An ultrasonic examination method, apparatus, and electronic device and storage medium

Abstract

The application discloses an ultrasonic examination method and device, an electronic device and a computer readable storage medium. The method comprises: acquiring cross-section images of different directions by using a two-dimensional ultrasonic probe, and determining a probe pose corresponding to the cross-section images; constructing three-dimensional body data based on at least one cross-section image and the corresponding probe pose, and segmenting the three-dimensional body data by using a three-dimensional semantic segmentation model to obtain three-dimensional masks corresponding to a plurality of target object instances; extracting a two-dimensional contour of a target object in at least one cross-section image by using a two-dimensional semantic segmentation model; and determining a target object instance matched with the two-dimensional contour based on the three-dimensional masks. The application realizes consistent scanning of a target object in two dimensions and three dimensions by using only a two-dimensional ultrasonic probe, and further realizes consistent measurement of the target object in two dimensions and three dimensions, thereby reducing the requirement of ultrasonic examination on an ultrasonic device.

Description

Technical Field

[0001] This application relates to the field of ultrasound examination technology, and more specifically, to an ultrasound examination method, apparatus, electronic device, and computer-readable storage medium. Background Technology

[0002] Two-dimensional ultrasound (2D) technology, due to its ease of operation and real-time display capabilities, is particularly suitable for observing dynamic processes in organs and tissues, such as heartbeats and fetal movements. However, it is limited by providing only single-view information, making it difficult to comprehensively determine the spatial location and size of lesions. In contrast, 3D ultrasound technology provides a more comprehensive anatomical view, aiding in the precise location and measurement of lesions. However, it is limited by high cost, lower resolution, and processing delays, making it unsuitable for real-time applications. In clinical practice, doctors typically first use 2D ultrasound to observe the target object, relying on experience and AI assistance for measurement, and then decide whether to use 3D ultrasound as needed. 3D ultrasound acquires stereoscopic images through a volumetric probe and also requires AI assistance or manual measurement. However, the types of ultrasound probes used for 2D and 3D ultrasound are often different, and they are scanned separately using independent scanning methods. The two types of ultrasound probes and independent scanning methods can lead to inconsistent scanning results.

[0003] Therefore, how to achieve consistent scanning of the target object in two dimensions and three dimensions, and reduce the requirements of ultrasound equipment for ultrasound examination, is a technical problem that needs to be solved by those skilled in the art. Summary of the Invention

[0004] The purpose of this application is to provide an ultrasound examination method, apparatus, electronic device, and computer-readable storage medium, which achieves consistent scanning of the target object in two dimensions and three dimensions using only a two-dimensional ultrasound probe, thereby reducing the requirements of ultrasound equipment for ultrasound examination.

[0005] To achieve the above objectives, this application provides an ultrasound examination method, comprising:

[0006] A two-dimensional ultrasound probe is used to acquire cross-sectional images from different orientations, and the probe pose corresponding to the cross-sectional images is determined.

[0007] Three-dimensional volume data is constructed based on at least one of the cross-sectional images and the corresponding probe pose, and the three-dimensional volume data is segmented using a three-dimensional semantic segmentation model to obtain three-dimensional masks corresponding to multiple target object instances.

[0008] A two-dimensional semantic segmentation model is used to extract the two-dimensional contour of the target object from at least one of the cross-sectional images;

[0009] The target object instance that matches the two-dimensional contour is determined based on the three-dimensional mask.

[0010] The method of acquiring cross-sectional images from different orientations using a two-dimensional ultrasound probe further includes:

[0011] Identify the key cross-section image from among the multiple cross-section images;

[0012] Accordingly, the construction of three-dimensional volume data based on at least one of the cross-sectional images and the corresponding probe pose includes:

[0013] At least one of the cross-sectional images and the corresponding probe pose are input into a 3D reconstruction network for training to obtain 3D volume data; wherein, at least one of the cross-sectional images includes the key cross-sectional image, and during the training of the 3D reconstruction network, the weight of the key cross-sectional image is greater than the weight of the non-key cross-sectional image.

[0014] Accordingly, the extraction of the two-dimensional contour of the target object from at least one of the cross-sectional images using a two-dimensional semantic segmentation model includes:

[0015] Two-dimensional semantic segmentation models are used to extract the two-dimensional contours of the target object from the key section image.

[0016] Specifically, at least one of the cross-sectional images and the corresponding probe pose are input into a 3D reconstruction network for training to obtain 3D volume data, including:

[0017] At least one of the cross-sectional images and the corresponding probe poses are input into an explicit 3D reconstruction network for training, so that the explicit 3D reconstruction network can directly output 3D volume data.

[0018] Specifically, at least one of the cross-sectional images and the corresponding probe pose are input into a 3D reconstruction network for training to obtain 3D volume data, including:

[0019] At least one of the cross-sectional images and the corresponding probe poses are input into an implicit 3D reconstruction network for training, so that the implicit 3D reconstruction network outputs an implicit representation of the 3D volume;

[0020] Multiple viewpoints are selected within a preset scanning viewpoint range, and multiple sampling points in three-dimensional space are determined for each viewpoint. The attributes of each sampling point are determined based on the implicit representation of the three-dimensional volume. If the attributes of the target sampling point do not exist in the implicit representation of the three-dimensional volume, the attributes of the adjacent sampling points of the target sampling point in the implicit representation of the three-dimensional volume are interpolated to obtain the attributes of the target sampling point.

[0021] The images corresponding to each viewpoint are determined based on the attributes of each sampling point corresponding to each viewpoint, and the images corresponding to each viewpoint are merged to obtain the three-dimensional volume data.

[0022] The target objects include at least one of the following: ovaries and follicles.

[0023] Wherein, if the target object includes ovaries and follicles, then determining the key section image among the multiple section images includes:

[0024] The number of follicles contained in each of the cross-sectional images is counted, and the cross-sectional images containing more than a preset value are determined as key cross-sectional images.

[0025] In the 3D mask corresponding to the kth target object instance, the mask value of the voxel point belonging to the kth target object instance is k, and the mask value of the voxel point not belonging to the kth target object instance is 0, 1≤k≤K, and K is the number of target object instances.

[0026] Accordingly, determining the target object instance matching the two-dimensional contour based on the three-dimensional mask includes:

[0027] Based on the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located, the two-dimensional contour is mapped to three-dimensional space to obtain the three-dimensional contour corresponding to the two-dimensional contour.

[0028] The center point coordinates of the three-dimensional contour are determined, and the corresponding mask value is queried in the three-dimensional mask corresponding to K target object instances using the center point coordinates as an index. The target object instance corresponding to the non-zero mask value is determined as the target object instance that matches the two-dimensional contour.

[0029] The step of mapping the two-dimensional contour to three-dimensional space based on the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located to obtain the three-dimensional contour corresponding to the two-dimensional contour includes:

[0030] In the two-dimensional contour, a plurality of contour points and the two-dimensional coordinates of the plurality of contour points are determined;

[0031] To obtain the ultrasound image scale, for any contour point: calculate the product of the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located, the ultrasound image scale, and the two-dimensional coordinates of the contour point, and use it as the three-dimensional coordinates of the contour point.

[0032] The three-dimensional contour corresponding to the two-dimensional contour in three-dimensional space is determined based on the three-dimensional coordinates of multiple contour points.

[0033] After determining the target object instance matching the two-dimensional contour based on the three-dimensional mask, the method further includes:

[0034] The measurement result of the target object is determined based on a two-dimensional contour in at least one of the cross-sectional images and a target object instance that matches the two-dimensional contour.

[0035] To achieve the above objectives, this application provides an ultrasound examination device, comprising:

[0036] The acquisition module is used to acquire cross-sectional images from different orientations using a two-dimensional ultrasound probe and determine the probe pose corresponding to the cross-sectional images.

[0037] A construction module is used to construct three-dimensional volume data based on at least one of the cross-sectional images and the corresponding probe poses, and to segment the three-dimensional volume data using a three-dimensional semantic segmentation model to obtain three-dimensional masks corresponding to multiple target object instances;

[0038] The extraction module is used to extract the two-dimensional contour of the target object from at least one of the cross-sectional images using a two-dimensional semantic segmentation model;

[0039] The matching module is used to determine the target object instance that matches the two-dimensional contour based on the three-dimensional mask.

[0040] To achieve the above objectives, this application provides an electronic device, comprising:

[0041] Memory, used to store computer programs;

[0042] A processor is used to execute the computer program to implement the steps of the ultrasound examination method described above.

[0043] To achieve the above objectives, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the ultrasound examination method described above.

[0044] The ultrasonic examination method provided in this application constructs three-dimensional volumetric data of the target object by combining two-dimensional cross-sectional images and probe pose information. Simultaneously, by matching two-dimensional contours with three-dimensional masks—that is, determining the target object instance corresponding to each two-dimensional contour in the three-dimensional volumetric data—the consistency of the two-dimensional and three-dimensional scanning results for the target object can be ensured. Therefore, this application achieves two-dimensional and three-dimensional scanning of the target object using only a two-dimensional ultrasonic probe, reducing the requirements for ultrasonic equipment in ultrasonic examinations. This application also discloses an ultrasonic examination device, an electronic device, and a computer-readable storage medium, which can achieve the same technical effects.

[0045] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. The drawings are used to provide a further understanding of this disclosure and constitute a part of the specification. They are used together with the following detailed description to explain this disclosure, but do not constitute a limitation of this disclosure. In the drawings:

[0047] Figure 1 This is a flowchart illustrating an ultrasound examination method according to an exemplary embodiment;

[0048] Figure 2 A flowchart illustrating another ultrasound examination method according to an exemplary embodiment;

[0049] Figure 3 This is a flowchart illustrating a combined two-dimensional and three-dimensional ovarian follicle measurement method according to an exemplary embodiment;

[0050] Figure 4 This is a structural diagram of an ultrasound examination device according to an exemplary embodiment;

[0051] Figure 5 This is a structural diagram of an electronic device according to an exemplary embodiment. Detailed Implementation

[0052] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. Furthermore, in the embodiments of this application, "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0053] This application discloses an ultrasound inspection method that achieves consistent scanning of the target object in two dimensions and three dimensions using only a two-dimensional ultrasound probe, thereby reducing the requirements of ultrasound equipment for ultrasound inspection.

[0054] See Figure 1 A flowchart illustrating an ultrasound examination method according to an exemplary embodiment, such as... Figure 1 As shown, it includes:

[0055] S101: Use a two-dimensional ultrasonic probe to acquire cross-sectional images from different orientations and determine the probe pose corresponding to the cross-sectional images.

[0056] In practice, a two-dimensional ultrasound probe is moved across the examination site to acquire a sequence of cross-sectional images from different angles, thus displaying a two-dimensional view of the examination site. After acquiring the cross-sectional image sequence, at least a portion of the acquired cross-sectional images can be used as the basis for subsequent processing. Furthermore, some cross-sectional image frames can be extracted as the basis for subsequent construction of three-dimensional volume data. Sequence frame extraction can reduce the running time of subsequent three-dimensional reconstruction algorithms. Sequence frame extraction can be achieved through methods such as equidistant frame extraction.

[0057] Furthermore, after determining the cross-sectional images used to construct the subsequent 3D volume data, the probe pose corresponding to each cross-sectional image is determined. The probe pose records the spatial position and orientation of the probe when acquiring each cross-sectional image. The probe pose can be directly determined using the inertial measurement unit within the probe.

[0058] The sensors in the inertial measurement unit can directly measure the angular velocity of the probe during its movement. And linear acceleration a. Pose at the starting frame (the section image initially acquired). As a reference coordinate system for the pose of subsequent cross-sectional images:

[0059] ;

[0060] Probe pose in the i-th frame slice image Including rotation matrix Translation vector , Used to describe the rotation information of the i-th frame slice image relative to the starting frame. This is used to describe the translation information of the i-th frame slice image relative to the starting frame. Therefore, the probe pose of the i-th frame slice image... It can be represented as:

[0061] ;

[0062] in, , , i and j are the frame numbers of the cross-sectional image. Let be the rotation matrix in the pose of the i-th frame slice image, used to describe the rotation information of the i-th frame slice image relative to the starting frame. Let be the rotation matrix in the pose of the (j-1)th frame slice image. The time interval between adjacent cross-sectional images. Let be the translation vector in the pose of the i-th frame slice image, used to describe the translation information of the i-th frame slice image relative to the starting frame. Let be the translation vector in the pose of the (j-1)th frame slice image. Let be the velocity vector of the probe corresponding to the (j-1)th frame of the cross-sectional image. Let be the linear acceleration of the probe corresponding to the j-th frame cross-sectional image.

[0063] S102: Construct three-dimensional volume data based on at least one of the cross-sectional images and the corresponding probe poses, and use a three-dimensional semantic segmentation model to segment the three-dimensional volume data to obtain three-dimensional masks corresponding to multiple target object instances.

[0064] It should be noted that, in this embodiment, the target object refers to the main object that needs to be detected, analyzed, or diagnosed during an ultrasound examination. It is usually a holistic concept used to describe the main object or area being examined. For example, in a medical ultrasound examination, the target object might be a fetus, an organ (such as the heart, liver, or ovary), or tissue. It emphasizes the subject and scope of the examination, without involving specific individual differences or the distinction between multiple objects. A target object instance, on the other hand, is the concretization and individualization of the target object in a specific scenario. It refers to a single entity with independent features and boundaries that is identified and segmented by a semantic segmentation model in three-dimensional volume data. For example, if the target object is a "follicle," then there may be multiple "instances" of follicles in the three-dimensional volume data; if the target object is a "fetus," then in an ultrasound examination of twins or multiple pregnancies, each fetus is a "target object instance."

[0065] In this step, images from different sections are spatially aligned based on the probe pose information to ensure their relative positions in three-dimensional space are accurate. Subsequently, image fusion techniques, such as voxel reconstruction or surface reconstruction, are used to merge these aligned two-dimensional images into a continuous three-dimensional model, i.e., three-dimensional volume data. The three-dimensional volume data contains all tissues detected during probe scanning, such as the ovaries, fallopian tubes, blood vessels, and uterus. Continuous three-dimensional scene representations can be obtained from the three-dimensional volume data; that is, inputting any scanning viewpoint yields the corresponding ultrasound image.

[0066] Furthermore, a 3D semantic segmentation model is used to analyze the 3D volume data. This model can identify and distinguish different tissues and structures, generating 3D masks corresponding to each target object instance. Different target object instances can correspond to different mask values. In the 3D mask corresponding to the k-th target object instance, the mask value of the voxels belonging to the k-th target object instance is k, and the mask value of the voxels not belonging to the k-th target object instance is 0, where 1≤k≤K, and K is the number of target object instances.

[0067] In one specific embodiment, a 3D semantic segmentation model is used to analyze the 3D volume data. Voxels belonging to the target object are marked as 1, and voxels not belonging to the target object are marked as 0. Then, connected component analysis is performed on the voxels marked as 1, which are the voxels corresponding to the target object, to obtain the corresponding 3D masks for multiple target object instances. Different target object instances correspond to different mask values. In the 3D mask corresponding to the k-th target object instance, the mask value of the voxels belonging to the k-th target object instance is k. That is, in the process of connected component analysis of the voxels corresponding to the target object, each connected component analyzed yields a corresponding target object instance, and a mask value is marked for that target object instance. For example, the mask value of the voxels belonging to the first target object instance is marked as 1, the mask value of the voxels belonging to the second target object instance is marked as 2, and so on, until the mask value of the voxels belonging to the K-th target object instance is marked as K.

[0068] In a preferred embodiment, after acquiring cross-sectional images from different orientations using a two-dimensional ultrasound probe, the method further includes: determining key cross-sectional images among the multiple cross-sectional images; correspondingly, constructing three-dimensional volume data based on at least one cross-sectional image and the corresponding probe pose includes: inputting at least one cross-sectional image and the corresponding probe pose into a three-dimensional reconstruction network for training to obtain three-dimensional volume data; wherein, at least one cross-sectional image contains the key cross-sectional image, and during the training process of the three-dimensional reconstruction network, the weight of the key cross-sectional image is greater than the weight of the non-key cross-sectional images.

[0069] Key section images refer to images containing important information selected from multiple section images through image feature analysis or manual selection by the physician. For example, these images may contain more details of the target object (such as follicles) or clearer tissue structures. In practice, the system filters key section images from multiple sections and assigns them higher weight in subsequent 3D reconstruction, thereby ensuring the accuracy and reliability of the reconstruction results.

[0070] In practice, after acquiring 2D cross-sectional images, the system identifies key cross-sectional images and inputs them, along with other cross-sectional images and their corresponding probe poses, into a 3D reconstruction network for training. The 3D reconstruction network maps the 2D cross-sectional images and probe poses into representations in 3D space. During training, the weights of key cross-sectional images are set higher than those of non-key cross-sectional images, allowing the network to focus more on these important images containing crucial information, thereby improving the accuracy and effectiveness of 3D reconstruction. After training, the 3D reconstruction network can generate high-quality representations of 3D volumes that reflect the geometric structure of the target object.

[0071] For the selection of the 3D reconstruction network, either an explicit 3D reconstruction network or an implicit 3D reconstruction network can be chosen. As a feasible implementation, at least one of the cross-sectional images and the corresponding probe poses are input into the 3D reconstruction network for training to obtain 3D volume data. This includes: inputting at least one of the cross-sectional images and the corresponding probe poses into an explicit 3D reconstruction network for training, so that the explicit 3D reconstruction network directly outputs 3D volume data, for example, in the form of voxels or meshes. The advantage of this method is that the output results are intuitive, facilitating subsequent processing and visualization.

[0072] As another feasible implementation, at least one of the cross-sectional images and the corresponding probe poses are input into a 3D reconstruction network for training to obtain 3D volume data. This includes: inputting at least one of the cross-sectional images and the corresponding probe poses into an implicit 3D reconstruction network for training, so that the implicit 3D reconstruction network outputs an implicit representation of the 3D volume; selecting multiple viewpoints within a preset scanning viewpoint range, determining multiple sampling points in 3D space for each viewpoint, and determining the attributes of each sampling point based on the implicit representation of the 3D volume; wherein, if the implicit representation of the 3D volume does not contain the attribute of the target sampling point, then interpolating the attributes of the adjacent sampling points of the target sampling point in the implicit representation of the 3D volume to obtain the attribute of the target sampling point; determining the image corresponding to each viewpoint according to the attributes of each sampling point corresponding to each viewpoint, and merging the images corresponding to each viewpoint to obtain 3D volume data.

[0073] The implicit 3D reconstruction network outputs an implicit representation of the 3D volume, mapping points in space to attributes such as density or color. This implicit representation not only accurately reflects the geometric structure of the target object but also captures complex details and texture information. Then, based on this implicit representation, 3D volume data can be generated by selecting multiple viewpoints within a preset scanning range and determining the sampling points for these viewpoints in 3D space. If an attribute of a sampling point exists in the implicit representation, it can be directly obtained. If an attribute of a sampling point does not exist in the implicit representation, it is obtained by interpolating the attributes of adjacent sampling points. For each viewpoint, the corresponding image is determined based on the attributes of its various sampling points. Finally, by merging the images corresponding to each viewpoint, complete 3D volume data is generated.

[0074] S103: Extract the two-dimensional contour of the target object from at least one of the cross-sectional images using a two-dimensional semantic segmentation model.

[0075] In practice, by applying a two-dimensional semantic segmentation model to each cross-sectional image, the two-dimensional contours of the target object can be identified and extracted. These contours provide the precise boundaries of the target object in each cross-sectional image, which helps to understand the shape and size of the target object.

[0076] As a preferred embodiment, the step of extracting the two-dimensional contour of the target object in at least one of the cross-sectional images using a two-dimensional semantic segmentation model includes: extracting the two-dimensional contour of the target object in the key cross-sectional image using a two-dimensional semantic segmentation model.

[0077] In practical implementation, when using a 2D semantic segmentation model to extract the 2D contour of a target object from cross-sectional images, special attention is paid to key cross-sectional images. This means that during image analysis and contour extraction, the model will prioritize processing those key images identified as containing important information or features. Because key cross-sectional images provide richer and clearer structural information, the contour of the target object can be more accurately identified and extracted from them, improving the efficiency and accuracy of contour extraction.

[0078] S104: Determine the target object instance that matches the two-dimensional contour based on the three-dimensional mask.

[0079] In this step, the two-dimensional contours are matched with the three-dimensional mask to determine the target object instance corresponding to each two-dimensional contour, ensuring that the two-dimensional contours extracted from different cross-sectional images are consistent with the target object instances in the three-dimensional model.

[0080] Optionally, spatial transformation based on the scale of the ultrasound image can transform the two-dimensional contour into three-dimensional space, obtaining the corresponding three-dimensional contour, denoted as contour-1. Using the probe pose corresponding to the keyframe (i.e., the key section image), the segmentation results on the corresponding section are obtained from the three-dimensional mask, yielding the three-dimensional contours corresponding to multiple target object instances, denoted as contour-2. Matching contour-1 and contour-2 in three-dimensional space yields a unique matching instance of each target object in the keyframe within the three-dimensional space. This matching of contour-1 and contour-2 in three-dimensional space can be achieved based on the feature coordinates of the three-dimensional contours, such as the center point coordinates.

[0081] As a feasible implementation, determining the target object instance matching the two-dimensional contour based on the three-dimensional mask includes: mapping the two-dimensional contour into three-dimensional space based on the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located to obtain the three-dimensional contour corresponding to the two-dimensional contour; determining the center point coordinates of the three-dimensional contour, and using the center point coordinates as an index to query the corresponding mask value in the three-dimensional masks corresponding to N target object instances, and determining the target object instance corresponding to the non-zero mask value as the target object instance matching the two-dimensional contour. In the three-dimensional mask corresponding to the k-th target object instance, the mask value of the voxel point belonging to the k-th target object instance is k, and the mask value of the voxel point not belonging to the k-th target object instance is 0, 1≤k≤K, where K is the number of target object instances.

[0082] In practice, based on the probe pose information corresponding to the cross-sectional image of the 2D contour, the 2D contour is mapped into 3D space to obtain the corresponding 3D contour. Then, the coordinates of the center point of this 3D contour are calculated, and this coordinate is used as an index to look up the corresponding mask value in the 3D mask. If the found mask value is not zero, the target object instance corresponding to that mask value is the instance that matches the 2D contour.

[0083] As a feasible implementation, the step of mapping the two-dimensional contour to a three-dimensional space based on the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located to obtain the three-dimensional contour corresponding to the two-dimensional contour includes: determining multiple contour points and the two-dimensional coordinates of the multiple contour points in the two-dimensional contour; obtaining an ultrasound image scale; for any contour point: calculating the product between the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located, the ultrasound image scale, and the two-dimensional coordinates of the contour point, as the three-dimensional coordinates of the contour point; and determining the three-dimensional contour corresponding to the two-dimensional contour in the three-dimensional space based on the three-dimensional coordinates of the multiple contour points.

[0084] In practice, firstly, multiple contour points and their two-dimensional coordinates are extracted from the two-dimensional contour. These contour points are key locations on the two-dimensional contour and are used for subsequent three-dimensional mapping. Next, the scale of the ultrasound image is obtained; this scale is a fixed parameter of the ultrasound equipment and is used to convert the pixel coordinates in the image into actual spatial dimensions. For each contour point, its three-dimensional coordinates are calculated. Specifically, the probe pose, the ultrasound image scale, and the two-dimensional coordinates of the contour point are multiplied to obtain the coordinates of that contour point in three-dimensional space. Finally, based on the three-dimensional coordinates of all contour points, the three-dimensional contour corresponding to the two-dimensional contour in three-dimensional space is constructed.

[0085] Taking medical ultrasound examination as an example, assuming the target object is a follicle, a cross-sectional image of the ovary is acquired using a two-dimensional ultrasound probe, and a three-dimensional semantic segmentation model is used to generate three-dimensional masks for K follicles, with mask values ​​of 1, 2, ..., K. When it is necessary to determine the follicle corresponding to a certain two-dimensional contour, the above matching method can accurately match the two-dimensional contour with the follicle instance in the three-dimensional mask, thereby achieving consistency analysis of two-dimensional and three-dimensional information.

[0086] In a preferred embodiment, after determining the target object instance matching the two-dimensional contour based on the three-dimensional mask, the method further includes: determining the measurement result of the target object based on the two-dimensional contour in at least one cross-sectional image and the target object instance matching the two-dimensional contour.

[0087] In practical implementation, after matching the target object instance in the 2D contour with the 3D mask, the measurement result of the target object is further determined by utilizing the 2D contour information and the matched 3D target object instance. This allows for consistent 2D and 3D measurement of the target object. The 2D contour provides the boundary information of the target object on the 2D plane, while the matched 3D target object instance provides information such as the target object's position, shape, and volume in 3D space. By integrating 2D and 3D data, the characteristics of the target object can be obtained more comprehensively, thus achieving more accurate measurement. For example, in medical ultrasound examination, the 2D contour can be used to measure the 2D dimensions of the target object, while the 3D target object instance can be used to calculate its volume or other 3D parameters. This measurement method, which combines 2D and 3D information, not only improves the accuracy and reliability of the measurement but also provides richer data support for clinical diagnosis.

[0088] The ultrasound examination method provided in this application constructs three-dimensional volumetric data of the target object by combining two-dimensional cross-sectional images and probe pose information. Simultaneously, by matching the two-dimensional contours with three-dimensional masks—that is, determining the target object instance corresponding to each two-dimensional contour in the three-dimensional volumetric data—the consistency of the two-dimensional and three-dimensional scanning results for the target object can be ensured. Therefore, this application achieves two-dimensional and three-dimensional scanning of the target object using only a two-dimensional ultrasound probe, reducing the requirements for ultrasound equipment in ultrasound examinations.

[0089] This application discloses an ultrasound examination method. Compared with the previous embodiment, this embodiment further explains and optimizes the technical solution. Specifically:

[0090] See Figure 2 A flowchart illustrating another ultrasound examination method according to an exemplary embodiment, such as... Figure 2 As shown, it includes:

[0091] S201: Use a two-dimensional ultrasonic probe to acquire cross-sectional images from different orientations and determine the probe pose corresponding to the cross-sectional images.

[0092] In this embodiment, a two-dimensional ultrasound probe scans the ovarian region to obtain cross-sectional images of the ovary from different angles, and simultaneously determines the probe pose corresponding to the cross-sectional images. The ovary may contain follicles, and information such as the number and size of follicles can be obtained from the ovarian cross-sectional images.

[0093] S202: Determine the key cross-section image among the plurality of cross-section images.

[0094] In this step, the determination of key section images can be achieved either through automated image feature analysis or manually by the physician. Specifically, a "keyframe button" can be configured in the ultrasound examination interface, allowing the physician to choose whether the current frame should be selected as a key section image based on their experience and judgment. If automated image feature analysis is used, multiple acquired section images are analyzed, and those whose features meet preset conditions are determined as key section images.

[0095] As a possible implementation method, determining the key section image among the multiple section images includes: counting the number of follicles contained in each section image, and determining the section image containing more than a preset value of follicles as the key section image.

[0096] In practice, by counting the number of follicles in each cross-sectional image, images with a higher number of follicles can be identified. When the number of follicles in a cross-sectional image exceeds a preset value, the image is marked as a key cross-sectional image. Key cross-sectional images contain richer follicular information and are of greater value for subsequent 3D reconstruction and follicle monitoring. This step helps improve the accuracy and efficiency of diagnosis, ensuring a comprehensive assessment of follicular development.

[0097] As another feasible implementation, determining the key section image among the plurality of section images includes: identifying the section image containing follicles larger than a preset size as the key section image. In a specific implementation, if a section image contains follicles larger than a preset size, then that section image is identified as the key section image.

[0098] If multiple cross-sectional images meet the aforementioned requirements, all of these cross-sectional images can be used as key cross-sectional images, or they can be selected from among them to determine the key cross-sectional images. For example, the cross-sectional image with the most follicles or the largest follicle size can be used as the key cross-sectional image.

[0099] S203: Input at least one of the cross-sectional images and the corresponding probe pose into an implicit 3D reconstruction network to train and obtain 3D volume data; wherein, at least one of the cross-sectional images includes the key cross-sectional image, and during the training process, the weight of the key cross-sectional image is greater than the weight of the non-key cross-sectional image.

[0100] 3D reconstruction is the process of converting multiple 2D cross-sectional images into a 3D model. Implicit 3D reconstruction networks can utilize cross-sectional images and probe pose information to train and generate 3D volumetric data. During training, key cross-sectional images, containing more follicle information, are given greater weight, meaning they have a greater impact on model training. This weighted strategy ensures that the 3D reconstruction model can more accurately capture the distribution and morphology of follicles, thereby improving the quality and reliability of the reconstruction results. In this way, a 3D model containing detailed structures of the ovary and follicles can be generated.

[0101] S204: Use a three-dimensional ovarian semantic segmentation model to segment the three-dimensional volume data to obtain three-dimensional masks corresponding to M ovarian instances, and use a three-dimensional follicle semantic segmentation model to segment the three-dimensional volume data to obtain three-dimensional masks corresponding to N follicle instances.

[0102] Wherein, the mask value of the voxel point belonging to the m-th ovary instance in the three-dimensional mask corresponding to the m-th ovary instance is m, 1≤m≤M, and the mask value of the voxel point belonging to the n-th follicle instance in the three-dimensional mask corresponding to the n-th follicle instance is n, 1≤n≤N.

[0103] In this step, a semantic segmentation model is used to identify and segment instances of ovaries and follicles. The 3D ovarian semantic segmentation model can identify and segment the boundaries of the ovary, generating a corresponding 3D mask, where each voxel point of an ovarian instance is assigned a unique identifier, i.e., a mask value. Similarly, the 3D follicle semantic segmentation model can identify and segment the boundaries of the follicle, generating a corresponding 3D mask, where each voxel point of a follicle instance is also assigned a unique identifier, i.e., a mask value.

[0104] S205: Use a two-dimensional ovarian semantic segmentation model to extract the two-dimensional contour of the ovary in the key section image, and use a two-dimensional follicle semantic segmentation model to extract the two-dimensional contour of the follicle in the key section image.

[0105] In this step, two-dimensional contours of the ovary and follicles are extracted from the key section images. The two-dimensional ovarian semantic segmentation model can identify the ovarian boundary in the key section image and extract its two-dimensional contour. Similarly, the two-dimensional follicle semantic segmentation model can identify the follicle boundary in the key section image and extract its two-dimensional contour. The m-th ovarian two-dimensional contour is represented as... The two-dimensional contour of the nth follicle is represented as: .

[0106] S206: Based on the probe pose corresponding to the key section image, the two-dimensional contour of the ovary and the two-dimensional contour of the follicle are mapped to three-dimensional space respectively to obtain the three-dimensional contour of the ovary and the three-dimensional contour of the follicle respectively.

[0107] In practice, the two-dimensional contours of the ovary and follicle are mapped into three-dimensional space to generate the three-dimensional contours of the ovary and follicle, respectively. Specifically, the product of the probe pose, ultrasound image scale, and ovarian contour corresponding to the cross-sectional image containing the ovarian contour is calculated as the ovarian three-dimensional contour in three-dimensional space. Similarly, the product of the probe pose, ultrasound image scale, and follicle contour corresponding to the cross-sectional image containing the follicle contour is calculated as the follicle three-dimensional contour in three-dimensional space.

[0108] The specific mapping method is as follows:

[0109] ;

[0110] ;

[0111] in, The m-th ovary has a two-dimensional outline; The m-th ovary's three-dimensional contour; The two-dimensional outline of the nth follicle; The three-dimensional contour of the nth follicle; s is the scale of the ultrasound image, a fixed parameter of the ultrasound equipment; T i The probe pose is the cross-sectional image (the i-th frame cross-sectional image) containing the m-th ovarian 2D contour and the n-th follicle 2D contour.

[0112] This mapping allows us to obtain the three-dimensional contours of the ovaries and follicles. These contours contain not only their shape information on a two-dimensional plane, but also their depth information in three-dimensional space.

[0113] S207: Calculate the center point coordinates of the three-dimensional contour of the ovary and the center point coordinates of the three-dimensional contour of the follicle, respectively.

[0114] In the specific implementation, the coordinates of the center point of the three-dimensional contour of the ovary are calculated respectively. Coordinates of the center point of the follicle's three-dimensional contour .

[0115] S208: Using the center point coordinates of the ovarian three-dimensional contour as an index, query the corresponding mask value in the three-dimensional masks corresponding to the M ovarian instances, and determine the ovarian instance corresponding to the non-zero mask value as the ovarian instance matching the ovarian two-dimensional contour.

[0116] In practical implementation, the coordinates of the center point of the ovarian three-dimensional contour are used. As an index, the 3D masks corresponding to M ovarian instances are queried to determine which ovarian instance matches the 2D contour. If a mask value corresponding to the center point coordinates is not zero, it indicates that the ovarian instance matches the 2D contour, ensuring consistency between the 2D contour and the 3D model, and providing an accurate correspondence for subsequent image analysis and clinical diagnosis. For example, the coordinates in the 3D mask of the m-th ovarian instance... If the mask value at a certain point is m, then the two-dimensional contour of the ovary is matched with the m-th ovary instance.

[0117] S209: Using the center point coordinates of the three-dimensional contour of the follicle as an index, query the corresponding mask value in the three-dimensional mask corresponding to the N follicle instances, and determine the follicle instance corresponding to the non-zero mask value as the follicle instance matching the two-dimensional contour of the follicle.

[0118] In practice, similar to the matching process for ovaries, the center point coordinates are used... As an index, the 3D masks corresponding to N follicle instances are queried to determine which follicle instance matches the 2D contour. If a mask value corresponding to the center point coordinates is not zero, it indicates that the follicle instance matches the 2D contour, ensuring consistency between the 2D contour and the 3D model. For example, the coordinates in the 3D mask of the nth follicle instance... If the mask value at a given location is n, then the two-dimensional contour of that follicle is matched with the nth follicle instance. In this way, each follicle can be accurately identified and located, providing crucial information for assessing follicle development and formulating treatment plans.

[0119] The flowchart of the combined two-dimensional and three-dimensional ovarian follicle measurement method is as follows: Figure 3 As shown, images are first acquired using a two-dimensional ultrasound probe. A sequence of cross-sectional images of the ovary is obtained and preprocessed for subsequent analysis. Keyframes are extracted from the preprocessed image sequence; these frames contain crucial information for the subsequent analysis. An implicit 3D reconstruction network is used to train and render the extracted contour information to construct 3D volume data. The 3D structures of the ovary and follicles are segmented from the 3D volume data. The 2D contours are matched with the 3D segmentation results to determine the corresponding 3D structures. Finally, the 2D and 3D measurements of the ovary and follicles in the keyframes are output.

[0120] The ultrasound examination method provided in this application utilizes a two-dimensional ultrasound probe to acquire two-dimensional ultrasound measurement results of the ovary and follicles. By combining the two-dimensional cross-sectional images and probe pose information, three-dimensional volumetric data of the ovary and follicles is constructed to obtain three-dimensional ultrasound measurement results. It is evident that this application embodiment achieves two-dimensional and three-dimensional scanning of the ovary and follicles using only a two-dimensional ultrasound probe, reducing the requirements for ultrasound equipment. Furthermore, this application embodiment matches the two-dimensional contour of the ovary with its three-dimensional mask, and matches the two-dimensional contour of the follicle with its three-dimensional mask, i.e., determining the corresponding ovarian instance and the corresponding follicle instance in the three-dimensional volumetric data for each two-dimensional ovarian contour. This ensures the consistency of the two-dimensional and three-dimensional scanning results for the ovary and follicles, thereby achieving consistent two-dimensional and three-dimensional measurements of the ovary and follicles.

[0121] The following describes an ultrasonic examination device provided in an embodiment of this application. The ultrasonic examination device described below and the ultrasonic examination method described above can be referred to each other.

[0122] See Figure 4 A structural diagram of an ultrasound examination device according to an exemplary embodiment is shown, as follows: Figure 4 As shown, it includes:

[0123] The acquisition module 100 is used to acquire cross-sectional images from different orientations using a two-dimensional ultrasound probe and determine the probe pose corresponding to the cross-sectional images.

[0124] The construction module 200 is used to construct three-dimensional volume data based on at least one of the cross-sectional images and the corresponding probe poses, and to segment the three-dimensional volume data using a three-dimensional semantic segmentation model to obtain three-dimensional masks corresponding to multiple target object instances.

[0125] Extraction module 300 is used to extract the two-dimensional contour of the target object from at least one of the cross-sectional images using a two-dimensional semantic segmentation model;

[0126] Matching module 400 is used to determine a target object instance that matches the two-dimensional contour based on the three-dimensional mask.

[0127] The ultrasonic examination apparatus provided in this application constructs three-dimensional volumetric data of the target object by combining two-dimensional cross-sectional images and probe pose information. Simultaneously, by matching the two-dimensional contours with three-dimensional masks—that is, determining the target object instance corresponding to each two-dimensional contour in the three-dimensional volumetric data—it ensures the consistency of the two-dimensional and three-dimensional scanning results for the target object. Therefore, this application achieves two-dimensional and three-dimensional scanning of the target object using only a two-dimensional ultrasonic probe, reducing the requirements for ultrasonic equipment in ultrasonic examinations.

[0128] Based on the above embodiments, as a preferred embodiment, it further includes:

[0129] A determination module is used to determine a key cross-section image among the multiple cross-section images;

[0130] Accordingly, the construction module 200 is specifically used to: input at least one of the cross-sectional images and the corresponding probe poses into a three-dimensional reconstruction network for training, so as to obtain three-dimensional volume data; wherein, at least one of the cross-sectional images includes the key cross-sectional image, and, during the training process of the three-dimensional reconstruction network, the weight of the key cross-sectional image is greater than the weight of the non-key cross-sectional image.

[0131] Accordingly, the extraction module 300 is specifically used to: extract the two-dimensional contour of the target object from the key section image using a two-dimensional semantic segmentation model.

[0132] Based on the above embodiments, as a preferred implementation, the construction module 200 is specifically used to: input at least one of the cross-sectional images and the corresponding probe poses into an explicit 3D reconstruction network for training, so that the explicit 3D reconstruction network can directly output 3D volume data.

[0133] Based on the above embodiments, as a preferred embodiment, the construction module 200 includes:

[0134] The training unit is used to input at least one of the cross-sectional images and the corresponding probe poses into the implicit 3D reconstruction network for training, so that the implicit 3D reconstruction network outputs an implicit representation of the 3D volume;

[0135] The determining unit is used to select multiple viewpoints within a preset scanning viewpoint range, determine multiple sampling points of each viewpoint in three-dimensional space, and determine the attributes of each sampling point based on the implicit representation of the three-dimensional volume; wherein, if the attributes of the target sampling point do not exist in the implicit representation of the three-dimensional volume, the attributes of the adjacent sampling points of the target sampling point in the implicit representation of the three-dimensional volume are interpolated to obtain the attributes of the target sampling point.

[0136] The merging unit is used to determine the image corresponding to each viewpoint based on the attributes of each sampling point corresponding to each viewpoint, and merge the images corresponding to each viewpoint to obtain three-dimensional volume data.

[0137] Based on the above embodiments, as a preferred embodiment, the target object includes at least one of the following: ovary and follicle.

[0138] Based on the above embodiments, as a preferred implementation, if the target object includes ovaries and follicles, the determining module is specifically used to: count the number of follicles contained in each of the cross-sectional images, determine the cross-sectional images containing a number of follicles greater than a preset value as key cross-sectional images, and / or, count the size of the follicles contained in each of the cross-sectional images, determine the cross-sectional images containing follicles with a size greater than a preset size as key cross-sectional images.

[0139] Based on the above embodiments, as a preferred implementation, the mask value of the voxel points belonging to the kth target object instance in the three-dimensional mask corresponding to the kth target object instance is k, and the mask value of the voxel points not belonging to the kth target object instance is 0, 1≤k≤K, where K is the number of target object instances.

[0140] Accordingly, the matching module 400 includes:

[0141] The mapping unit is used to map the two-dimensional contour to a three-dimensional space based on the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located, so as to obtain the three-dimensional contour corresponding to the two-dimensional contour.

[0142] The matching unit is used to determine the center point coordinates of the three-dimensional contour, and use the center point coordinates as an index to query the corresponding mask value in the three-dimensional masks corresponding to N target object instances, and determine the target object instance corresponding to the non-zero mask value as the target object instance that matches the two-dimensional contour.

[0143] Based on the above embodiments, as a preferred implementation, the mapping unit includes:

[0144] The first determining subunit is used to determine multiple contour points and the two-dimensional coordinates of the multiple contour points in the two-dimensional contour.

[0145] The calculation subunit is used to obtain the ultrasound image scale. For any contour point, the product of the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located, the ultrasound image scale, and the two-dimensional coordinates of the contour point is calculated as the three-dimensional coordinates of the contour point.

[0146] The second determining subunit is used to determine the three-dimensional contour corresponding to the two-dimensional contour in three-dimensional space based on the three-dimensional coordinates of the multiple contour points.

[0147] Based on the above embodiments, as a preferred embodiment, it further includes:

[0148] A measurement module is used to determine the measurement result of the target object based on a two-dimensional contour in at least one of the cross-sectional images and a target object instance that matches the two-dimensional contour.

[0149] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0150] Based on the hardware implementation of the above program modules, and in order to implement the method of the embodiments of this application, the embodiments of this application also provide an electronic device. Figure 5 This is a structural diagram of an electronic device according to an exemplary embodiment, such as... Figure 5 As shown, the electronic device includes:

[0151] Communication interface 1 enables information exchange with other devices, such as network devices;

[0152] Processor 2 is connected to communication interface 1 to enable information exchange with other devices and, when running a computer program, executes the ultrasound examination method provided by one or more of the above-mentioned technical solutions. The computer program is stored in memory 3.

[0153] Of course, in practical applications, the various components in an electronic device are coupled together through bus system 4. It can be understood that bus system 4 is used to achieve communication and connection between these components. In addition to the data bus, bus system 4 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 5 The general will label all buses as Bus System 4.

[0154] The memory 3 in this embodiment is used to store various types of data to support the operation of the electronic device. Examples of such data include any computer program used to operate on the electronic device.

[0155] It is understood that memory 3 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), ferromagnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); magnetic surface memory can be disk storage or magnetic tape storage. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), SyncLink Dynamic Random Access Memory (SLDRAM), and Direct Rambus Random Access Memory (DRRAM).The memory 3 described in the embodiments of this application is intended to include, but is not limited to, these and any other suitable types of memory.

[0156] The methods disclosed in the embodiments of this application can be applied to processor 2, or implemented by processor 2. Processor 2 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 2 or by instructions in the form of software. The processor 2 may be a general-purpose processor, DSP, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 2 can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in memory 3. Processor 2 reads the program in memory 3 and completes the steps of the aforementioned method in combination with its hardware.

[0157] When processor 2 executes the program, it implements the corresponding processes in the various methods of the embodiments of this application. For the sake of brevity, these will not be described in detail here.

[0158] In an exemplary embodiment, this application also provides a storage medium, namely a computer storage medium, specifically a computer-readable storage medium, such as a memory 3 that stores a computer program, which can be executed by a processor 2 to complete the steps described in the aforementioned method. The computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, or CD-ROM.

[0159] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0160] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0161] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An ultrasound examination method, characterized in that, include: A two-dimensional ultrasound probe is used to acquire cross-sectional images from different orientations, and the probe pose corresponding to the cross-sectional images is determined. Three-dimensional volume data is constructed based on at least one of the cross-sectional images and the corresponding probe pose, and the three-dimensional volume data is segmented using a three-dimensional semantic segmentation model to obtain three-dimensional masks corresponding to multiple target object instances. A two-dimensional semantic segmentation model is used to extract the two-dimensional contour of the target object from at least one of the cross-sectional images; The target object instance that matches the two-dimensional contour is determined based on the three-dimensional mask.

2. The ultrasound examination method according to claim 1, characterized in that, After acquiring cross-sectional images from different orientations using a two-dimensional ultrasound probe, the method further includes: Identify the key cross-section image from among the multiple cross-section images; Accordingly, the construction of three-dimensional volume data based on at least one of the cross-sectional images and the corresponding probe pose includes: At least one of the cross-sectional images and the corresponding probe pose are input into a 3D reconstruction network for training to obtain 3D volume data; wherein, at least one of the cross-sectional images includes the key cross-sectional image, and during the training of the 3D reconstruction network, the weight of the key cross-sectional image is greater than the weight of the non-key cross-sectional image. Accordingly, the extraction of the two-dimensional contour of the target object from at least one of the cross-sectional images using a two-dimensional semantic segmentation model includes: Two-dimensional semantic segmentation models are used to extract the two-dimensional contours of the target object from the key section image.

3. The ultrasound examination method according to claim 2, characterized in that, At least one of the cross-sectional images and the corresponding probe poses are input into a 3D reconstruction network for training to obtain 3D volume data, including: At least one of the cross-sectional images and the corresponding probe poses are input into an explicit 3D reconstruction network for training, so that the explicit 3D reconstruction network can directly output 3D volume data.

4. The ultrasound examination method according to claim 2, characterized in that, At least one of the cross-sectional images and the corresponding probe poses are input into a 3D reconstruction network for training to obtain 3D volume data, including: At least one of the cross-sectional images and the corresponding probe poses are input into an implicit 3D reconstruction network for training, so that the implicit 3D reconstruction network outputs an implicit representation of the 3D volume; Multiple viewpoints are selected within a preset scanning viewpoint range, and multiple sampling points in three-dimensional space are determined for each viewpoint. The attributes of each sampling point are determined based on the implicit representation of the three-dimensional volume. If the attributes of the target sampling point do not exist in the implicit representation of the three-dimensional volume, the attributes of the adjacent sampling points of the target sampling point in the implicit representation of the three-dimensional volume are interpolated to obtain the attributes of the target sampling point. The images corresponding to each viewpoint are determined based on the attributes of each sampling point corresponding to each viewpoint, and the images corresponding to each viewpoint are merged to obtain the three-dimensional volume data.

5. The ultrasound examination method according to claim 2, characterized in that, The target objects include at least one of the following: ovaries and follicles.

6. The ultrasound examination method according to claim 5, characterized in that, If the target object includes follicles, then determining the key section image among the multiple section images includes: The number of follicles contained in each of the cross-sectional images is counted, and the cross-sectional images containing more than a preset value are determined as key cross-sectional images; And / or, The size of the follicles contained in each of the cross-sectional images is counted, and the cross-sectional images containing follicles with a size larger than a preset size are determined as key cross-sectional images.

7. The ultrasound examination method according to claim 1, characterized in that, In the 3D mask corresponding to the kth target object instance, the mask value of the voxel point belonging to the kth target object instance is k, and the mask value of the voxel point not belonging to the kth target object instance is 0, 1≤k≤K, where K is the number of target object instances; Accordingly, determining the target object instance matching the two-dimensional contour based on the three-dimensional mask includes: Based on the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located, the two-dimensional contour is mapped to three-dimensional space to obtain the three-dimensional contour corresponding to the two-dimensional contour. The center point coordinates of the three-dimensional contour are determined, and the corresponding mask value is queried in the three-dimensional mask corresponding to K target object instances using the center point coordinates as an index. The target object instance corresponding to the non-zero mask value is determined as the target object instance that matches the two-dimensional contour.

8. The ultrasound examination method according to claim 7, characterized in that, The step of mapping the two-dimensional contour to three-dimensional space based on the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located to obtain the three-dimensional contour corresponding to the two-dimensional contour includes: In the two-dimensional contour, a plurality of contour points and the two-dimensional coordinates of the plurality of contour points are determined; To obtain the ultrasound image scale, for any contour point: calculate the product of the probe pose corresponding to the cross-sectional image where the two-dimensional contour is located, the ultrasound image scale, and the two-dimensional coordinates of the contour point, and use it as the three-dimensional coordinates of the contour point. The three-dimensional contour corresponding to the two-dimensional contour in three-dimensional space is determined based on the three-dimensional coordinates of multiple contour points.

9. The ultrasound examination method according to any one of claims 1 to 8, characterized in that, After determining the target object instance matching the two-dimensional contour based on the three-dimensional mask, the method further includes: The measurement result of the target object is determined based on a two-dimensional contour in at least one of the cross-sectional images and a target object instance that matches the two-dimensional contour.

10. An ultrasonic examination device, characterized in that, include: The acquisition module is used to acquire cross-sectional images from different orientations using a two-dimensional ultrasound probe and determine the probe pose corresponding to the cross-sectional images. A construction module is used to construct three-dimensional volume data based on at least one of the cross-sectional images and the corresponding probe poses, and to segment the three-dimensional volume data using a three-dimensional semantic segmentation model to obtain three-dimensional masks corresponding to multiple target object instances; The extraction module is used to extract the two-dimensional contour of the target object from at least one of the cross-sectional images using a two-dimensional semantic segmentation model; The matching module is used to determine the target object instance that matches the two-dimensional contour based on the three-dimensional mask.

11. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the ultrasound examination method as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, implements the steps of the ultrasound examination method as described in any one of claims 1 to 9.

Images