A method and system for three-dimensional localization of a medical diagnostic ultrasound position

Abstract

The application discloses a method and system for three-dimensional positioning of a medical diagnosis position by ultrasound. The method comprises the following steps: obtaining ultrasound images of different ultrasound positions and ultrasound angles; detecting features of each ultrasound image to obtain an ultrasound feature map containing information of a detected object; converting a gray value of the ultrasound image into a distance from the detected object; constructing a three-dimensional position matrix capable of representing the distance by a position and containing the ultrasound features; detecting features of a diagonal line using the three-dimensional position matrix; detecting features of multiple ultrasound positions and multiple ultrasound angles as values for judging whether the detected diagnosis position is accurate; and fusing the features of the diagonal line when detecting the diagnosis position.

Description

Technical Field

[0001] The present invention relates to the field of ultrasonic medical technology, and in particular to a method and system for three-dimensional positioning of ultrasonic medical diagnostic positions. Background Art

[0002] Ultrasound examination is a medical imaging diagnostic technology based on ultrasound waves that can visualize human tissue and thus detect lesions within it. When performing an ultrasound examination using an ultrasound probe, only two-dimensional ultrasound images are generally obtained. Screening for lesions using these images depends primarily on the experience and technique of the ultrasound physician. For two-dimensional ultrasound images, factors such as the ultrasound physician's lack of experience can cause the ultrasound examination to fail to achieve the desired effect. Therefore, three-dimensional reconstruction of two-dimensional ultrasound images is necessary to provide more intuitive, rich, and comprehensive diagnostic information about the ultrasound examination site, thereby improving the diagnostic capabilities of ultrasound examinations. Ensuring that two-dimensional ultrasound images have three-dimensional spatial position information is a necessary condition for achieving three-dimensional reconstruction of two-dimensional ultrasound images.

[0003] Currently, ultrasound probes achieve 3D positioning primarily through the use of multiple external sensors, which are expensive and bulky. Alternatively, 3D positioning methods employ the conversion of 2D ultrasound images to 3D images. However, the conversion process, combined with the lack of secondary testing of the results, can lead to inaccurate judgments. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for three-dimensional positioning of ultrasonic medical diagnostic positions to solve the above-mentioned problems existing in the prior art.

[0005] In a first aspect, an embodiment of the present invention provides a method for three-dimensional positioning of an ultrasonic medical diagnostic position, comprising:

[0006] Acquire multiple ultrasonic plane images and corresponding ultrasonic positions and ultrasonic angles; the ultrasonic position is the position of the ultrasonic device; the ultrasonic angle represents the angle between the ultrasonic device and the object under test; the multiple ultrasonic plane images represent ultrasonic images of a test object at multiple angles and multiple positions;

[0007] Detecting features of the ultrasonic plane image to obtain an ultrasonic feature map; the ultrasonic feature map represents a correlation between the grayscale value of the ultrasonic plane image and the distance from the ultrasonic device; and obtaining multiple ultrasonic feature maps corresponding to multiple ultrasonic plane images;

[0008] Based on the ultrasound position and ultrasound angle, a plurality of ultrasound feature maps are constructed into three dimensions to obtain a three-dimensional ultrasound feature matrix;

[0009] Based on the three-dimensional ultrasonic feature matrix, features of multiple ultrasonic positions and multiple ultrasonic angles are detected to obtain a diagnosis position; the diagnosis position represents a three-dimensional position of the detection object.

[0010] Optionally, based on the ultrasonic position and the ultrasonic angle, multiple ultrasonic feature maps are constructed into three dimensions to obtain a three-dimensional ultrasonic feature matrix, including:

[0011] Based on the ultrasonic feature map, a correlation relationship in the ultrasonic feature map is detected to obtain a three-dimensional position matrix;

[0012] Multiple ultrasonic feature maps correspond to obtain multiple three-dimensional position matrices;

[0013] Based on the multiple three-dimensional position matrices, mapping is performed at different ultrasonic positions and ultrasonic angles to obtain a three-dimensional ultrasonic feature matrix.

[0014] Optionally, based on the three-dimensional ultrasonic feature matrix, features of multiple ultrasonic positions and multiple ultrasonic angles are detected to obtain a diagnosis position, including:

[0015] A first three-dimensional convolution kernel of 3*3*3 is obtained;

[0016] Four diagonal lines are obtained in the three-dimensional ultrasonic feature matrix;

[0017] With a step size of 1, the center point of the first three-dimensional convolution kernel is matched with the positions on the diagonal lines, and then convolution is performed on the three-dimensional ultrasonic feature matrix to obtain diagonal convolution features;

[0018] Four diagonal lines correspond to obtain four diagonal convolution features;

[0019] Based on the four diagonal convolution features and the three-dimensional ultrasonic feature matrix, a diagnosis position is obtained.

[0020] Optionally, based on the four diagonal convolution features and the three-dimensional ultrasonic feature matrix, a diagnosis position is obtained, including:

[0021] Based on the four diagonal convolution features, a diagnosis position accuracy value is obtained;

[0022] If the diagnosis position accuracy value is greater than a diagnosis threshold value, convolution is performed based on the diagonal convolution features and the three-dimensional ultrasonic feature matrix to obtain a second feature, and a diagnosis position is obtained.

[0023] Optionally, based on the four diagonal convolution features, a diagnosis position accuracy value is obtained, including:

[0024] The four diagonal convolution features are fused to obtain a first fused feature;

[0025] acquire a first ultrasound plane image and a second ultrasound plane image; the first ultrasound plane image and the second ultrasound plane image are two different ultrasound plane images with an angle difference less than 90 degrees;

[0026] take a point of a position overlapped in the first ultrasound plane image and the second ultrasound plane image as a labeled detection point;

[0027] input the first fusion feature into a first neural network to obtain a predicted detection point;

[0028] take a shortest distance between the labeled detection point and the predicted detection point as a diagnostic position accuracy value.

[0029] Optionally, if the diagnostic position accuracy value is greater than a diagnostic threshold value, a second feature is obtained by performing convolution based on the diagonal convolution feature and the three-dimensional ultrasound feature matrix to obtain a diagnostic position, including:

[0030] acquire a second three-dimensional convolution kernel of 3*3*3;

[0031] convolve the second three-dimensional convolution kernel on the three-dimensional ultrasound feature matrix with a step of 1 to obtain an ultrasound convolution feature;

[0032] extract values of diagonal lines of the ultrasound convolution feature to obtain a plurality of second diagonal convolution features; one diagonal line corresponds to one diagonal convolution feature, and one diagonal convolution feature corresponds to one second diagonal convolution feature;

[0033] fuse the diagonal convolution feature and the corresponding second diagonal convolution feature to obtain a fusion diagonal feature;

[0034] replace values of corresponding diagonal lines in the ultrasound convolution feature with the fusion diagonal feature to obtain a fusion ultrasound convolution feature;

[0035] input the fusion ultrasound convolution feature into a three-dimensional convolution network to obtain a diagnostic position.

[0036] Optionally, the three-dimensional position matrix is obtained by detecting a correlation relationship in the ultrasound wave feature map based on the ultrasound wave feature map, including:

[0037] detect a distance between each pixel point of each ultrasound plane image and a detection object by inputting the ultrasound wave feature map into a distance detection network to obtain a detection distance matrix;

[0038] construct an initial three-dimensional matrix; a number of rows of the initial three-dimensional matrix corresponds to a length of the ultrasound wave feature map; a number of columns of the initial three-dimensional matrix corresponds to a width of the ultrasound wave feature map; a number of pages of the initial three-dimensional matrix is a fixed value; and the fixed value is less than a minimum difference value in the detection distance matrix;

[0039] an initial value of the initial three-dimensional matrix is 0;

[0040] Subtracting the difference between the maximum value and the minimum value in the detection distance matrix from a fixed value, a page distance length is obtained; the page distance length represents the range of actual distances contained in a page of 1;

[0041] Subtracting the difference between the minimum value and the value in the detection distance matrix from the page distance length, a detection page number is obtained; one position in the detection distance matrix corresponds to one detection page number;

[0042] Finding the position corresponding to the ultrasonic feature map in the initial three-dimensional matrix, setting the value corresponding to the detection page number as the value of the position corresponding to the ultrasonic feature map, and obtaining a three-dimensional position matrix.

[0043] Optionally, the plurality of three-dimensional position matrices are mapped based on different ultrasonic positions and ultrasonic angles to obtain a three-dimensional ultrasonic feature matrix, including:

[0044] Constructing a position coordinate; the position coordinate is a three-dimensional matrix;

[0045] Mapping the values in the three-dimensional position matrix in the position coordinate according to the ultrasonic position and the ultrasonic angle to obtain a plurality of mapped three-dimensional positions;

[0046] According to the x-axis, the y-axis and the z-axis parallel to the position coordinate, the plurality of mapped three-dimensional positions are divided into a position three-dimensional ultrasonic feature matrix.

[0047] Optionally, the features of the ultrasonic plane image are detected to obtain an ultrasonic feature map, including:

[0048] Obtaining a two-dimensional convolution kernel of 2*2;

[0049] Convoluting the two-dimensional convolution kernel in the ultrasonic plane image with a step of 1 to obtain an ultrasonic plane feature.

[0050] In a second aspect, an embodiment of the present application provides a system for three-dimensional positioning of an ultrasonic medical diagnosis position, including:

[0051] An acquisition module is configured to acquire a plurality of ultrasonic plane images, corresponding ultrasonic positions and ultrasonic angles; the ultrasonic position is the position of an ultrasonic device; the ultrasonic angle represents the angle between the ultrasonic device and a detection object; the plurality of ultrasonic plane images represent the ultrasonic wave images of a detection object at a plurality of angles and a plurality of positions;

[0052] A feature extraction module is configured to detect the features of the ultrasonic plane image to obtain an ultrasonic feature map; the ultrasonic feature map represents the correlation between the gray value of the ultrasonic plane image and the length from the ultrasonic device; the plurality of ultrasonic plane images correspond to obtain a plurality of ultrasonic feature maps;

[0053] a three-dimensional module, configured to construct a three-dimensional ultrasonic feature matrix from the plurality of ultrasonic feature maps based on the ultrasonic position and ultrasonic angle;

[0054] The detection module is used to detect the characteristics of multiple ultrasound positions and multiple ultrasound angles based on the three-dimensional ultrasound feature matrix to obtain a diagnostic position; the diagnostic position represents the three-dimensional position of the detection object.

[0055] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0056] The embodiments of the present invention also provide a method and system for three-dimensional positioning of ultrasonic medical diagnosis positions.

[0057] The present invention uses ultrasonic images at different ultrasonic positions and angles to detect the features of each ultrasonic image and obtain an ultrasonic feature map containing information about the detection object. The grayscale value of the ultrasonic image is converted into the distance to the detection object. A three-dimensional position matrix is ​​constructed that can represent the distance using position and contains ultrasonic features. This achieves the goal of using a feature three-dimensional position matrix to represent the distance to the detection object and the features of the detection object in ultrasonic response. The three-dimensional position matrix is ​​then used to detect the features of the diagonal line, and the predicted values ​​of two points that are different on the two-dimensional ultrasonic plane image but have the same three-dimensional position coordinates are obtained as values ​​for judging whether the detection diagnostic position is accurate. In addition, the features of the diagonal line are integrated when detecting the diagnostic position, so that the diagnostic position is more accurate, thereby achieving a technical effect of more accurate three-dimensional positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a flow chart of a method for three-dimensional positioning of ultrasonic medical diagnostic positions provided by an embodiment of the present invention.

[0059] Figure 2 This is a convolution diagram of a first three-dimensional convolution kernel in a method for three-dimensional positioning of an ultrasonic medical diagnostic position provided by an embodiment of the present invention.

[0060] Figure 3 It is a schematic diagram of position coordinates in a method for three-dimensional positioning of an ultrasonic medical diagnosis position provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The present invention will be described in detail below with reference to the accompanying drawings.

[0062] Example 1

[0063] like Figure 1 As shown, an embodiment of the present invention provides a method for three-dimensional positioning of an ultrasonic medical diagnosis position, the method comprising:

[0064] S101: Obtain a plurality of ultrasonic plane images, corresponding ultrasonic positions and ultrasonic angles; the ultrasonic position is the position of the ultrasonic device; the ultrasonic angle represents the angle between the ultrasonic device and the detection object; the plurality of ultrasonic plane images represent the ultrasonic wave images of the detection object at a plurality of angles and a plurality of positions.

[0065] In the embodiment, the ultrasonic device is a handheld ultrasonic device, and the detection object is a human body that has been detected by ultrasonic waves.

[0066] In the embodiment, the angle between the plane in which the ultrasonic device emits ultrasonic waves and the front surface of the detection object is taken as the ultrasonic angle.

[0067] In the embodiment, the position of the ultrasonic device on the geographical position coordinates is taken as the ultrasonic position. The relationship between the time length and the actual distance can be known from the ultrasonic ranging formula.

[0068] In the embodiment, the ultrasonic plane image is the image reflected by the ultrasonic wave when encountering an obstacle.

[0069] S102: Detect the features of the ultrasonic plane image to obtain an ultrasonic feature map; the ultrasonic feature map represents the correlation between the gray value of the ultrasonic plane image and the length from the ultrasonic device; a plurality of ultrasonic plane images correspondingly obtain a plurality of ultrasonic feature maps.

[0070] S103: Based on the ultrasonic position and the ultrasonic angle, construct a three-dimensional ultrasonic feature matrix by using the plurality of ultrasonic feature maps.

[0071] S104: Based on the three-dimensional ultrasonic feature matrix, detect the features of the plurality of ultrasonic positions and the plurality of ultrasonic angles to obtain a diagnosis position; the diagnosis position represents the three-dimensional position of the detection object.

[0072] Optionally, the method further includes the following steps.

[0073] Based on the ultrasonic feature map, detect the correlation in the ultrasonic feature map to obtain a three-dimensional position matrix.

[0074] The three-dimensional position matrix is used to extract the distance features in the ultrasonic plane image.

[0075] A plurality of ultrasonic feature maps correspondingly obtain a plurality of three-dimensional position matrices.

[0076] Based on the plurality of three-dimensional position matrices, map the different ultrasonic positions and ultrasonic angles to obtain a three-dimensional ultrasonic feature matrix.

[0077] Optionally, the detecting features of multiple ultrasound positions and multiple ultrasound angles based on a three-dimensional ultrasound feature matrix to obtain a diagnostic position includes:

[0078] Get the first three-dimensional convolution kernel of 3*3*3.

[0079] Acquiring four diagonal lines in the three-dimensional ultrasound feature matrix;

[0080] With a step size of 1, the center point of the first three-dimensional convolution kernel is matched with the position on the diagonal line, and then convolution is performed on the three-dimensional ultrasound feature matrix to obtain diagonal convolution features.

[0081] Among them, the convolution of the first three-dimensional convolution kernel is as follows Figure 2 shown.

[0082] The 4 diagonal lines correspond to 4 diagonal convolution features;

[0083] Based on the four diagonal convolution features and the three-dimensional ultrasound feature matrix, the diagnostic position is obtained.

[0084] Optionally, obtaining the diagnostic position based on the four diagonal convolution features and the three-dimensional ultrasound feature matrix includes:

[0085] Based on the four diagonal convolution features, a diagnostic position accuracy value is obtained; the diagnostic position accuracy value represents the accuracy of the diagnostic position detection;

[0086] If the diagnostic position accuracy value is greater than the diagnostic threshold, convolution is performed based on the diagonal convolution feature and the three-dimensional ultrasound feature matrix to obtain a second feature and obtain the diagnostic position.

[0087] In this embodiment, the diagnostic threshold is 0.9.

[0088] Optionally, obtaining the diagnostic position accuracy value based on the four diagonal convolution features includes:

[0089] The four diagonal convolution features are fused to obtain the first fused feature.

[0090] In this embodiment, fusion is performed by averaging corresponding positions;

[0091] A first ultrasonic plane image and a second ultrasonic plane image are acquired; the first ultrasonic plane image and the second ultrasonic plane image are two different ultrasonic plane images with an angle difference of less than 90 degrees.

[0092] Using points at overlapping positions in the first ultrasonic plane image and the second ultrasonic plane image as marked detection points;

[0093] The first fusion feature is input into the first neural network to obtain the predicted detection point.

[0094] The first neural network is a Fully Connected Neural Network (FCNN). The first neural network has only one output neuron.

[0095] The shortest distance between the marked detection point and the predicted detection point is taken as the diagnostic position accuracy value.

[0096] In this embodiment, the Euclidean distance is used to calculate the shortest distance.

[0097] Optionally, if the diagnostic position accuracy value is greater than a diagnostic threshold, performing convolution based on the diagonal convolution feature and the three-dimensional ultrasound feature matrix to obtain a second feature and obtain the diagnostic position, including:

[0098] Get the second three-dimensional convolution kernel of 3*3*3;

[0099] The second three-dimensional convolution kernel is convolved on the three-dimensional ultrasonic feature matrix with a step size of 1 to obtain an ultrasonic convolution feature; the ultrasonic convolution feature represents a feature of the three-dimensional position image corresponding to the constructed three-dimensional ultrasonic feature matrix.

[0100] Among them, 3D Convolutional Neural Networks (CNN) is used for convolution.

[0101] Extracting the diagonal value of the ultrasonic convolution feature to obtain multiple second diagonal convolution features; one diagonal line corresponds to one diagonal convolution feature and one second diagonal convolution feature;

[0102] The diagonal convolution feature is fused with the corresponding second diagonal convolution feature to obtain a fused diagonal feature.

[0103] In this embodiment, the fusion is performed by using a method of averaging corresponding positions.

[0104] The fused diagonal features are used to replace the corresponding diagonal values ​​in the ultrasonic convolution features to obtain the fused ultrasonic convolution features;

[0105] The fused ultrasound convolution features are input into a three-dimensional convolutional network to obtain a diagnosis position.

[0106] Wherein, the third-dimensional convolutional network is a three-dimensional convolutional neural network (3D Convolutional Neural Networks, CNN).

[0107] In this embodiment, the diagnostic position is the smallest target frame containing the detection object, including the center point, length, width, and height of the detection object.

[0108] Optionally, detecting the internal depth based on the ultrasonic characteristic graph to obtain a three-dimensional position matrix includes:

[0109] According to inputting the ultrasonic characteristic graph into a distance detection network, the distance between each pixel point of each ultrasonic plane image and the detection object is detected to obtain a detection distance matrix.

[0110] Among them, the distance detection network is a deconvolutional convolutional neural network (CNN).

[0111] Construct an initial three-dimensional matrix; the number of rows of the initial three-dimensional matrix corresponds to the length of the ultrasonic characteristic graph; the number of columns of the initial three-dimensional matrix corresponds to the width of the ultrasonic characteristic graph; the number of pages of the initial three-dimensional matrix is ​​a fixed value; the fixed value is less than the minimum difference in the detection distance matrix.

[0112] The fixed value indicates that the initial three-dimensional matrix is ​​set to have a fixed number of pages according to a method in which one page of the initial three-dimensional matrix only contains one value of the detection distance matrix.

[0113] The initial value of the initial three-dimensional matrix is ​​0.

[0114] The initial value 0 represents the distance between each position of the unknown detection object and the ultrasonic device.

[0115] The difference between the maximum value and the minimum value in the detection distance matrix is ​​divided by a fixed value to obtain a page distance length; the page distance length represents the actual distance range contained in the page of 1.

[0116] Among them, if the fixed value is 10, the minimum detection value is 4 cm, the maximum detection value is 6 cm, and the page distance length is 2 mm.

[0117] The difference between the value in the detection distance matrix and the minimum value is divided by the page distance length to obtain the number of detection pages; one position in the detection distance matrix corresponds to one detection page number.

[0118] In this embodiment, the subscripts of the page numbers of the initial three-dimensional matrix start from 0.

[0119] If the value in the detection distance matrix is ​​4.3 centimeters, then (4.3-4) / 0.2=1, and the page number corresponding to the subscript 1 in the initial three-dimensional matrix is ​​set as the detection page number.

[0120] The position corresponding to the ultrasonic characteristic pattern is found in the initial three-dimensional matrix, and the value corresponding to the detection page number is set as the value of the position corresponding to the ultrasonic characteristic pattern to obtain a three-dimensional position matrix.

[0121] Wherein, the value of 4.3 cm in the detection distance matrix is 1, column 2, then find the position of row 1, column 2, page 1 in the initial three-dimensional matrix is set to 1.

[0122] Optionally, the three-dimensional ultrasound feature matrix is obtained by mapping the plurality of three-dimensional position matrices at different ultrasound positions and ultrasound angles.

[0123] Constructing position coordinates.

[0124] In this embodiment, the position coordinates represent the coordinates of the position of the detection object relative to the actual geographical position. The position coordinates are as shown in the following table. Figure 3

[0125] According to the x-axis, y-axis and z-axis parallel to the position coordinates, the plurality of mapped three-dimensional positions are segmented into a three-dimensional ultrasound feature matrix.

[0126] The three-dimensional position point cloud is converted into a three-dimensional ultrasound feature matrix.

[0127] Optionally, the features of the ultrasound plane image are detected to obtain an ultrasound wave feature map, including:

[0128] A 2*2 two-dimensional convolution kernel is obtained.

[0129] The two-dimensional convolution kernel is convolved in the ultrasound plane image with a step size of 1 to obtain an ultrasound plane feature.

[0130] Wherein, the convolution is performed by a convolutional neural network (CNN).

[0131] Embodiment 2

[0132] Based on the above-mentioned method for three-dimensional positioning of an ultrasound medical diagnosis position, the embodiment of the present application further provides a three-dimensional positioning system for an ultrasound medical diagnosis position, which comprises an acquisition module, a feature extraction module, a three-dimensional module and a detection module.

[0133] The acquisition module is used to acquire a plurality of ultrasound plane images and corresponding ultrasound positions and ultrasound angles; the ultrasound position is the position of the ultrasound device; the ultrasound angle represents the angle between the ultrasound device and the detection object; and the plurality of ultrasound plane images represent the ultrasound wave images of a detection object at a plurality of angles and a plurality of positions.

[0134] The feature extraction module is used to detect the features of the ultrasound plane image to obtain an ultrasound wave feature map; the ultrasound wave feature map represents the correlation between the gray value of the ultrasound plane image and the length from the ultrasound device; and a plurality of ultrasound wave feature maps are obtained corresponding to the plurality of ultrasound plane images.

[0135] ​The three-dimensional module is used to construct three dimensions of the plurality of ultrasonic feature maps based on the ultrasonic position and the ultrasonic angle to obtain a three-dimensional ultrasonic feature matrix.

[0136] The detection module is used to detect the characteristics of multiple ultrasound positions and multiple ultrasound angles based on the three-dimensional ultrasound feature matrix to obtain a diagnostic position; the diagnostic position represents the three-dimensional position of the detection object.

[0137] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general-purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing this type of system. In addition, the present invention is not directed to any specific programming language. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the above description of specific languages ​​is for the purpose of disclosing the best mode of the present invention.

[0138] In the description provided herein, numerous specific details are described. However, it is understood that embodiments of the present invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

[0139] The various component embodiments of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components in the apparatus according to an embodiment of the present invention. The present invention can also be implemented as a device or device program (e.g., a computer program and a computer program product) for executing a part or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

Claims

1. A method for three-dimensional positioning of ultrasonic medical diagnostic positions, characterized in that: include: Acquire multiple ultrasonic plane images and corresponding ultrasonic positions and ultrasonic angles; the ultrasonic position is the position of the ultrasonic device; the ultrasonic angle represents the angle between the ultrasonic device and the object under test; the multiple ultrasonic plane images represent ultrasonic images of a test object at multiple angles and multiple positions; Detecting features of the ultrasonic plane image to obtain an ultrasonic feature map; the ultrasonic feature map represents a correlation between the grayscale value of the ultrasonic plane image and the distance from the ultrasonic device; and obtaining multiple ultrasonic feature maps corresponding to multiple ultrasonic plane images; Based on the ultrasound position and ultrasound angle, a plurality of ultrasound feature maps are constructed into three dimensions to obtain a three-dimensional ultrasound feature matrix; Based on the three-dimensional ultrasound feature matrix, the features of multiple ultrasound positions and multiple ultrasound angles are detected to obtain a diagnosis position; the diagnosis position represents the three-dimensional position of the detection object; The method constructs a three-dimensional ultrasonic feature matrix based on the ultrasonic position and ultrasonic angle to obtain a three-dimensional ultrasonic feature matrix, including: Based on the ultrasonic characteristic graph, detecting the correlation relationship in the ultrasonic characteristic graph to obtain a three-dimensional position matrix; Multiple three-dimensional position matrices are obtained corresponding to multiple ultrasonic feature maps; Based on the multiple three-dimensional position matrices, mapping is performed at different ultrasound positions and ultrasound angles to obtain a three-dimensional ultrasound feature matrix; The detecting the correlation relationship in the ultrasonic characteristic graph based on the ultrasonic characteristic graph to obtain a three-dimensional position matrix includes: Inputting the ultrasonic feature map into a distance detection network, detecting the distance between each pixel point of each ultrasonic plane image and the detection object, and obtaining a detection distance matrix; Constructing an initial three-dimensional matrix; the number of rows of the initial three-dimensional matrix corresponds to the length of the ultrasonic characteristic graph; the number of columns of the initial three-dimensional matrix corresponds to the width of the ultrasonic characteristic graph; the number of pages of the initial three-dimensional matrix is ​​a fixed value; the fixed value is less than the minimum difference in the detection distance matrix; The initial value of the initial three-dimensional matrix is ​​0; The difference between the maximum value and the minimum value in the detection distance matrix is ​​divided by a fixed value to obtain a page distance length; the page distance length represents the range of actual distances contained in a page with 1; The difference between the value in the detection distance matrix and the minimum value is divided by the page distance length to obtain the number of detection pages; one position in the detection distance matrix corresponds to one detection page number; Find the position corresponding to the ultrasonic characteristic pattern in the initial three-dimensional matrix, set the value corresponding to the detected page number as the value of the position corresponding to the ultrasonic characteristic pattern, and obtain a three-dimensional position matrix; The mapping is performed based on the multiple three-dimensional position matrices at different ultrasound positions and ultrasound angles to obtain a three-dimensional ultrasound feature matrix, including: Constructing position coordinates; the position coordinates are a three-dimensional matrix; According to the ultrasound position and ultrasound angle, the values ​​in the three-dimensional position matrix are mapped into the position coordinates to obtain a plurality of mapped three-dimensional positions; According to the x-axis, y-axis and z-axis parallel to the position coordinates, the multiple mapped three-dimensional positions are segmented into a three-dimensional ultrasound feature matrix.

2. The method for three-dimensional positioning of ultrasonic medical diagnostic positions according to claim 1, characterized in that: The method of detecting features of multiple ultrasound positions and multiple ultrasound angles based on a three-dimensional ultrasound feature matrix to obtain a diagnosis position includes: Get the first three-dimensional convolution kernel of 3*3*3; Acquire four diagonal lines in the three-dimensional ultrasound feature matrix; With a step size of 1, the center point of the first three-dimensional convolution kernel is matched with the position on the diagonal line, and then convolution is performed on the three-dimensional ultrasound feature matrix to obtain diagonal convolution features; The 4 diagonal lines correspond to 4 diagonal convolution features; Based on the four diagonal convolution features and the three-dimensional ultrasound feature matrix, the diagnostic position is obtained.

3. The method for three-dimensional positioning of ultrasonic medical diagnostic positions according to claim 2, characterized in that: The diagnostic position is obtained based on the four diagonal convolution features and the three-dimensional ultrasound feature matrix, including: Based on the four diagonal convolution features, the diagnostic location accuracy is obtained; If the diagnostic position accuracy value is greater than the diagnostic threshold, convolution is performed based on the diagonal convolution feature and the three-dimensional ultrasound feature matrix to obtain a second feature and obtain the diagnostic position.

4. The method for three-dimensional positioning of ultrasonic medical diagnostic positions according to claim 3, characterized in that: The diagnostic location accuracy is obtained based on the four diagonal convolution features, including: Fuse the four diagonal convolution features to obtain the first fused feature; Acquire a first ultrasonic plane image and a second ultrasonic plane image; the first ultrasonic plane image and the second ultrasonic plane image are two different ultrasonic plane images with an angle difference of less than 90 degrees; Using points at overlapping positions in the first ultrasonic plane image and the second ultrasonic plane image as marked detection points; Inputting the first fusion feature into the first neural network to obtain a predicted detection point; The shortest distance between the marked detection point and the predicted detection point is taken as the diagnostic position accuracy value.

5. The method for three-dimensional positioning of ultrasonic medical diagnosis position according to claim 3, characterized in that: If the diagnostic position accuracy value is greater than the diagnostic threshold, performing convolution based on the diagonal convolution feature and the three-dimensional ultrasound feature matrix to obtain a second feature and obtain the diagnostic position, including: Get the second three-dimensional convolution kernel of 3*3*3; Convolving the second three-dimensional convolution kernel on the three-dimensional ultrasound feature matrix with a step size of 1 to obtain an ultrasound convolution feature; Extracting the diagonal value of the ultrasonic convolution feature to obtain multiple second diagonal convolution features; one diagonal line corresponds to one diagonal convolution feature and one second diagonal convolution feature; Fuse the diagonal convolution feature with the corresponding second diagonal convolution feature to obtain a fused diagonal feature; The fused diagonal features are used to replace the corresponding diagonal values ​​in the ultrasonic convolution features to obtain the fused ultrasonic convolution features; The fused ultrasound convolution features are input into a three-dimensional convolutional network to obtain a diagnosis position.

6. The method for three-dimensional positioning of ultrasonic medical diagnostic positions according to claim 1, characterized in that: The detecting the features of the ultrasonic plane image to obtain the ultrasonic feature map includes: Get a 2*2 two-dimensional convolution kernel; The two-dimensional convolution kernel is convolved on the ultrasound plane image with a step size of 1 to obtain ultrasound plane features.

7. A three-dimensional positioning system for ultrasonic medical diagnosis, characterized in that: include: an acquisition module, configured to acquire a plurality of ultrasonic plane images and corresponding ultrasonic positions and ultrasonic angles; the ultrasonic position being the position of the ultrasonic device; the ultrasonic angle being the angle between the ultrasonic device and the object under test; and the plurality of ultrasonic plane images representing ultrasonic images of a single object under test at a plurality of angles and a plurality of positions; A feature extraction module is used to detect features of an ultrasonic plane image and obtain an ultrasonic feature map; the ultrasonic feature map represents the correlation between the grayscale value of the ultrasonic plane image and the distance from the ultrasonic device; multiple ultrasonic feature maps are obtained corresponding to multiple ultrasonic plane images; a three-dimensional module, configured to construct a three-dimensional ultrasonic feature matrix from the plurality of ultrasonic feature maps based on the ultrasonic position and the ultrasonic angle; a detection module configured to detect features of multiple ultrasound positions and multiple ultrasound angles based on a three-dimensional ultrasound feature matrix to obtain a diagnostic position representing a three-dimensional position of a detection object; The method constructs a three-dimensional ultrasonic feature matrix based on the ultrasonic position and ultrasonic angle to obtain a three-dimensional ultrasonic feature matrix, including: Based on the ultrasonic characteristic graph, detecting the correlation relationship in the ultrasonic characteristic graph to obtain a three-dimensional position matrix; Multiple three-dimensional position matrices are obtained corresponding to multiple ultrasonic feature maps; Based on the multiple three-dimensional position matrices, mapping is performed at different ultrasound positions and ultrasound angles to obtain a three-dimensional ultrasound feature matrix; The detecting the correlation relationship in the ultrasonic characteristic graph based on the ultrasonic characteristic graph to obtain a three-dimensional position matrix includes: Inputting the ultrasonic feature map into a distance detection network, detecting the distance between each pixel point of each ultrasonic plane image and the detection object, and obtaining a detection distance matrix; Constructing an initial three-dimensional matrix; the number of rows of the initial three-dimensional matrix corresponds to the length of the ultrasonic characteristic graph; the number of columns of the initial three-dimensional matrix corresponds to the width of the ultrasonic characteristic graph; the number of pages of the initial three-dimensional matrix is ​​a fixed value; the fixed value is less than the minimum difference in the detection distance matrix; The initial value of the initial three-dimensional matrix is ​​0; The difference between the maximum value and the minimum value in the detection distance matrix is ​​divided by a fixed value to obtain a page distance length; the page distance length represents the range of actual distances contained in a page with 1; The difference between the value in the detection distance matrix and the minimum value is divided by the page distance length to obtain the number of detection pages; one position in the detection distance matrix corresponds to one detection page number; Find the position corresponding to the ultrasonic characteristic pattern in the initial three-dimensional matrix, set the value corresponding to the detected page number as the value of the position corresponding to the ultrasonic characteristic pattern, and obtain a three-dimensional position matrix; The mapping is performed based on the multiple three-dimensional position matrices at different ultrasound positions and ultrasound angles to obtain a three-dimensional ultrasound feature matrix, including: Constructing position coordinates; the position coordinates are a three-dimensional matrix; According to the ultrasound position and ultrasound angle, the values ​​in the three-dimensional position matrix are mapped into the position coordinates to obtain a plurality of mapped three-dimensional positions; According to the x-axis, y-axis and z-axis parallel to the position coordinates, the multiple mapped three-dimensional positions are segmented into a three-dimensional ultrasound feature matrix.

Images