetal head and symphysis pubis regions on the standard plane image;

[0082] The fitting module 703 is used to perform elliptical fitting on the contour of the segmented fetal head region to obtain a fitted fetal head ellipse equation;

[0083] The calculation module 704 is used to calculate the labor progress angle and the fetal head-symphysis pubis distance based on the fetal head ellipse equation.

[0084] This application designs a multi-task deep learning network that integrates standard plane recognition and labor process parameter measurement, and uses image classification, image segmentation, and feature point localization to automatically identify the standard plane and achieve automatic measurement of the labor progress angle and the fetal head-pubic symphysis distance. Compared with a single-task single model, sharing a single model for multiple tasks can reduce memory usage and speed up the inference speed. At the same time, multiple related tasks complement each other by sharing information, which can improve the network performance. Experimental results show that the multi-task algorithm of this application can effectively identify the standard plane, segment the fetal head-pubic symphysis region, and locate the endpoints of the pubic symphysis. This application can reduce the burden on doctors, help improve doctor work efficiency, and eliminate the subjectivity drawbacks of different measurement results of the labor progress angle and the fetal head-pubic symphysis distance for the same pregnant woman by different doctors due to different professional levels and experiences.

[0085] To solve the above technical problems, the embodiments of this application also provide a computer device. For details, please refer to Figure 8 , Figure 8 which is the basic structural block diagram of the computer device in this embodiment.

[0086] The computer device 80 includes a memory 81, a processor 82, and a network interface 83 that are communicatively connected to each other through a system bus. It should be noted that only the computer device 80 with components 81-83 is shown in the figure, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be alternatively implemented. Among them, those skilled in the art of this technology can understand that a computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0087] The computer device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer device can interact with users through means such as a keyboard, a mouse, a remote control, a touchpad, or a voice control device.

[0088] The memory 81 includes at least one type of readable storage medium, which includes flash memory, hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 81 may be an internal storage unit of the computer device 80, such as the hard disk or memory of the computer device 80. In other embodiments, the memory 81 may also be an external storage device of the computer device 80, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. equipped on the computer device 80. Of course, the memory 81 may also include both the internal storage unit and the external storage device of the computer device 80. In this embodiment, the memory 81 is generally used to store the operating system and various application software installed on the computer device 80, such as computer-readable instructions for the standard surface recognition and production process parameter measurement method based on a multi-task network. In addition, the memory 81 may also be used to temporarily store various data that have been output or will be output.

[0089] In some embodiments, the processor 82 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 82 is generally used to control the overall operation of the computer device 80. In this embodiment, the processor 82 is used to run the computer-readable instructions stored in the memory 81 or process data, such as running the computer-readable instructions for the standard surface recognition and production process parameter measurement method based on a multi-task network.

[0090] The network interface 83 may include a wireless network interface or a wired network interface, and the network interface 83 is generally used to establish a communication connection between the computer device 80 and other electronic devices.

[0091] This application designs a multi-task deep learning network that integrates standard plane recognition and intrapartum parameter measurement, and uses image classification, image segmentation, and feature point localization to automatically identify the standard plane and achieve automatic measurement of the intrapartum progress angle and the fetal head-pubic symphysis distance. Compared with a single-task single model, sharing a single model for multiple tasks can reduce memory usage and speed up the inference speed. At the same time, multiple related tasks complement each other by sharing information, which can improve the network performance. Experimental results show that the multi-task algorithm of this application can effectively identify the standard plane, segment the fetal head-pubic symphysis region, and locate the pubic symphysis endpoints. This application can reduce the burden on doctors, help improve doctors' work efficiency, and eliminate the subjectivity drawbacks of different measurement results of the intrapartum progress angle and the fetal head-pubic symphysis distance of the same pregnant woman by different doctors due to different professional levels and experiences.

[0092] This application also provides another implementation manner, that is, to provide a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor, so that the at least one processor executes the steps of the standard plane recognition and intrapartum parameter measurement method based on the multi-task network as described above.

[0093] This application designs a multi-task deep learning network that integrates standard plane recognition and intrapartum parameter measurement, and uses image classification, image segmentation, and feature point localization to automatically identify the standard plane and achieve automatic measurement of the intrapartum progress angle and the fetal head-pubic symphysis distance. Compared with a single-task single model, sharing a single model for multiple tasks can reduce memory usage and speed up the inference speed. At the same time, multiple related tasks complement each other by sharing information, which can improve the network performance. Experimental results show that the multi-task algorithm of this application can effectively identify the standard plane, segment the fetal head-pubic symphysis region, and locate the pubic symphysis endpoints. This application can reduce the burden on doctors, help improve doctors' work efficiency, and eliminate the subjectivity drawbacks of different measurement results of the intrapartum progress angle and the fetal head-pubic symphysis distance of the same pregnant woman by different doctors due to different professional levels and experiences.

[0094] Through the description of the above implementation manners, those skilled in the art can clearly understand that the above-described example methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in various embodiments of this application.

[0095] Obviously, the embodiments described above are only a part of the embodiments of this application, rather than all of them. The preferred embodiments of this application are given in the accompanying drawings, but they do not limit the patent scope of this application. This application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosed content of this application more thorough and comprehensive. Although this application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions described in the foregoing specific embodiments, or perform equivalent replacements on some of the technical features. Any equivalent structure made by using the content of this application's specification and drawings, directly or indirectly applied in other related technical fields, is equally within the scope of the patent protection of this application.

Claims

1. A method for standard face recognition and labor parameter measurement based on a multi-task network, characterized in that: The steps include: S1, collect transperineal ultrasound images; S2, using a preset multi-branch convolutional neural network MT-UNet to classify a standard surface image with clear fetal head and pubic symphysis structures in the transperineal ultrasound image, and segmenting the fetal head and pubic symphysis regions on the standard surface image; S3, performing ellipse fitting on the segmented fetal head region contour to obtain a fitted fetal head ellipse equation; S4. Calculate the labor progress angle and the fetal head-pubic symphysis distance based on the fetal head ellipse equation.

2. The method for standard face recognition and labor parameter measurement based on multi-task network according to claim 1, characterized in that: The step S1 also includes: dividing the transperineal ultrasound image into two data sets: iterative optimization data set and measurement data set.

3. The method for standard face recognition and labor parameter measurement based on multi-task network according to claim 1, characterized in that: The step S2 specifically includes the following steps: S21. Design a multi-branch convolutional neural network MT-UNet. The network MT-UNet has a shared encoding part and a decoding part. The decoding part includes an image classification part, an image segmentation part and a feature point positioning part. The shared encoding part extracts feature information from the intrapartum ultrasound image. The image classification part determines whether the ultrasound image belongs to a standard surface image. The image segmentation part extracts the fetal head and pubic symphysis area. The feature point positioning part uses a Gaussian heat map to locate the two endpoints of the pubic symphysis. S22. Use the back propagation algorithm to iteratively optimize the multi-branch convolutional neural network MT-UNet according to the data set; S23. Use the iteratively optimized multi-branch convolutional neural network MT-UNet to perform standard face recognition and image segmentation on ultrasound images.

4. The method for standard face recognition and labor parameter measurement based on multi-task network according to claim 3 is characterized in that: The shared coding part in step S21 includes five convolution modules, each of which includes two convolution layers and one downsampling layer.

5. The method for standard face recognition and labor parameter measurement based on multi-task network according to claim 3 is characterized in that: The image classification part in step S21 includes two residual connection convolution blocks, each residual connection convolution block includes a convolution layer, a regularization layer and an activation layer, the two residual connection convolution blocks are connected to an average pooling operation, and the average pooling is further connected to a fully connected layer and an activation layer. The classification feature map has two channels, respectively indicating whether it belongs to a standard surface.

6. The method for standard face recognition and labor parameter measurement based on multi-task network according to claim 3 is characterized in that: The image segmentation part in step S21 includes four upsampling convolution modules and activation layers. Each upsampling convolution module includes an upsampling layer and two convolution layers. Bilinear sampling is used here to enlarge the image size and halve the number of image channels. Then, a jump connection is used to splice the feature map of equal resolution in the contraction stage with the feature map of the current expansion stage.

7. The method for standard face recognition and labor parameter measurement based on multi-task network according to claim 1, characterized in that: The step S3 specifically comprises the following steps: S31, screening the fetal head and pubic symphysis regions, and retaining the connected domain with the largest pixel area through connected domain processing; S32, extracting the contour of the fetal head region using a Sobel edge detection algorithm; S33. Use the least square method to perform ellipse fitting on the contour of the fetal head area.

8. A standard face recognition and labor parameter measurement device based on a multi-task network, characterized in that: include: An acquisition module, used for acquiring transperineal ultrasound images; A classification and segmentation module, used to classify the transperineal ultrasound image into a standard surface image with clear fetal head and pubic symphysis structures using a preset multi-branch convolutional neural network MT-UNet, and segment the fetal head and pubic symphysis regions on the standard surface image; A fitting module is used to perform ellipse fitting on the segmented fetal head area contour to obtain a fitted fetal head ellipse equation; A calculation module is used to calculate the labor progress angle and the fetal head-pubic symphysis distance based on the fetal head ellipse equation.

9. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the standard face recognition and labor parameter measurement method based on a multi-task network as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the method for standard face recognition and labor parameter measurement based on a multi-task network as described in any one of claims 1 to 7.

Images