Mammary gland volume ultrasonic focus positioning and grading method, medium and terminal

Abstract

The invention is suitable for the technical field of intelligent diagnosis, and relates to a breast volume ultrasonic focus positioning and grading method, a medium and a terminal.The method comprises the steps that S10, after a U-shaped network is prepared, normalization processing is conducted on coronal plane two-dimensional image sequence data through Z-score standardization operation; s20, taking the normalized 3D image data as the input of a U-point cloud neural network model, carrying out the training of the U-point cloud neural network model, and outputting a probability distribution result; s30, segmenting a breast volume ultrasound focus position according to a probability distribution result, storing a three-dimensional image of the focus position, and taking the three-dimensional image as input data of the point cloud; and S40, processing the data input in the step S30 by using a point cloud network, learning global and local features from a point set, and realizing a 3D classification task and a segmentation task of the breast volume ultrasound focus. The accuracy and efficiency of breast focus detection are effectively improved, and the workload of doctors is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent diagnosis, and in particular relates to a method, medium and terminal for locating and grading breast volume ultrasonic lesions. Background Art

[0002] With the rapid development of medical imaging technology, early diagnosis and treatment of breast cancer have become increasingly important. Breast Volume Ultrasound (BVU), as a non-invasive, radiation-free examination method, has become an important tool for breast cancer screening and diagnosis. BVU can provide high-resolution three-dimensional images of breast tissue, which helps doctors observe breast structure, detect lumps, cysts and other abnormalities. However, due to the complexity of breast tissue and the instability of ultrasound images, it is still a challenge to accurately identify and grade lesions from BVU images.

[0003] Traditional lesion detection methods mainly rely on doctors' experience and manual operations, which is not only time-consuming and subjective, but may also lead to missed or misdiagnosis due to doctors' fatigue or lack of experience, resulting in low accuracy of lesion detection; in addition, with the increase in the number of breast cancer patients, doctors' workload has also increased, which further exacerbates the difficulty of diagnosis. In recent years, deep learning technology has made significant progress in the fields of image processing and pattern recognition, especially in the increasingly widespread application of medical image analysis. However, most existing deep learning research focuses on two-dimensional images, and the processing of three-dimensional ultrasound images still faces challenges. The patent with announcement number CN113239951B provides a classification method, device and storage medium of ultrasonic breast lesions, including obtaining target ultrasonic breast information to be classified, wherein the target ultrasonic breast information is a target ultrasonic breast image or a target ultrasonic breast video, wherein the target ultrasonic breast video includes at least two frames of target ultrasonic breast images; determining whether the target ultrasonic breast image is a breast image corresponding to a breast section through a breast tissue evaluation network; if the target ultrasonic breast image is a breast image corresponding to the breast section, identifying the lesion area in the target ultrasonic breast image; dividing the identified lesion area into n1*n2 data blocks; converting each data block into p1 *p2*c dimensional vector data; wherein n1=H / p1, n2=W / p2; H is the height of the input image, W is the width of the input image, p1 is the height of the segmented data block, and p2 is the width of the segmented data block; the vector data corresponding to the n1*n2 data blocks are merged to obtain a two-dimensional data matrix of n1n2×p1p2c; according to the position of each data block in the target ultrasound breast image, a position coding vector corresponding to the position is generated, and the position coding vector is added to the two-dimensional data matrix to obtain a data matrix to be processed; the data matrix to be processed is input into a pre-trained image classification network to obtain the lesion property classification corresponding to the target ultrasound breast image. This patent also studies breast lesions based on two-dimensional images, and has the same drawbacks as the prior art.

[0004] Therefore, how to improve the accuracy and efficiency of breast lesion detection and reduce the workload of doctors is an urgent problem to be solved by people in this technical field. Summary of the invention

[0005] In view of the deficiencies of the prior art, the purpose of the present invention is to provide a method for locating and grading breast volume ultrasonic lesions to solve the problems of low accuracy and efficiency of lesion detection in the prior art; in addition, the present invention also provides a breast volume ultrasonic lesion locating and grading medium and terminal.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for locating and grading breast volume ultrasound lesions, comprising the following steps:

[0008] S10, when the U-shaped network is ready, the coronal two-dimensional image sequence data is normalized using the Z-score normalization operation;

[0009] S20, using the normalized 3D image data as input of the U-point cloud neural network model, training the U-point cloud neural network model, and outputting a probability distribution result;

[0010] S30, segmenting the breast volume ultrasound lesion location according to the probability distribution result, saving the three-dimensional image of the lesion location, and using it as input data of the point cloud;

[0011] S40, using a point cloud network to process the data input in step S30, learning global and local features from the point set, and achieving 3D classification and segmentation tasks of breast volume ultrasound lesions.

[0012] Furthermore, in step S10, the data is normalized using the Z-score standardization operation. Assuming that the original data is S t+i , the normalization operation is expressed as:

[0013]

[0014] Wherein, μ and δ are the mean and standard deviation pre-calculated in step S10, and the data satisfies the normal distribution after being normalized by the Z-score standardization operation, with a mean of 0 and a standard deviation of 1.

[0015] Furthermore, in step S20, the U-point cloud neural network model includes an input layer, a U-type network module, a point cloud module and an output layer.

[0016] Furthermore, the U-shaped network module includes an analysis path and a synthesis path, each of which contains four resolution levels. In the analysis path, each layer includes two 3×3×3 convolutional layers, each convolutional layer is followed by a ReLU activation function, followed by a 2×2×2 maximum pooling layer with a pooling step of 2, which is used to extract high-level features.

[0017] Furthermore, in the synthesis path, each layer is first a 2×2×2 upconvolution layer with a stride of 2, which is used to restore the spatial resolution of the feature map. The upconvolution layer is followed by two 3×3×3 convolution layers, each of which is followed by a ReLU activation function. After each upsampling, the obtained feature image is concatenated with the corresponding downsampled feature image. The final softmax layer converts the output image into a segmentation result of the same size as the input image, and the segmentation result is a three-dimensional probability map, in which each voxel is assigned a probability distribution.

[0018] Furthermore, in step S30, a threshold selection function is provided according to the output probability distribution result for segmenting the lesion position, data whose probability exceeds the threshold is recorded as 1, and data whose probability does not exceed the threshold is recorded as 0, and the three-dimensional position of the image of the data recorded as 1 is saved and used as the input data of the point cloud.

[0019] Furthermore, the specific steps of step S40 are as follows:

[0020] S401, the point cloud network receives an unordered point set as input, each point is represented by its three-dimensional coordinates;

[0021] S402, the point cloud network aligns the input point cloud data through an input transformation network, the input transformation network includes several convolutional layers and fully connected layers, and finally outputs a 3×3 transformation matrix for performing affine transformation on the coordinates of the input point cloud;

[0022] S403, the point cloud data output by the transformation matrix enters a shared multi-layer perceptron, which includes several convolution layers, and each point undergoes the same convolution operation to extract local features;

[0023] S404, the point cloud network uses a feature transformation network to align the extracted local features. The feature transformation network includes several convolutional layers and fully connected layers, and finally outputs a 64×64 transformation matrix for transforming the feature vector;

[0024] S405, the point cloud network uses the maximum pooling layer to extract the global feature vector from the local features of all points for subsequent classification and segmentation tasks.

[0025] Furthermore, in step S405, for the classification task, the point cloud will extract the global feature vector through the fully connected layer of the classification branch, and finally output the probability of each category. The classification branch is composed of several fully connected layers. The fully connected layer processes the global feature vector to generate the final classification result, so as to classify the entire point cloud and identify the category to which the point cloud belongs; for the segmentation task, the point cloud will concatenate the global feature vector with the local feature of each point, and then pass the convolution layer of the segmentation branch to output the segmentation result of each point. The segmentation branch is composed of several convolution layers. The convolution layer processes the concatenated feature vector to generate a segmentation label for each point, so as to perform fine-grained segmentation on each point in the point cloud and identify the category to which each point belongs.

[0026] In a second aspect, the present invention further provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0027] In a third aspect, the present invention further provides an electronic terminal, comprising: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes the method as described above.

[0028] Compared with the prior art, the breast volume ultrasound lesion localization and grading method, medium and terminal provided by the present invention have at least the following beneficial effects:

[0029] Traditional lesion detection methods mainly rely on the doctor's experience and manual operation, which is not only time-consuming and subjective, but also may lead to missed diagnosis or misdiagnosis due to doctor fatigue or lack of experience, resulting in low accuracy of lesion detection; in addition, with the increase in the number of breast cancer patients, the workload of doctors also increases, which further aggravates the difficulty of diagnosis. The present invention uses a 3D U-Net network to perform three-dimensional segmentation, can accurately identify and locate the lesion area, reduce the risk of missed diagnosis and misdiagnosis, and use a point cloud network to accurately locate and grade the lesion area, thereby improving the overall accuracy of lesion detection. The automated detection process significantly reduces the doctor's manual operation and subjective judgment, improves the diagnostic efficiency, enables doctors to devote more time and energy to the treatment and management of patients, and reduces the workload. In addition, when processing three-dimensional medical image data, the present invention can make full use of the stereoscopic information of the image, provide more comprehensive lesion features, and provide strong support for the early diagnosis and treatment of breast cancer. In summary, the present invention solves the shortcomings of traditional methods that rely on doctor experience and manual operation through deep learning technology, especially in the application of three-dimensional segmentation and point cloud classification, and provides an effective tool for the early diagnosis and treatment of breast cancer. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the scheme of the present invention, a brief introduction is given below to the figures required for use in the description of the embodiments. Obviously, the figures described below are some embodiments of the present invention. For ordinary technicians in this field, other figures can be obtained based on these figures without paying any creative work.

[0031] Figure 1 A flow chart of a method for locating and grading breast volume ultrasound lesions provided by an embodiment of the present invention;

[0032] Figure 2 A schematic diagram of the working process of a U-point cloud neural network model of a breast volume ultrasound lesion localization and grading method provided by an embodiment of the present invention;

[0033] Figure 3 A U-shaped network architecture diagram of a breast volume ultrasound lesion localization and grading method provided in an embodiment of the present invention;

[0034] Figure 4 A data conversion flow chart of a method for locating and grading breast volume ultrasound lesions provided by an embodiment of the present invention;

[0035] Figure 5 A point cloud network schematic diagram of a breast volume ultrasound lesion localization and grading method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. The preferred embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0038] The present invention provides a method for locating and grading breast volume ultrasound lesions, which is applied in the process of detecting breast volume ultrasound lesions. The method for locating and grading breast volume ultrasound lesions comprises the following steps:

[0039] S10. When the U-type network is ready, the coronal two-dimensional image sequence data is normalized using the Z-score standardization operation; S20. The normalized 3D image data is used as the input of the U-point cloud neural network model, the U-point cloud neural network model is trained, and the probability distribution result is output; S30. The breast volume ultrasound lesion location is segmented according to the probability distribution result, and the three-dimensional image of the lesion location is saved and used as the input data of the point cloud; S40. The data input in step S30 is processed using the point cloud network, and the global and local features are learned from the point set to achieve the 3D classification task and segmentation task of the breast volume ultrasound lesion.

[0040] The present invention effectively improves the accuracy and efficiency of breast lesion detection and reduces the workload of doctors.

[0041] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0042] The present invention provides a method for locating and grading breast volume ultrasound lesions, which is applied in the detection process of breast volume ultrasound lesions. Figures 1 to 5 In this embodiment, the breast volume ultrasound lesion localization and grading method includes the following steps:

[0043] S10. When the U-shaped network is ready, the coronal two-dimensional image sequence data is normalized using the Z-score normalization operation.

[0044] Specifically, in this embodiment, when the U-shaped network is ready, the coronal plane 2D image sequence data is normalized using the Z-score normalization operation. After normalization, the data eliminates the dimension difference, and the convergence speed is faster during training. The normalized 3D image data is used as the input of the model, and the data is normalized using the Z-score normalization operation. After normalization, the data eliminates the dimension, and the convergence speed is faster during training. Assuming that the original data is S t+i , the normalization operation can be expressed as:

[0045]

[0046] Wherein, μ and δ are the mean and standard deviation pre-calculated in step S10. After the data is normalized by the Z-score standardization operation, it satisfies the normal distribution, with a mean of 0 and a standard deviation of 1.

[0047] S20, using the normalized 3D image data as the input of the U-point cloud model, training the U-point cloud model, and outputting the probability distribution result.

[0048] Specifically, in this embodiment, the processed data is input into a U-point cloud model, which consists of an input, a U-type network module, a point cloud module and an output.

[0049] Furthermore, in this embodiment, the U-shaped network architecture is mainly divided into an analysis path (i.e., the encoder part) and a synthesis path (i.e., the decoder part), each of which includes four resolution levels. The point cloud data is a data set composed of a large number of three-dimensional coordinate points, which can describe the geometric shape and boundary information of the target area in detail, and is used for subsequent point cloud feature extraction to provide the geometric features of the target area.

[0050] Furthermore, in this embodiment, in the analysis path, each layer consists of two 3×3×3 convolutional layers, each of which is followed by a ReLU activation function, followed by a 2×2×2 maximum pooling layer with a pooling step of 2 for extracting high-level features. This series of operations helps capture the global characteristics of the input data.

[0051] The convolutional layer is:

[0052]

[0053] in, Represents the convolution operation, x is the input, W is the convolution kernel, b is the bias term, y is the output, and ReLU is the activation function, which is defined as:

[0054] ReLU(z)=max(0,z).

[0055] The maximum pooling layer is:

[0056] Assume that the shape of the input data is H×W×C, where H is the height, W is the width, C is the number of channels, the pooling kernel size is k×k, and the step size is s. Divide the input feature map X into several non-overlapping local areas according to the size k×k of the pooling kernel and the step size s. The shape of each local area is k×k.

[0057] Select the maximum value in each local area to form a new feature map Y. The height H′ and width W′ of the new feature map can be calculated by the following formula:

[0058]

[0059] The shape of the output feature map Y is H′×W′×C, and each element Y i，j，k is the maximum value of the corresponding local area in the input feature map X:

[0060]

[0061] Among them, region(i,j) represents a k×k local region centered at position (i,j).

[0062] Furthermore, in this embodiment, in the synthesis path, each layer is first a 2×2×2 up-convolution layer (also called a deconvolution layer or a transposed convolution layer) with a step size of 2, which is used to restore the spatial resolution of the feature map. The up-convolution layer is followed by two 3×3×3 convolution layers, and each convolution layer is also followed by a ReLU activation function. After each upsampling, the obtained feature image is concatenated with the corresponding down-sampled feature image. The final Softmax layer of the network converts the output image into a segmentation result of the same size as the input image, and the segmentation result is a three-dimensional probability map in which each voxel is assigned a probability distribution.

[0063] Data z enters the upper convolution layer, and the convolution formula is:

[0064]

[0065] Among them, ConvTranspose is a transposed convolution operation, which is used to restore the spatial resolution of the feature map.

[0066] After the upper convolution is completed, the data enters the convolution layer for calculation. The convolution layer formula is:

[0067]

[0068] Then the upsampled feature map y u And the upsampled feature map y d The two feature maps are concatenated in the channel dimension through the Concatenate operation.

[0069] Finally, the final output P is obtained through the Softmax layer, which is the probability distribution result.

[0070] The Softmax function formula is:

[0071]

[0072] Furthermore, in this embodiment, after receiving the output result P, a threshold selection function is provided to segment the lesion location, and the data with a probability exceeding the threshold is recorded as 1 and vice versa. The three-dimensional position of the image of the data with a probability of 1 is then saved and passed to the next level point cloud network as input data of the point cloud.

[0073] S30, segmenting the breast volume ultrasound lesion location according to the probability distribution result, saving the three-dimensional image of the lesion location, and using it as input data of the point cloud.

[0074] S40, using the point cloud network to process the data input in step S30, learning global and local features from the point set, and realizing the 3D classification task and segmentation task of breast volume ultrasound lesions.

[0075] Specifically, in this embodiment, the point cloud network is a deep learning model specifically used to process point cloud data. It can directly learn global and local features from point sets for 3D classification and segmentation tasks. In this model, only the classification function of the point cloud network is used. The point cloud receives an unordered point set as input, and each point is represented by its three-dimensional coordinates (x, y, z). The point cloud can be regarded as a set of discrete points in three-dimensional space, for example: {P i |i=1,...,n}, where each P i =(x i ,y i , z i ).

[0076] Furthermore, in this embodiment, the specific steps of step S40 are as follows:

[0077] S401, the point cloud network receives an unordered point set as input, each point is represented by its three-dimensional coordinates.

[0078] S402. The point cloud network aligns the input point cloud data through the input transformation network (Input Transform Net). The input transformation network consists of several convolutional layers and fully connected layers, and finally outputs a 3×3 transformation matrix. This transformation matrix is ​​used to perform affine transformation on the coordinates of the input point cloud, making the point cloud more regular and symmetrical in space. By aligning the input point cloud, the model can better capture the geometric structure of the point cloud.

[0079] S403, the point cloud data output by the transformation matrix enters the shared multi-layer perceptron (Shared MLP) to extract local features for each point. The shared multi-layer perceptron consists of several convolutional layers. Each point undergoes the same convolution operation to extract local features. These local features retain the local geometric information of the point cloud and provide a basis for further feature extraction. In order to further enhance the alignment of features, the point cloud also uses a feature transformation network (Feature Transform Net) to align the extracted local features. The feature transformation network also consists of several convolutional layers and fully connected layers, and finally outputs a 64×64 transformation matrix. This transformation matrix is ​​used to transform the feature vector to make the features more regular and symmetrical in the feature space.

[0080] S404. After extracting local features and aligning them, the point cloud uses the maximum pooling layer (Max Pooling) to extract the global feature vector from the local features of all points. The maximum pooling layer aggregates the local features into a global feature vector by performing the maximum value operation on the features of all points. This global feature vector represents the global characteristics of the entire point cloud and can be used for subsequent classification and segmentation tasks.

[0081] S405, the point cloud network uses the maximum pooling layer to extract the global feature vector from the local features of all points for subsequent classification and segmentation tasks.

[0082] Specifically, for classification tasks, the point cloud extracts the global feature vector through the fully connected layer of the classification branch, and finally outputs the probability of each category. The classification branch consists of several fully connected layers, which process the global feature vector to generate the final classification result. In this way, the point cloud can classify the entire point cloud and identify the category to which the point cloud belongs. For segmentation tasks, the point cloud concatenates the global feature vector with the local feature of each point, and then outputs the segmentation result of each point through the convolution layer of the segmentation branch. The segmentation branch consists of several convolution layers, which process the concatenated feature vector to generate the segmentation label of each point. In this way, the point cloud can perform fine-grained segmentation on each point in the point cloud and identify the category to which each point belongs.

[0083] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, any one of the methods in the present embodiment is implemented.

[0084] An embodiment of the present invention further provides an electronic terminal, comprising: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes any one of the methods in this embodiment.

[0085] The computer-readable storage medium in this embodiment can be understood by ordinary technicians in this field: all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to the computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other media that can store program codes.

[0086] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication with each other. The memory is used to store computer programs, the communication interface is used to communicate, and the processor and the transceiver are used to run computer programs so that the electronic terminal executes each step of the above method.

[0087] The breast volume ultrasound lesion localization and grading method, medium and terminal described in the above embodiment, compared with the prior art, the traditional lesion detection method mainly relies on the doctor's experience and manual operation, which is not only time-consuming and subjective, but also may lead to missed diagnosis or misdiagnosis due to doctor fatigue or lack of experience, resulting in low accuracy of lesion detection; in addition, with the increase in the number of breast cancer patients, the workload of doctors also increases, which further aggravates the difficulty of diagnosis. The present invention performs three-dimensional segmentation through a 3D U-Net network, can accurately identify and locate the lesion area, reduces the risk of missed diagnosis and misdiagnosis, and uses a point cloud network to accurately locate and grade the lesion area, thereby improving the overall accuracy of lesion detection. The automated detection process significantly reduces the doctor's manual operation and subjective judgment, improves the diagnostic efficiency, enables the doctor to devote more time and energy to the treatment and management of the patient, and reduces the workload. In addition, when processing three-dimensional medical image data, the present invention can make full use of the stereoscopic information of the image, provide more comprehensive lesion characteristics, and provide strong support for the early diagnosis and treatment of breast cancer. In summary, the present invention solves the shortcomings of traditional methods that rely on doctor experience and manual operation through the application of deep learning technology, especially in three-dimensional segmentation and point cloud classification, and provides an effective tool for early diagnosis and treatment of breast cancer.

[0088] Obviously, the embodiments described above are only preferred embodiments of the present invention, rather than all embodiments. The preferred embodiments of the present invention are shown in the accompanying drawings, but they do not limit the patent scope of the present invention. The present invention can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive. Although the present invention has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific embodiments, or to perform equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of the present invention, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present invention.

Claims

1. A method for locating and grading breast volume ultrasound lesions, characterized in that: The following steps are involved: S10, when the U-shaped network is ready, the coronal two-dimensional image sequence data is normalized using the Z-score normalization operation; S20, using the normalized 3D image data as input of the U-point cloud neural network model, training the U-point cloud neural network model, and outputting a probability distribution result; S30, segmenting the breast volume ultrasound lesion location according to the probability distribution result, saving the three-dimensional image of the lesion location, and using it as input data of the point cloud; S40, using a point cloud network to process the data input in step S30, learning global and local features from the point set, and achieving 3D classification and segmentation tasks of breast volume ultrasound lesions.

2. A method for locating and grading breast volume ultrasound lesions according to claim 1, characterized in that: In step S10, the data is normalized using the Z-score standardization operation. Assuming that the original data is S t+i , the normalization operation is expressed as: Wherein, μ and δ are the mean and standard deviation pre-calculated in step S10, and the data satisfies the normal distribution after being normalized by the Z-score standardization operation, with a mean of 0 and a standard deviation of 1.

3. A method for locating and grading breast volume ultrasound lesions according to claim 2, characterized in that: In step S20, the U-point cloud neural network model includes an input layer, a U-type network module, a point cloud module and an output layer.

4. A method for locating and grading breast volume ultrasound lesions according to claim 3, characterized in that: The U-shaped network module includes an analysis path and a synthesis path, each of which contains four resolution levels. In the analysis path, each layer includes two 3×3×3 convolutional layers, each convolutional layer is followed by a ReLU activation function, followed by a 2×2×2 maximum pooling layer with a pooling step of 2, which is used to extract high-level features.

5. A method for locating and grading breast volume ultrasound lesions according to claim 4, characterized in that: In the synthesis path, each layer is first a 2×2×2 upconvolution layer with a stride of 2, which is used to restore the spatial resolution of the feature map. The upconvolution layer is followed by two 3×3×3 convolution layers, each of which is followed by a ReLU activation function. After each upsampling, the obtained feature image is concatenated with the corresponding downsampled feature image. The final softmax layer converts the output image into a segmentation result of the same size as the input image, and the segmentation result is a three-dimensional probability map in which each voxel is assigned a probability distribution.

6. A method for locating and grading breast volume ultrasound lesions according to claim 5, characterized in that: In step S30, a threshold selection function is provided for segmenting the lesion position according to the output probability distribution result, data with probability exceeding the threshold is recorded as 1, and data with probability not exceeding the threshold is recorded as 0, and the three-dimensional position of the image of the data recorded as 1 is saved and used as the input data of the point cloud.

7. A method for locating and grading breast volume ultrasound lesions according to claim 1, characterized in that: The specific steps of step S40 are as follows: S401, the point cloud network receives an unordered point set as input, each point is represented by its three-dimensional coordinates; S402, the point cloud network aligns the input point cloud data through an input transformation network, the input transformation network includes several convolutional layers and fully connected layers, and finally outputs a 3×3 transformation matrix for performing affine transformation on the coordinates of the input point cloud; S403, the point cloud data output by the transformation matrix enters a shared multi-layer perceptron, which includes several convolution layers, and each point undergoes the same convolution operation to extract local features; S404, the point cloud network uses a feature transformation network to align the extracted local features. The feature transformation network includes several convolutional layers and fully connected layers, and finally outputs a 64×64 transformation matrix for transforming the feature vector; S405, the point cloud network uses the maximum pooling layer to extract the global feature vector from the local features of all points for subsequent classification and segmentation tasks.

8. A method for locating and grading breast volume ultrasound lesions according to claim 7, characterized in that: In the step S405, for the classification task, the point cloud will extract the global feature vector through the fully connected layer of the classification branch, and finally output the probability of each category. The classification branch is composed of several fully connected layers. The fully connected layer processes the global feature vector to generate the final classification result, so as to classify the entire point cloud and identify the category to which the point cloud belongs; for the segmentation task, the point cloud will concatenate the global feature vector with the local feature of each point, and then pass the convolution layer of the segmentation branch to output the segmentation result of each point. The segmentation branch is composed of several convolution layers. The convolution layer processes the concatenated feature vector to generate a segmentation label for each point, so as to perform fine-grained segmentation on each point in the point cloud and identify the category to which each point belongs.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

10. An electronic terminal, characterized in that: include: Processor and memory; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes the method according to any one of claims 1 to 8.

Images