Key point identification method and device, model training method and device, equipment and medium

Abstract

The invention discloses a key point recognition model training method and device, a key point recognition method, electronic equipment and a storage medium. The key point recognition model training method comprises the steps that original training body data collected through ultrasonic signals is acquired; wherein the original training volume data is volume data before a digital scanning converter performs signal conversion; determining a key point marked in the original training body data based on a conversion relation between the original training body data and the reconstructed training body data in the digital scanning converter and the key point marked in the reconstructed training body data; and training a key point identification model by using the original training volume data marked with the key points, wherein the trained key point identification model is used for determining the key points in the target volume data. According to the key point recognition model training method provided by the invention, the generalization of the key point recognition model is improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and more specifically, to a method and apparatus for training a key point recognition model, a method for recognizing key points, an electronic device, and a storage medium. Background Art

[0002] In the field of ultrasonic image analysis, an artificial intelligence (AI) model can be used to recognize key points in ultrasonic images. The volume data currently used in training the AI model is prone to poor generalization of the trained AI model due to accuracy loss and the like.

[0003] Therefore, how to improve the generalization of the key point recognition model is a technical problem that needs to be solved by those skilled in the art. Summary of the Invention

[0004] The purpose of the present application is to provide a method and apparatus for training a key point recognition model, an electronic device, and a computer-readable storage medium, which improve the generalization of the key point recognition model.

[0005] To achieve the above purpose, the present application provides a method for training a key point recognition model, including:

[0006] Obtain original training volume data collected through ultrasonic signals; wherein, the original training volume data is the volume data before signal conversion by a digital scan converter;

[0007] Based on the conversion relationship between the original training volume data and the reconstructed training volume data in the digital scan converter and the key points marked in the reconstructed training volume data, determine the key points marked in the original training volume data;

[0008] Use the original training volume data marked with key points to train a key point recognition model, and the trained key point recognition model is used to determine key points in the target volume data.

[0009] Wherein, the determining the key points marked in the original training volume data based on the conversion relationship between the original training volume data and the reconstructed training volume data in the digital scan converter and the key points marked in the reconstructed training volume data includes:

[0010] Use the digital scan converter to perform signal conversion on the original training volume data to obtain reconstructed training volume data;

[0011] Mark key points in the reconstructed training volume data;

[0012] Map the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the original training volume data to determine the marked key points in the original training volume data.

[0013] Among them, mapping the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the original training volume data includes:

[0014] Map the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the first coordinate system; wherein, the first coordinate system is a coordinate system established with the probe position as the origin and the symmetry axis of the region of interest as the Y-axis, and the XOY plane of the first coordinate system is the plane where the region of interest is located;

[0015] Determine the coordinates of the first intersection point in the XOY plane of the first coordinate system based on the coordinates of the key point in the first coordinate system, the first distance, and the first angle, and determine the second angle based on the coordinates of the key point in the first coordinate system and the first distance; wherein, the first distance is the distance between the projection point and the probe swing rotation axis, the projection point is the projection point of the key point on the XOY plane of the first coordinate system, the first angle is the angle between the symmetry axis of the region of interest and the center line of the probe, the first intersection point is the intersection point of the key point swinging around the probe swing rotation axis and the XOY plane of the first coordinate system, and the second angle is the swinging angle when the key point swings around the probe swing rotation axis until it intersects with the XOY plane of the first coordinate system;

[0016] Determine the second distance based on the coordinates of the first intersection point in the XOY plane of the first coordinate system and the probe scanning radius, and determine the third angle based on the coordinates of the first intersection point in the first coordinate system; wherein, the second distance is the distance between the first intersection point and the boundary of the region of interest, and the third angle is the acute angle between the line connecting the first intersection point and the origin of the first coordinate system and the Y-axis of the first coordinate system;

[0017] Determine the coordinates of the key points in the original training volume data based on the second distance, the second angle, the third angle, the number of points in the X, Y, and Z directions in the original training volume data, the physical distance in the X direction in the original training volume data, the scanning angle of the XOY plane of the first coordinate system, and the motor swing angle.

[0018] Among them, mapping the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the first coordinate system includes:

[0019] Determine the mapping factor based on the physical distance in the X direction in the reconstructed training volume data and the number of points in the X direction in the reconstructed training volume data;

[0020] Determine the X - direction coordinate value of the key point in the first coordinate system by summing the first product and the minimum physical coordinate in the X - direction of the reconstructed training volume data; wherein, the first product is the product of the X - direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor;

[0021] Determine the Y - direction coordinate value of the key point in the first coordinate system by subtracting the second product from the maximum physical coordinate in the Y - direction of the reconstructed training volume data; wherein, the first product is the product of the Y - direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor;

[0022] Determine the Z - direction coordinate value of the key point in the first coordinate system by summing the third product and the minimum physical coordinate in the Z - direction of the reconstructed training volume data; wherein, the third product is the product of the Z - direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor.

[0023] Wherein, determining the coordinates of the first intersection point in the XOY plane of the first coordinate system based on the coordinates of the key point in the first coordinate system, the first distance, and the first angle includes:

[0024] Determine the first angle by summing half of the scanning angle of the XOY plane of the first coordinate system and the starting angle of the region of interest;

[0025] Calculate the first product of the coordinate value in the X - direction of the key point in the first coordinate system and the sine value of the first angle, and calculate the second product of the coordinate value in the Y - direction of the key point in the first coordinate system and the cosine value of the first angle;

[0026] Determine the first distance by subtracting the difference between the first product, the second product, and the scanning radius from the motor swing radius;

[0027] Calculate the square root of the coordinate value in the Z - direction of the first coordinate system and the first distance, and calculate the difference between the square root and the first distance to obtain the target difference;

[0028] Calculate the third product of the target difference and the sine value of the first angle, and calculate the fourth product of the target difference and the cosine value of the first angle;

[0029] Determine the X - direction coordinate value of the first intersection point in the XOY plane of the first coordinate system by subtracting the third product from the coordinate value in the X - direction of the key point in the first coordinate system, and determine the Y - direction coordinate value of the first intersection point in the XOY plane of the first coordinate system by subtracting the fourth product from the coordinate value in the Y - direction of the key point in the first coordinate system.

[0030] Among them, determining the second included angle based on the coordinates of the key point in the first coordinate system and the first distance includes:

[0031] Calculating a first ratio of the coordinate value of the key point in the Z direction in the first coordinate system to the first distance, and determining the second included angle based on the first ratio using the arctangent function.

[0032] Among them, determining the coordinates of the key point in the original training volume data based on the second distance, the second included angle, the third included angle, the number of points in the X, Y, and Z directions in the original training volume data, the physical distance in the X direction in the original training volume data, the scanning angle of the XOY plane of the first coordinate system, and the motor swing angle includes:

[0033] Calculating a fifth product of the second distance and the number of points in the X direction in the original training volume data, and determining the coordinate value of the key point in the X direction in the original training volume data as the ratio of the fifth product to the physical distance in the X direction in the original training volume data;

[0034] Calculating a sixth product of the third included angle and the number of points in the Y direction in the original training volume data, calculating a second ratio of the sixth product to the scanning angle of the XOY plane of the first coordinate system, and determining the coordinate value of the key point in the Y direction in the original training volume data as the sum of half of the number of points in the Y direction in the original training volume data and the second ratio;

[0035] Calculating a seventh product of the second included angle and the number of points in the Z direction in the original training volume data, calculating a third ratio of the seventh product to the motor swing angle, and determining the coordinate value of the key point in the Z direction in the original training volume data as the sum of half of the number of points in the Z direction in the original training volume data and the third ratio.

[0036] To achieve the above object, the present application provides a key point recognition method, including:

[0037] Obtaining original target volume data collected by ultrasonic signals, and performing signal conversion on the original target volume data using a digital scan converter to obtain reconstructed target volume data;

[0038] Inputting the original target volume data into a key point recognition model trained by the above key point recognition model training method to predict the key points in the original target volume data;

[0039] Mapping the key points in the original target volume data into the reconstructed target volume data to obtain the key points in the reconstructed target volume data.

[0040] To achieve the above object, the present application provides an apparatus for training a key point recognition model, including:

[0041] An acquisition module, configured to acquire original training volume data collected by an ultrasonic signal; wherein, the original training volume data is volume data before signal conversion by a digital scan converter;

[0042] A determination module, configured to determine the key points labeled in the original training volume data based on the conversion relationship between the original training volume data and the reconstructed training volume data in the digital scan converter and the key points labeled in the reconstructed training volume data;

[0043] A training module, configured to train a key point recognition model by using the original training volume data with labeled key points, and the trained key point recognition model is used to determine key points in target volume data.

[0044] To achieve the above object, the present application provides an electronic device, including:

[0045] A memory, configured to store a computer program;

[0046] A processor, configured to implement the steps of the above key point recognition model training method or the above key point recognition method when executing the computer program.

[0047] To achieve the above object, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above key point recognition model training method or the above key point recognition method are implemented.

[0048] As can be seen from the above solutions, a key point recognition model training method provided by the present application includes: acquiring original training volume data collected by an ultrasonic signal; wherein, the original training volume data is volume data before signal conversion by a digital scan converter; determining the key points labeled in the original training volume data based on the conversion relationship between the original training volume data and the reconstructed training volume data in the digital scan converter and the key points labeled in the reconstructed training volume data; training a key point recognition model by using the original training volume data with labeled key points, and the trained key point recognition model is used to determine key points in target volume data.

[0049] The key point recognition model training method provided by this application determines the labeled key points from the original training volume data before DSC reconstruction, and trains the key point recognition model using the original training volume data with labeled key points, that is, trains the key point recognition model using the original training volume data before DSC reconstruction. Since the accuracy of the original training volume data before DSC reconstruction is higher than that of the volume data after DSC reconstruction, the generalization of the trained key point recognition model is relatively high. Thus, it can be seen that the key point recognition model training method provided by this application improves the generalization of the key point recognition model. This application also discloses a key point recognition model training device, a key point recognition method, an electronic device, and a computer-readable storage medium, which can also achieve the above technical effects.

[0050] It should be understood that the above general description and the following detailed description are only exemplary and do not limit this application. Brief Description of the Drawings

[0051] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. The drawings are used to provide a further understanding of this application, and constitute a part of the specification, and are used together with the following specific embodiments to explain this application, but do not constitute a limitation to this application. In the drawings:

[0052] Figure 1 It is a flowchart of a key point recognition model training method shown according to an exemplary embodiment;

[0053] Figure 2 It is a schematic diagram of key points labeled in the reconstructed training volume data shown according to an exemplary embodiment;

[0054] Figure 3 For Figure 2 The corresponding schematic diagram of key points labeled in the original training volume data;

[0055] Figure 4 It is another schematic diagram of key points labeled in the reconstructed training volume data shown according to an exemplary embodiment;

[0056] Figure 5 For Figure 4 The corresponding schematic diagram of key points labeled in the original training volume data;

[0057] Figure 6Flowchart of a method for mapping key points annotated in reconstructed training volume data to original training volume data according to an exemplary embodiment;

[0058] Figure 7 Schematic diagram of a region of interest according to an exemplary embodiment;

[0059] Figure 8 Flowchart of a key point recognition method according to an exemplary embodiment;

[0060] Figure 9 Structural diagram of a key point recognition model training device according to an exemplary embodiment;

[0061] Figure 10 Structural diagram of a key point recognition device according to an exemplary embodiment;

[0062] Figure 11 Structural diagram of an electronic device according to an exemplary embodiment. Detailed implementation manners

[0063] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application. In addition, in the embodiments of the present application, "first", "second", etc. are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence.

[0064] The embodiments of the present application disclose a key point recognition model training method, which improves the generalization of the key point recognition model.

[0065] Optionally, the key point recognition model training can be executed by an ultrasound device or other electronic devices other than the ultrasound device. For the latter case, after obtaining the trained key point recognition model, the electronic device can send it to the ultrasound device, and the ultrasound device can determine key points in the target volume data based on the trained key point recognition model.

[0066] See Figure 1 , a flowchart of a key point recognition model training method according to an exemplary embodiment, as Figure 1 shown, includes:

[0067] S101: Obtain original training volume data collected through ultrasound signals; wherein, the original training volume data is the volume data before signal conversion by a digital scan converter.

[0068] The purpose of this embodiment is to train a key point recognition model. In this step, the original training volume data collected by ultrasonic signals is obtained, and this original training volume data is the volume data before DSC reconstruction.

[0069] Optionally, ultrasonic scanning can be performed on the scanned object containing the region of interest through an ultrasonic device to obtain the corresponding volume data, and at least part of the volume data in the volume data is determined as the original training volume data. The region of interest can be a specific body part or a specific lesion tissue.

[0070] The inventor found that since most of the volume data after DSC (Digital Scan Converter) reconstruction will have a certain accuracy loss. For example, the original volume data is about 90M, and the volume data after DSC reconstruction may only be 20M - 30M. Therefore, using the volume data after DSC reconstruction to train the AI model will result in a relatively poor generalization ability of the trained AI model. This embodiment determines the original training volume data based on the volume data before signal conversion by the digital scan converter, and then trains the model. High-precision training data can obtain a model with higher accuracy and better generalization ability.

[0071] S102: Determine the key points marked in the original training volume data based on the conversion relationship between the original training volume data and the reconstructed training volume data in the digital scan converter and the key points marked in the reconstructed training volume data.

[0072] Among them, the key points can be the feature points or feature regions corresponding to the region of interest.

[0073] As a feasible implementation manner, that is, based on the conversion relationship between the original training volume data and the reconstructed training volume data in the digital scan converter and the key points marked in the reconstructed training volume data, to determine the key points marked in the original training volume data, including: using the digital scan converter to perform signal conversion on the original training volume data to obtain the reconstructed training volume data; marking the key points in the reconstructed training volume data; mapping the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the original training volume data, so as to determine the key points marked in the original training volume data. In specific implementation, the original training volume data can be reconstructed to obtain the reconstructed training volume data, and the mapping parameters from the original training volume data to the reconstructed training volume data are saved. Then, the key points are marked in the volume data after DSC reconstruction, that is, the reconstructed training volume data. According to the mapping parameters from the original training volume data to the reconstructed training volume data, the key points marked in the reconstructed training volume data are mapped to the key points in the original training volume data. It can be understood that the volume data after DSC reconstruction is easy to mark. Marking the key points in the reconstructed training volume data can reduce the marking difficulty of the key points, improve the marking efficiency of the key points, and thus improve the training efficiency of the key point recognition model.

[0074] The type of the key points marked in this step can be flexibly set according to different application scenarios. For example, if the key point recognition model trained in this embodiment is used to determine the boundary of the target object in the target volume data, then the key points marked in this step are the boundary points of the target object, and the key points marked in the reconstructed training volume data are as shown in Figure 2 and its conversion to the key points marked in the original training volume data is as shown in Figure 3 . Another example is that if the key point recognition model trained in this embodiment is used to determine the feature points concerned by the user in the target volume data, then the key points marked in this step are the feature points concerned by the user, and the key points marked in the reconstructed training volume data are as shown in Figure 4 and its conversion to the key points marked in the original training volume data is as shown in Figure 5 .

[0075] S103: Training a key point recognition model using the original training volume data with marked key points, and the trained key point recognition model is used to determine key points in the target volume data.

[0076] In this step, the original training volume data with marked key points is used to train a key point recognition model. The key point recognition model can be a neural network model, for example, a convolutional neural network (CNN) model, etc., which is not specifically limited herein. In the inference stage, the trained key point recognition model is used to determine key points in the target volume data. The target volume data is the volume data for which key point recognition is required, and can be the volume data directly collected through ultrasonic signals or the volume data after DSC reconstruction, which is different from the original training volume data and the reconstructed training volume data.

[0077] The key point recognition model training method provided by the embodiments of the present application determines the marked key points in the original training volume data before DSC reconstruction, and uses the original training volume data with marked key points to train the key point recognition model, that is, uses the original training volume data before DSC reconstruction to train the key point recognition model. Since the accuracy of the original training volume data before DSC reconstruction is higher than that of the volume data after DSC reconstruction, the generalization of the trained key point recognition model is relatively high. And the key points in the original training volume data are obtained by converting the key points marked in the reconstructed training volume data, without the need for actual marking in the large amount of original training volume data, which can effectively reduce the marking cost of key points. It can be seen that the key point recognition model training method provided by the embodiments of the present application improves the generalization of the key point recognition model without significantly increasing the training data marking cost.

[0078] This embodiment introduces the method of mapping the key points marked in the reconstructed training volume data to the original training volume data. It should be noted that some mapping parameters are required in the process of reconstructing the original training volume data to obtain the reconstructed training volume data. This embodiment exactly uses these mapping parameters to map the key points marked in the reconstructed training volume data back to the original training volume data. The mapping parameters can include: the physical distance in the X direction in the reconstructed training volume data, the number of points in the X direction in the reconstructed training volume data, the minimum physical coordinate in the X direction in the reconstructed training volume data, the maximum physical coordinate in the Y direction in the reconstructed training volume data, the minimum physical coordinate in the Z direction in the reconstructed training volume data, the starting angle of the region of interest, the scanning angle of the XOY plane of the first coordinate system, the motor swing angle, the motor swing radius, the scanning radius, the physical distance in the X direction in the original training volume data, the number of points in the X direction in the original training volume data, the number of points in the Y direction in the original training volume data, the number of points in the Z direction in the original training volume data, etc.

[0079] See Figure 6 , a flowchart of a method for mapping the key points marked in the reconstructed training volume data to the original training volume data shown according to an exemplary embodiment, as Figure 6 shown, includes:

[0080] S201: Map the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the first coordinate system. Among them, the first coordinate system is a coordinate system established with the probe position as the origin and the symmetry axis of the region of interest as the Y-axis. The XOY plane of the first coordinate system is the plane where the region of interest is located.

[0081] Optionally, a certain point on the probe or a certain point near the probe can be determined as the origin. Taking the convex array probe as an example, the intersection point of the extension lines of the convex array element directions can be determined as the origin.

[0082] In a specific implementation, as Figure 7 shown, establish the first coordinate system XOY with the probe position as the origin and the symmetry axis of the region of interest as the Y-axis. The XOY plane of the first coordinate system is the plane where the region of interest is located. Establish the second coordinate system X’OY’ with the probe position as the origin and the center line of the probe as the Y’-axis. The X’OY’ plane of the first coordinate system is also the plane where the region of interest is located. Figure 7 The fan-shaped area in it is the region of interest (ROI).

[0083] In this step, map the coordinates (xpos, ypox, zpos) of the key point P marked in the reconstructed training volume data to the coordinates (xWorld, yWorld, zWorld) in the first coordinate system. As a feasible implementation manner, mapping the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the first coordinate system includes: determining the mapping factor based on the physical distance in the X direction in the reconstructed training volume data and the number of points in the X direction in the reconstructed training volume data; determining the sum of the first product and the minimum physical coordinate in the X direction in the reconstructed training volume data as the X-direction coordinate value of the key point in the first coordinate system. Among them, the first product is the product of the X-direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor; the difference between the maximum physical coordinate in the Y direction in the reconstructed training volume data and the second product is determined as the Y-direction coordinate value of the key point in the first coordinate system. Among them, the first product is the product of the Y-direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor; determining the sum of the third product and the minimum physical coordinate in the Z direction in the reconstructed training volume data as the Z-direction coordinate value of the key point in the first coordinate system. Among them, the third product is the product of the Z-direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor.

[0084] In a specific implementation, first determine the ratio of the physical distance fXSizeMmDSC in the X direction in the reconstructed training volume data to the number of points iXSizeDSC in the X direction in the reconstructed training volume data minus one as the mapping factor dstPixelToWorld, that is, dstPixelToWorld = fXSizeMmDSC / (iXSizeDSC - 1).

[0085] Secondly, according to the mapping factor dstPixelToWorld, map the coordinates (xpos, ypox, zpos) of the key points marked in the reconstructed training volume data to the coordinates (xWorld, yWorld, zWorld) in the first coordinate system. Specifically:

[0086] xWorld = xPoS * dstPixelToWorld + minX;

[0087] yWorld = -yPos * dstPixelToWorld + maxY;

[0088] zWorld = zPoS * dstPixelToWorld + minZ;

[0089] Wherein, minX is the minimum physical coordinate in the X direction of the reconstructed training volume data, maxY is the maximum physical coordinate in the Y direction of the reconstructed training volume data, and minZ is the minimum physical coordinate in the Z direction of the reconstructed training volume data.

[0090] S202: Determine the coordinates of the first intersection point in the XOY plane of the first coordinate system based on the coordinates of the key point in the first coordinate system, the first distance, and the first angle. Determine the second angle based on the coordinates of the key point in the first coordinate system and the first distance. The first distance is the distance between the projection point and the probe swing rotation axis. The projection point is the projection point of the key point on the XOY plane of the first coordinate system. The first angle is the angle between the symmetry axis of the region of interest and the probe center line. The first intersection point is the intersection point of the key point swinging around the probe swing rotation axis and the XOY plane of the first coordinate system. The second angle is the swinging angle when the key point swings around the probe swing rotation axis until it intersects with the XOY plane of the first coordinate system.

[0091] In a specific implementation, determine the projection point S of the key point P on the XOY plane of the first coordinate system, determine the first distance TS between the projection point S and the probe swing rotation axis L, determine the first angle θ between the symmetry axis of the region of interest and the probe center line, that is, the angle θ between the Y axis and the Y' axis. Swing the key point P around the probe swing rotation axis L, that is, swing PT around the probe swing rotation axis L, so that it intersects with the XOY plane to obtain the first intersection point W, WT = PT, and the swinging angle is the second angle phi.

[0092] Determine the sum of half of the scanning angle of the XOY plane of the first coordinate system and the starting angle of the region of interest as the first angle. θ = SpanAngle / 2 + fStartAngle, where SpanAngle is the scanning angle of the XOY plane and fStartAngle is the starting angle of the region of interest.

[0093] Calculate the first product of the coordinate value of the key point in the X direction in the first coordinate system and the sine value of the first angle, and calculate the second product of the coordinate value of the key point in the Y direction in the first coordinate system and the cosine value of the first angle; determine the first distance by taking the difference between the motor swing radius and the first product, the second product, and the scanning radius. TS = -(sinθ * xWorld + cosθ * yWorld + XYRadius - ZRadius), where XYRadius is the scanning radius, that is, the distance from the center of the probe to the boundary of the region of interest, and ZRadius is the motor swing radius, that is, the distance from the center of the motor swing to the boundary of the region of interest.

[0094] In this step, first determine the coordinates (X1, Y1) of the first intersection point W in the XOY plane according to the coordinates (xWorld, yWorld, zWorld) of the key point in the first coordinate system, the first distance TS, and the first angle θ. As a feasible implementation method, determining the coordinates of the first intersection point in the XOY plane of the first coordinate system based on the coordinates of the key point in the first coordinate system, the first distance, and the first angle includes: calculating the square root of the coordinate value in the Z direction in the first coordinate system and the first distance, calculating the difference between the square root and the first distance to obtain the target difference; calculating the third product of the target difference and the sine value of the first angle, and calculating the fourth product of the target difference and the cosine value of the first angle; determining the X-direction coordinate value of the first intersection point in the XOY plane of the first coordinate system by taking the difference between the coordinate value of the key point in the X direction in the first coordinate system and the third product, and determining the Y-direction coordinate value of the first intersection point in the XOY plane of the first coordinate system by taking the difference between the coordinate value of the key point in the Y direction in the first coordinate system and the fourth product.

[0095]

[0096]

[0097] Secondly, determine the second angle phi according to the coordinates (xWorld, yWorld, zWorld) of the key point in the first coordinate system and the first distance TS. As a feasible implementation method, determining the second angle based on the coordinates of the key point in the first coordinate system and the first distance includes: calculating the first ratio of the coordinate value in the Z direction in the first coordinate system to the first distance, and using the arctangent function to determine the second angle based on the first ratio. That is, phi = arctan(zWorld / TS).

[0098] S203: Determine the second distance based on the coordinates of the first intersection point in the XOY plane of the first coordinate system and the probe scanning radius, and determine the third angle based on the coordinates of the first intersection point in the first coordinate system; wherein, the second distance is the distance between the first intersection point and the boundary of the region of interest, and the third angle is the acute angle between the line connecting the first intersection point and the origin of the first coordinate system and the Y-axis of the first coordinate system.

[0099] In this step, first, according to the coordinates (X1, Y1) of point W in the XOY plane and the probe scanning radius XYRadius, determine the second distance R between point W and the boundary of the region of interest. The second distance R is the difference between the distance between point W and the origin and the probe scanning radius, that is

[0100] Secondly, according to the coordinates (X1, Y1) of point W in the XOY plane, determine the third angle theta. The third angle theta is the angle between OW and the Y-axis, that is theta = arctan(X1 / -Y1).

[0101] S204: Determine the coordinates of the key points in the original training volume data based on the second distance, the second angle, the third angle, the number of points in the X, Y, and Z directions in the original training volume data, the physical distance in the X direction in the original training volume data, the scanning angle of the XOY plane of the first coordinate system, and the motor swing angle.

[0102] In this step, according to the second distance R, the second angle phi, the third angle the, the number of points SRC_XSIZE in the X direction in the original training volume data, the number of points SRC_YSIZE in the Y direction in the original training volume data, the number of points SRC_ZSIZE in the Z direction in the original training volume data, the physical distance srcXsizeMM in the X direction in the original training volume data, the scanning angle SpanAngle of the XOY plane, and the motor swing angle MotorAngle, determine the coordinates (srcX, srcY, srcZ) of the key points in the original training volume data.

[0103] As a feasible implementation manner, the coordinates of key points in the original training volume data are determined based on the second distance, the second included angle, the third included angle, the number of points in the X, Y, and Z directions in the original training volume data, the physical distance in the X direction in the original training volume data, the scanning angle of the XOY plane of the first coordinate system, and the motor swing angle, including: calculating the fifth product of the second distance and the number of points in the X direction in the original training volume data, and determining the coordinate value in the X direction of the key point in the original training volume data as the ratio of the fifth product to the physical distance in the X direction in the original training volume data; calculating the sixth product of the third included angle and the number of points in the Y direction in the original training volume data, calculating the second ratio of the sixth product to the scanning angle of the XOY plane of the first coordinate system, and determining the coordinate value in the Y direction of the key point in the original training volume data as the sum of half of the number of points in the Y direction in the original training volume data and the second ratio; calculating the seventh product of the second included angle and the number of points in the Z direction in the original training volume data, calculating the third ratio of the seventh product to the motor swing angle, and determining the coordinate value in the Z direction of the key point in the original training volume data as the sum of half of the number of points in the Z direction in the original training volume data and the third ratio.

[0104] srcX = R * SRC_XSIZE / srcXsizeMM;

[0105] srcY = theta * SRC_YSIZE / SpanAngle + SRC_YSIZE / 2;

[0106] srcZ = phi * SRC_ZSIZE / MotorAngle + SRC_ZSIZE / 2.

[0107] It can be seen that in this embodiment, the key points marked in the reconstructed training volume data are mapped to the original training volume data through the mapping parameters.

[0108] An embodiment of the present application discloses a method for training a key point recognition model. Refer to Figure 8 , according to the flowchart of a key point recognition method shown in an exemplary embodiment, as Figure 8 shown, including:

[0109] S301: Obtain the original target volume data collected through ultrasonic signals, and perform signal conversion on the original target volume data by using a digital scan converter to obtain the reconstructed target volume data;

[0110] The original target volume data may be volume data collected in real time by an ultrasonic device by emitting ultrasonic signals, or may be historical volume data collected in advance.

[0111] S302: Input the original target volume data into the key point recognition model trained by the above key point recognition model training method to predict the key points in the original target volume data;

[0112] S303: Map the key points in the original target volume data to the reconstructed target volume data to obtain the key points in the reconstructed target volume data.

[0113] In the inference stage, obtain the original target volume data collected through ultrasonic signals, perform DSC reconstruction on the original target volume data to obtain the reconstructed target volume data, save the mapping parameters from the original target volume data to the reconstructed target volume data, input the original target volume data into the trained key point recognition model to predict the key points in the original target volume data, and then map the key points in the original target volume data to the reconstructed target volume data according to the mapping parameters from the original target volume data to the reconstructed target volume data. The key points mapped in the reconstructed target volume data can be determined as the key point recognition result of the reconstructed target volume data.

[0114] Furthermore, based on the reconstructed target volume data and its key points, the corresponding ultrasonic image and the corresponding key point identifier (the key point identifier can be a marker with an identification function such as a wireframe, text, etc.) can be displayed on the display screen. In this way, accurate identification of specific body parts, lesion tissues, etc. can be achieved and accurate indication can be given to the user.

[0115] Next, an apparatus for training a key point recognition model provided in an embodiment of the present application will be introduced. The apparatus for training a key point recognition model described below can be referred to in mutual reference with the method for training a key point recognition model described above.

[0116] See Figure 9 , a structural diagram of an apparatus for training a key point recognition model shown according to an exemplary embodiment, as Figure 9 shown, includes:

[0117] An acquisition module 101, configured to acquire original training volume data collected through ultrasonic signals; wherein, the original training volume data is the volume data before signal conversion by a digital scan converter;

[0118] A determination module 102, configured to determine the key points labeled in the original training volume data based on the conversion relationship between the original training volume data and the reconstructed training volume data in the digital scan converter and the key points labeled in the reconstructed training volume data;

[0119] A training module 103, configured to train a key point recognition model using the original training volume data with labeled key points, and the trained key point recognition model is used to determine key points in the target volume data.

[0120] The key point recognition model training device provided by the embodiment of the present application determines the labeled key points in the original training volume data before DSC reconstruction, and trains the key point recognition model by using the original training volume data with labeled key points, that is, trains the key point recognition model by using the original training volume data before DSC reconstruction. Since the accuracy of the original training volume data before DSC reconstruction is higher than that of the volume data after DSC reconstruction, the generalization of the trained key point recognition model is relatively high. It can be seen that the key point recognition model training device provided by the embodiment of the present application improves the generalization of the key point recognition model.

[0121] As a preferred implementation manner, the determining module 102 includes:

[0122] A conversion sub-module, configured to perform signal conversion on the original training volume data by using a digital scan converter to obtain reconstructed training volume data;

[0123] A labeling sub-module, configured to label key points in the reconstructed training volume data;

[0124] A mapping sub-module, configured to map the coordinates of the key points labeled in the reconstructed training volume data to the coordinates in the original training volume data, so as to determine the labeled key points in the original training volume data.

[0125] As a preferred implementation manner, the mapping sub-module includes:

[0126] A mapping unit, configured to map the coordinates of the key points labeled in the reconstructed training volume data to the coordinates in the first coordinate system; wherein, the first coordinate system is a coordinate system established with the probe position as the origin and the symmetry axis of the region of interest as the Y axis, and the XOY plane of the first coordinate system is the plane where the region of interest is located;

[0127] A first determining unit, configured to determine the coordinates of the first intersection point in the XOY plane of the first coordinate system based on the coordinates of the key point in the first coordinate system, the first distance, and the first angle, and determine the second angle based on the coordinates of the key point in the first coordinate system and the first distance; wherein, the first distance is the distance between the projection point and the probe swing rotation axis, the projection point is the projection point of the key point on the XOY plane of the first coordinate system, the first angle is the angle between the symmetry axis of the region of interest and the probe center line, the first intersection point is the intersection point of the key point swinging around the probe swing rotation axis and the XOY plane of the first coordinate system, and the second angle is the swinging angle when the key point swings around the probe swing rotation axis until it intersects with the XOY plane of the first coordinate system;

[0128] A second determination unit, configured to determine a second distance based on the coordinates of the first intersection point in the XOY plane of the first coordinate system and the probe scanning radius, and determine a third angle based on the coordinates of the first intersection point in the first coordinate system; wherein, the second distance is the distance between the first intersection point and the boundary of the region of interest, and the third angle is the acute angle between the line connecting the first intersection point and the origin of the first coordinate system and the Y axis of the first coordinate system;

[0129] A third determination unit, configured to determine the coordinates of the key points in the original training volume data based on the second distance, the second angle, the third angle, the number of points in the X, Y, and Z directions in the original training volume data, the physical distance in the X direction in the original training volume data, the scanning angle of the XOY plane of the first coordinate system, and the motor swing angle.

[0130] As a preferred embodiment, the mapping unit is specifically configured to: determine a mapping factor based on the physical distance in the X direction in the reconstructed training volume data and the number of points in the X direction in the reconstructed training volume data; determine the sum of the first product and the minimum physical coordinate in the X direction in the reconstructed training volume data as the X-direction coordinate value of the key point in the first coordinate system; wherein, the first product is the product of the X-direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor; determine the difference between the maximum physical coordinate in the Y direction in the reconstructed training volume data and the second product as the Y-direction coordinate value of the key point in the first coordinate system; wherein, the first product is the product of the Y-direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor; determine the sum of the third product and the minimum physical coordinate in the Z direction in the reconstructed training volume data as the Z-direction coordinate value of the key point in the first coordinate system; wherein, the third product is the product of the Z-direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor.

[0131] As a preferred embodiment, the first determining unit is specifically configured to: determine the sum of half of the scanning angle of the XOY plane of the first coordinate system and the starting angle of the region of interest as the first included angle; calculate a first product of the coordinate value of the key point in the X direction in the first coordinate system and the sine value of the first included angle, and calculate a second product of the coordinate value of the key point in the Y direction in the first coordinate system and the cosine value of the first included angle; determine the difference between the motor swing radius and the first product, the second product, and the scanning radius as the first distance; calculate the square root of the coordinate value in the Z direction in the first coordinate system and the first distance, and calculate the difference between the square root and the first distance to obtain a target difference; calculate a third product of the target difference and the sine value of the first included angle, and calculate a fourth product of the target difference and the cosine value of the first included angle; determine the difference between the coordinate value of the key point in the X direction in the first coordinate system and the third product as the X-direction coordinate value of the first intersection point in the XOY plane of the first coordinate system, and determine the difference between the coordinate value of the key point in the Y direction in the first coordinate system and the fourth product as the Y-direction coordinate value of the first intersection point in the XOY plane of the first coordinate system.

[0132] As a preferred embodiment, the first determining unit is specifically configured to: calculate a first ratio of the coordinate value of the key point in the Z direction in the first coordinate system to the first distance, and determine the second included angle based on the first ratio using the arctangent function.

[0133] As a preferred embodiment, the third determining unit is specifically configured to: calculate a fifth product of the second distance and the number of points in the X direction in the original training volume data, and determine the ratio of the fifth product to the physical distance in the X direction in the original training volume data as the coordinate value of the key point in the X direction in the original training volume data; calculate a sixth product of the third included angle and the number of points in the Y direction in the original training volume data, calculate a second ratio of the sixth product to the scanning angle of the XOY plane of the first coordinate system, and determine the sum of half of the number of points in the Y direction in the original training volume data and the second ratio as the coordinate value of the key point in the Y direction in the original training volume data; calculate a seventh product of the second included angle and the number of points in the Z direction in the original training volume data, calculate a third ratio of the seventh product to the motor swing angle, and determine the sum of half of the number of points in the Z direction in the original training volume data and the third ratio as the coordinate value of the key point in the Z direction in the original training volume data.

[0134] Next, a key point recognition device provided by an embodiment of the present application will be introduced. The key point recognition device described below can be referred to with the key point recognition method described above.

[0135] See Figure 10 , a structural diagram of a key point recognition device shown according to an exemplary embodiment, as Figure 10 shown, includes:

[0136] A conversion module 201, configured to obtain original target body data collected by an ultrasonic signal, and perform signal conversion on the original target body data by using a digital scan converter to obtain reconstructed target body data;

[0137] An input module 202, configured to input the original target body data into a key point recognition model trained by the above-mentioned key point recognition model training device to predict key points in the original target body data;

[0138] A mapping module 203, configured to map key points in the original target body data to the reconstructed target body data to obtain key points in the reconstructed target body data.

[0139] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment related to the method, and will not be elaborated here.

[0140] Based on the hardware implementation of the above program modules, and in order to implement the method of the embodiments of the present application, the embodiments of the present application also provide an electronic device, Figure 11 which is a structural diagram of an electronic device shown according to an exemplary embodiment, as Figure 11 shown, the electronic device includes:

[0141] A communication interface 1, capable of performing information interaction with other devices such as network devices;

[0142] A processor 2, connected to the communication interface 1 to implement information interaction with other devices, and when running a computer program, configured to execute the key point recognition model training or key point recognition method provided by the above one or more technical solutions. And the computer program is stored on a memory 3.

[0143] Of course, in actual application, each component in the electronic device is coupled together through a bus system 4. It can be understood that the bus system 4 is used to realize the connection and communication between these components. The bus system 4 includes, in addition to a data bus, a power bus, a control bus, and a status signal bus. However, for the sake of clarity, in Figure 11 all kinds of buses are labeled as the bus system 4.

[0144] The memory 3 in the embodiments of the present application is used to store various types of data to support the operation of the electronic device. Examples of these data include: any computer program for operating on the electronic device.

[0145] It can be understood that the memory 3 can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM, Read Only Memory), a programmable read-only memory (PROM, Programmable Read-Only Memory), an erasable programmable read-only memory (EPROM, Erasable Programmable Read-Only Memory), an electrically erasable programmable read-only memory (EEPROM, Electrically Erasable Programmable Read-Only Memory), a ferromagnetic random access memory (FRAM, ferromagnetic random access memory), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM, Compact Disc Read-Only Memory); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM, Random Access Memory), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as a static random access memory (SRAM, Static Random Access Memory), a synchronous static random access memory (SSRAM, Synchronous Static Random Access Memory), a dynamic random access memory (DRAM, Dynamic Random Access Memory), a synchronous dynamic random access memory (SDRAM, Synchronous Dynamic Random Access Memory), a double data rate synchronous dynamic random access memory (DDR SDRAM, Double Data Rate Synchronous Dynamic Random Access Memory), an enhanced synchronous dynamic random access memory (ESDRAM, Enhanced Synchronous Dynamic Random Access Memory), a sync link dynamic random access memory (SLDRAM, SyncLink Dynamic Random Access Memory), a direct rambus random access memory (DRRAM, Direct Rambus Random Access Memory).The memory 3 described in the embodiments of the present application is intended to include, but not limited to, these and any other suitable types of memories.

[0146] The method disclosed in the embodiments of the present application above can be applied to the processor 2 or implemented by the processor 2. The 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 in the hardware of the processor 2 or the instructions in the form of software. The above-mentioned processor 2 may be a general-purpose processor, DSP, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 2 can implement or execute each method, step, and logic block diagram disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or any conventional processor, etc. Combining the steps of the method disclosed in the embodiments of the present application, it can be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by the combination of the hardware and software modules in the decoding processor. The software module may be located in the storage medium, which is located in the memory 3. The processor 2 reads the program in the memory 3 and combines its hardware to complete the steps of the foregoing method.

[0147] When the processor 2 executes the program, it implements the corresponding processes in each method of the embodiments of the present application. For the sake of brevity, it will not be elaborated here.

[0148] In an exemplary embodiment, the embodiments of the present application further provide a storage medium, namely a computer storage medium, specifically a computer-readable storage medium, such as a memory 3 including a stored computer program. The above computer program can be executed by the processor 2 to complete the steps described in the foregoing method. The computer-readable storage medium may be a FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, CD-ROM, or other memories.

[0149] Those of ordinary skill in the art can understand that all or part of the steps to implement the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above method embodiments; and the foregoing storage medium includes: removable storage devices, ROM, RAM, magnetic disks, or optical discs and other media that can store program codes.

[0150] Alternatively, if the above integrated units in the present application are implemented in the form of software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present application essentially or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing an electronic device (which can be a personal computer, a server, a network device, etc.) to execute all or part of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes, such as removable storage devices, ROM, RAM, magnetic disks, or optical discs.

[0151] As described above, the foregoing is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A method for training a key point recognition model, characterized in that, Including: Obtaining original training volume data collected through ultrasonic signals; wherein, the original training volume data is the volume data before signal conversion by a digital scan converter; Determining the key points marked in the original training volume data based on the conversion relationship between the original training volume data and the reconstructed training volume data in the digital scan converter and the key points marked in the reconstructed training volume data; Training a key point recognition model using the original training volume data with marked key points, and the trained key point recognition model is used to determine key points in the target volume data.

2. The method for training a key point recognition model according to claim 1, characterized in that, The determining the key points marked in the original training volume data based on the conversion relationship between the original training volume data and the reconstructed training volume data in the digital scan converter and the key points marked in the reconstructed training volume data includes: Performing signal conversion on the original training volume data using the digital scan converter to obtain reconstructed training volume data; Marking key points in the reconstructed training volume data; Mapping the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the original training volume data to determine the key points marked in the original training volume data.

3. The method for training a key point recognition model according to claim 2, characterized in that, Mapping the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the original training volume data includes: Mapping the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the first coordinate system; wherein, the first coordinate system is a coordinate system established with the probe position as the origin and the symmetry axis of the region of interest as the Y-axis, and the XOY plane of the first coordinate system is the plane where the region of interest is located; Determining the coordinates of the first intersection point in the XOY plane of the first coordinate system based on the coordinates of the key point in the first coordinate system, the first distance, and the first angle, and determining the second angle based on the coordinates of the key point in the first coordinate system and the first distance; wherein, the first distance is the distance between the projection point and the probe swing rotation axis, the projection point is the projection point of the key point on the XOY plane of the first coordinate system, the first angle is the angle between the symmetry axis of the region of interest and the probe center line, the first intersection point is the intersection point of the key point swinging around the probe swing rotation axis and the XOY plane of the first coordinate system, and the second angle is the swinging angle when the key point swings around the probe swing rotation axis until it intersects with the XOY plane of the first coordinate system; Determining the second distance based on the coordinates of the first intersection point in the XOY plane of the first coordinate system and the probe scanning radius, and determining the third angle based on the coordinates of the first intersection point in the first coordinate system; wherein, the second distance is the distance between the first intersection point and the boundary of the region of interest, and the third angle is the acute angle between the line connecting the first intersection point and the origin of the first coordinate system and the Y-axis of the first coordinate system; Determine the coordinates of the key points in the original training volume data based on the second distance, the second included angle, the third included angle, the number of points in the X, Y, and Z directions in the original training volume data, the physical distance in the X direction in the original training volume data, the scanning angle of the XOY plane of the first coordinate system, and the motor swing angle.

4. The method for training a key point recognition model according to claim 3, characterized in that, Map the coordinates of the key points marked in the reconstructed training volume data to the coordinates in the first coordinate system, including: Determine the mapping factor based on the physical distance in the X direction in the reconstructed training volume data and the number of points in the X direction in the reconstructed training volume data; Determine the X-direction coordinate value of the key point in the first coordinate system as the sum of the first product and the minimum physical coordinate in the X direction in the reconstructed training volume data; wherein, the first product is the product of the X-direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor; Determine the Y-direction coordinate value of the key point in the first coordinate system as the difference between the maximum physical coordinate in the Y direction in the reconstructed training volume data and the second product; wherein, the first product is the product of the Y-direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor; Determine the Z-direction coordinate value of the key point in the first coordinate system as the sum of the third product and the minimum physical coordinate in the Z direction in the reconstructed training volume data; wherein, the third product is the product of the Z-direction coordinate value of the key point marked in the reconstructed training volume data and the mapping factor.

5. The method for training a key point recognition model according to claim 3, characterized in that, The method for determining the coordinates of the first intersection point in the XOY plane of the first coordinate system based on the coordinates of the key point in the first coordinate system, the first distance, and the first included angle includes: Determine the first included angle as the sum of half of the scanning angle of the XOY plane of the first coordinate system and the starting angle of the region of interest; Calculate the first product of the coordinate value in the X direction of the key point in the first coordinate system and the sine value of the first included angle, and calculate the second product of the coordinate value in the Y direction of the key point in the first coordinate system and the cosine value of the first included angle; Determine the first distance as the difference between the motor swing radius and the difference between the first product, the second product, and the scanning radius; Calculate the square root of the coordinate value in the Z direction in the first coordinate system and the first distance, and calculate the difference between the square root and the first distance to obtain the target difference; Calculate the third product of the target difference and the sine value of the first included angle, and calculate the fourth product of the target difference and the cosine value of the first included angle; Determine the X-direction coordinate value of the first intersection point in the XOY plane of the first coordinate system as the difference between the coordinate value in the X direction of the key point in the first coordinate system and the third product, and determine the Y-direction coordinate value of the first intersection point in the XOY plane of the first coordinate system as the difference between the coordinate value in the Y direction of the key point in the first coordinate system and the fourth product.

6. The method for training a key point recognition model according to claim 5, characterized in that, The method for determining the second included angle based on the coordinates of the key point in the first coordinate system and the first distance includes: Calculate a first ratio of the coordinate value of the key point in the Z direction in the first coordinate system to the first distance, and use the arctangent function to determine the second included angle based on the first ratio.

7. The method for training a key point recognition model according to claim 3, characterized in that, Determining the coordinates of the key point in the original training volume data based on the second distance, the second included angle, the third included angle, the number of points in the X, Y, and Z directions in the original training volume data, the physical distance in the X direction in the original training volume data, the scanning angle of the XOY plane of the first coordinate system, and the motor swing angle includes: Calculate a fifth product of the second distance and the number of points in the X direction in the original training volume data, and determine the coordinate value in the X direction of the key point in the original training volume data as the ratio of the fifth product to the physical distance in the X direction in the original training volume data; Calculate a sixth product of the third included angle and the number of points in the Y direction in the original training volume data, calculate a second ratio of the sixth product to the scanning angle of the XOY plane of the first coordinate system, and determine the coordinate value in the Y direction of the key point in the original training volume data as the sum of half of the number of points in the Y direction in the original training volume data and the second ratio; Calculate a seventh product of the second included angle and the number of points in the Z direction in the original training volume data, calculate a third ratio of the seventh product to the motor swing angle, and determine the coordinate value in the Z direction of the key point in the original training volume data as the sum of half of the number of points in the Z direction in the original training volume data and the third ratio.

8. A key point recognition method, characterized in that Including: Obtain the original target volume data collected by the ultrasonic signal, and use a digital scan converter to perform signal conversion on the original target volume data to obtain the reconstructed target volume data; Input the original target volume data into the key point recognition model trained by the key point recognition model training method according to any one of claims 1 to 7 to predict the key points in the original target volume data; Map the key points in the original target volume data into the reconstructed target volume data to obtain the key points in the reconstructed target volume data.

9. A key point recognition model training device, characterized in that Including: An acquisition module for acquiring the original training volume data collected by the ultrasonic signal; wherein, the original training volume data is the volume data before signal conversion by the digital scan converter; A determination module for determining the key points labeled in the original training volume data based on the conversion relationship between the original training volume data and the reconstructed training volume data in the digital scan converter and the key points labeled in the reconstructed training volume data; A training module for training a key point recognition model using the original training volume data with labeled key points, and the trained key point recognition model is used to determine key points in the target volume data.

10. An electronic device, characterized in that Including: A memory for storing a computer program; A processor for implementing the steps of the key point recognition model training method according to any one of claims 1 to 7 or the key point recognition method according to claim 8 when executing the computer program.

11. A computer-readable storage medium, characterized in that A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps of the key point recognition model training method according to any one of claims 1 to 7 or the key point recognition method according to claim 8 are implemented.

Images