Breast ultrasound image acquisition method, device, equipment and storage medium

By judging the similarity and signal-to-noise ratio of breast ultrasound image acquisition methods and using the object detection model, the problem of image quality dependence on doctors' experience in the prior art is solved, and the standardized acquisition and efficient evaluation of breast ultrasound images are realized, which improves the The unity and normative nature of medical resources.

CN114612462BActive Publication Date: 2024-11-01CHENGDU AIJIAFEISHITE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210299134.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-25
Publication Date
2024-11-01
Estimated Expiration
2042-03-25

AI Technical Summary

Technical Problem

The quality of existing breast ultrasound image acquisition is easily affected by the personal experience of doctors, making it difficult to guarantee image quality, especially in small and medium-sized hospitals.

Method used

A breast ultrasound image acquisition method is used to compare the similarity and signal-to-noise ratio between the current frame and the previous frame image. If the preset range is met, the target detection model completed by the training is input for lesion detection. If the lesion is detected and The image is saved in the center of the image, and the perceived hashing algorithm and deep neural network are used to reduce the influence of subjective factors.

Benefits of technology

The objective evaluation and standardized acquisition of breast ultrasound image quality is achieved, which reduces the influence of doctors' personal experience, improves the stability and efficiency of image acquisition, and promotes the unified normative of medical resources.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114612462B_ABST
    Figure CN114612462B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of image processing technology, and discloses a breast ultrasound image acquisition method, device, equipment and storage medium, the method comprising: acquiring a first ultrasound image of the current frame; calculating the similarity between the first ultrasound image and the second ultrasound image of the previous frame, and comparing the similarity with a first preset range; if the similarity is within the first preset range, calculating the signal-to-noise ratio of the first ultrasound image, and comparing the signal-to-noise ratio with a second preset range; if the signal-to-noise ratio is within the second preset range, inputting the first ultrasound image into a trained target detection model to obtain a detection result; if the detection result indicates that there is a lesion in the first ultrasound image and the lesion is located in the center of the image, then saving the first ultrasound image. The present invention solves the problem that the quality of ultrasound images acquired in existing breast ultrasound examination operations is easily affected by the personal experience level of doctors, resulting in the difficulty in ensuring the quality of ultrasound images.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a breast ultrasound image acquisition method, device, equipment and storage medium. Background Art

[0002] Breast cancer (BC) is a leading cause of cancer-related death in women worldwide. It is a malignant tumor that develops in the female breast. With the Westernization of social lifestyles and increased physical and mental stress, the incidence of breast cancer is increasing with age. According to the National Cancer Center's 2020 Annual Work Report and a 2021 analysis of cancer mortality data over the past 30 years, cancer incidence in China is generally on the rise, with a significant increase in women, with breast cancer ranking first among cancers.

[0003] Meanwhile, breast cancer mortality rates in Europe and the United States have been declining over the past decade. Research indicates that early diagnosis and treatment, primarily through breast cancer screenings, are the primary drivers of this decline. Therefore, addressing breast cancer, a major health threat to women, is crucial. Furthermore, data shows that most breast cancer patients seek medical attention only in the middle or late stages of the disease, when symptoms worsen. At this stage, the disease is more likely to metastasize to multiple organs, making treatment less effective.

[0004] Breast cancer is currently one of two malignancies recognized by the World Health Organization for which early diagnosis can reduce mortality. Breast ultrasound is a widely used imaging method for early breast cancer diagnosis. It is non-invasive, safe, effective, and relatively inexpensive. This method allows for the early initiation of comprehensive treatment for breast cancer, effectively reducing mortality.

[0005] However, the quality of existing ultrasound image acquisition relies on the radiologist's personal experience, which is easily affected by the physician's subjective emotions, fatigue, and clinical ability, and lacks a relatively objective evaluation system. Furthermore, for some small and medium-sized hospitals with inexperienced physicians, the image quality of ultrasound reports is even more difficult to control. Summary of the Invention

[0006] Based on the above technical problems, the present invention provides a breast ultrasound image acquisition method, device, equipment and storage medium, which solves the problem that the quality of the ultrasound images collected in the existing breast ultrasound examination operation is easily affected by the doctor's personal experience level, resulting in the difficulty in ensuring the quality of the ultrasound image.

[0007] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows:

[0008] A breast ultrasound image acquisition method, comprising:

[0009] Acquire a first ultrasound image of a current frame;

[0010] Calculating the similarity between the first ultrasonic image and the second ultrasonic image of the previous frame, and comparing the similarity with a first preset range;

[0011] If the similarity is within a first preset range, calculating a signal-to-noise ratio of the first ultrasound image and comparing the signal-to-noise ratio with a second preset range;

[0012] If the signal-to-noise ratio is within a second preset range, inputting the first ultrasound image into the trained target detection model to obtain a detection result;

[0013] If the detection result indicates that there is a lesion in the first ultrasound image and the lesion is located in the center of the image, the first ultrasound image is saved.

[0014] A breast ultrasound image acquisition device, comprising:

[0015] An image acquisition module, the image acquisition module is used to acquire a first ultrasound image of a current frame;

[0016] A similarity calculation module is used to calculate the similarity between the first ultrasonic image and the second ultrasonic image of the previous frame, and compare the similarity with a first preset range;

[0017] a signal-to-noise ratio calculation module, the signal-to-noise ratio calculation module being configured to calculate the signal-to-noise ratio of the first ultrasound image if the similarity is within a first preset range, and compare the signal-to-noise ratio with a second preset range;

[0018] A target detection module, the target detection module is used to input the first ultrasound image into a trained target detection model to obtain a detection result if the signal-to-noise ratio is within a second preset range;

[0019] The image acquisition module is used to save the first ultrasound image if the detection result shows that there is a lesion in the first ultrasound image and the lesion is located in the center of the image.

[0020] A computer device includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of the above-mentioned breast ultrasound image acquisition method.

[0021] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the processor executes the steps of the above-mentioned breast ultrasound image acquisition method.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] The present invention provides a method for acquiring breast ultrasound images. Through a series of ultrasound image quality judgment operations, the influence of the doctor's personal experience level can be eliminated, and ultimately breast ultrasound images with stable and excellent quality can be obtained for use by medical personnel in subsequent diagnosis and treatment stages.

[0024] In addition, by automatically completing the quality assessment, acquisition and storage of breast ultrasound images, the efficiency of doctors can be improved to a certain extent.

[0025] In addition, by adopting a unified breast ultrasound image acquisition standard and deploying such a standardized breast ultrasound acquisition system in hospitals of different levels, it will help promote the unification and standardization of standards and reduce the unfairness of medical resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings described herein are intended to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an improper limitation of the present invention. Among them:

[0027] Figure 1 Schematic diagram of the process of breast ultrasound image acquisition method.

[0028] Figure 2 Schematic diagram of the process of training the target detection model.

[0029] Figure 3 A flowchart of inputting a first ultrasound image into a trained target detection model to obtain a detection result. DETAILED DESCRIPTION

[0030] To make the purpose, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be clearly and completely described below in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0031] Unless otherwise defined, the technical or scientific terms used in this application should have the usual meanings understood by people with ordinary skills in the field to which this application belongs. The "first", "second" and similar words used in this application do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0032] See Figure 1 In some embodiments, a breast ultrasound image acquisition method includes:

[0033] S101, acquiring a first ultrasound image of a current frame;

[0034] S102, calculating the similarity between the first ultrasonic image and the second ultrasonic image of the previous frame, and comparing the similarity with a first preset range;

[0035] Specifically, the perceptual hashing algorithm can be used to calculate the similarity between the first and second ultrasound images. The basic principle of this algorithm is to generate a fingerprint (in string format) for each image and then compare the similarity between the fingerprints to calculate the similarity between the images. The more similar the fingerprints of the two images are, the more similar they are considered to be.

[0036] The implementation steps of perceptual hashing mainly consist of the following aspects:

[0037] First, reduce the image size. To optimize file size and facilitate transmission, while removing differences in image sizes and proportions, retaining only basic information such as structure and brightness, the image is reduced to a size of 32x32.

[0038] Secondly, convert the image to grayscale. The RGB in the original image can provide very little information, so directly converting the reduced image to grayscale facilitates subsequent calculations.

[0039] Next, the DCT (discrete cosine transform) is calculated. Research has shown that, in addition to possessing general orthogonal transform properties, the basis vectors of the DCT matrix closely resemble the eigenvectors of the Toeplitz matrix, which embodies the relevant characteristics of human speech and image signals. It is similar to the discrete Fourier transform (DFT), but uses only real numbers. The DFT requires complex operations, and while the FFT (fast Fourier transform) can improve computational speed, it is inconvenient for image coding, especially in real-time processing. While the DFT is rarely used in practical image communication systems, it does have theoretical guidance. Based on the properties of the DFT, which states that the Fourier transform of a real even function contains only real cosine terms, a real-domain transform—the discrete cosine transform (DCT)—is constructed.

[0040] The discrete cosine transform (DCT) is often used in signal and image processing for lossy data compression of signals and images (both still and moving images). This is due to the DCT's strong "energy concentration" property: the energy of most natural signals (including sound and images) is concentrated in the low-frequency portion of the DCT. This low-frequency portion determines the overall structure, while the high-frequency portion refines the details. Therefore, when comparing images for similarity, we focus more on the differences in the mid- and low-frequency portions. Therefore, to convert images from the spatial domain to the frequency domain, the DCT is used. For the scanned image, only the 8*8 matrix in the upper left corner is taken; this portion represents the lowest frequencies in the image.

[0041] There are 8 forms of one-dimensional DCT, the most commonly used forms are as follows:

[0042]

[0043] Among them, F(u) represents the coefficient after DCT transformation, c(u) represents the compensation coefficient, i represents the traversal pointer, f(i) represents the original signal, and u represents the generalized frequency variable;

[0044] The specific formula of c(u) is as follows:

[0045]

[0046] Wherein, N represents the total number of one-dimensional elements, that is, 8, the c(u) coefficient makes the DCT transform matrix an orthogonal matrix, and u represents a generalized frequency variable.

[0047] For two-dimensional DCT transform, the specific formula is as follows:

[0048]

[0049] Among them, f(i, j) is the pixel point of the corresponding element, j represents the traversal pointer, v represents the generalized frequency variable, and the specific formulas of c(u) and c(v) are as follows:

[0050]

[0051] Next, calculate the average value. The above formula gives F(u, v), where F(0, 0) is the DC coefficient, called the DC coefficient, and the others are the AC coefficients.

[0052] Next, a hash value is calculated. The grayscale of each pixel is compared with the average value of the DCT coefficients. If it is greater than or equal to the average value, it is recorded as 1; if it is less than the average value, it is recorded as 0. The results are combined to form a 64-bit binary integer, which is the fingerprint of the image.

[0053] Finally, compare the image fingerprints. Once the image fingerprint is obtained, the fingerprints of different images can be compared to calculate how many of the 64 bits are different. Generally speaking, if the number of different bits does not exceed 5, the two images are very similar; if it is greater than 10, they are different.

[0054] Specifically, the similarity can be determined by the ratio of the number of identical data bits in the image fingerprint to the total number of 64 bits.

[0055] S103, if the similarity is within the first preset range, calculating the signal-to-noise ratio of the first ultrasound image, and comparing the signal-to-noise ratio with a second preset range;

[0056] Specifically, for the calculation of the signal-to-noise ratio of an ultrasound image, the specific formula for calculating the signal-to-noise ratio is as follows:

[0057]

[0058] Where SNR stands for signal-to-noise ratio, represents the signal mean, Represents the standard deviation of signal data relative to random noise;

[0059] Among them, the signal mean The specific formula is as follows:

[0060]

[0061] Among them, sig represents the signal, the pixel value of sig is between 0 and 255, which is also applicable to three channels, and n represents the number of pixels contained in the image;

[0062] Among them, the standard deviation The specific formula is as follows:

[0063]

[0064] in, Represents the pixel value of the i-th pixel.

[0065] Preferably, if the similarity is outside the first preset range, the scanning speed of the ultrasound probe is adjusted, and the process proceeds to the step of acquiring the first ultrasound image of the current frame;

[0066] If the similarity is greater than a first preset range, the scanning speed of the ultrasound probe is increased;

[0067] At this time, it indicates that the first ultrasonic image and the second ultrasonic image are too similar. The scanning speed of the ultrasonic probe can be increased so that the similarity between the ultrasonic image of the next frame to be collected and the ultrasonic image of the current frame can meet the conditions, thereby obtaining a reasonable ultrasonic probe inspection scanning speed.

[0068] If the similarity is less than the first preset range, the scanning speed of the ultrasound probe is reduced.

[0069] At this time, it indicates that the first ultrasound image and the second ultrasound image are too different. The scanning speed of the ultrasound probe can be reduced so that the similarity between the ultrasound image of the next frame to be collected and the ultrasound image of the current frame can meet the conditions, thereby obtaining a reasonable ultrasound probe inspection scanning speed.

[0070] S104: If the signal-to-noise ratio is within a second preset range, inputting the first ultrasound image into the trained target detection model to obtain a detection result;

[0071] Specifically, before detecting the first ultrasound image using the target detection model, it is necessary to first construct the target detection model and train it.

[0072] Using a deep neural network-based object detection model to precisely locate breast lesions achieves high accuracy and reduces the influence of subjective factors in the judgment process. The trained model is robust and highly reliable, helping doctors accurately judge scan images, reducing workload and the influence of subjective factors.

[0073] See Figure 2 ,Specifically, training the target detection model includes:

[0074] S201, obtaining a training set, wherein standard ultrasound images in the training set are annotated with ground truth frames;

[0075] Among them, the true frame refers to the labeling frame of the breast lesion on the standard ultrasound image, which is usually manually labeled.

[0076] S202, inputting the training set into the target detection model to obtain the output result of the target detection model;

[0077] The target detection model mainly includes Backbone, Neck, and Heads. Backbone uses ResNet, Neck uses PAN, and Heads uses YOLOV3's Heads. The design of this target detection model mainly includes the following aspects:

[0078] First, the backbone network specifically refers to the feature extraction network, whose function is to extract information from the image for use by subsequent networks. The dense module includes multiple convolutional layers, each of which consists of batch normalization, ReLU activation function, and subsequent convolution. Combining multiple dense modules with convolutional pooling to construct DenseNet, we choose to use multi-scale feature maps for detection, using both large and small feature maps for detection: the larger one can be used to detect smaller objects, while the smaller one is used to detect larger objects, thereby obtaining more accurate information.

[0079] Secondly, the neck network, specifically located between the backbone and head networks, serves to better utilize the backbone network. Similar to the FPN approach, for object detection prediction at different layers, FPN upsamples the top-down data stream by a factor of 2 and adds it to the adjacent bottom-up layer. The result is then passed to a 3×3 convolution kernel to reduce upsampling artifacts and create a feature map for the head. SPP, on the other hand, uses a slightly different approach, replacing the final pooling layer (after the final convolutional layer) with a spatial pyramid pooling layer. Its feature map is spatially divided into M*M bins, where M can be 1, 2, 4, and so on. Max pooling is then applied to each bin for each channel, resulting in a fixed-length representation that can be further analyzed using fully connected layers. While many CNN-based models only accept input images of a specific size, SPP can accommodate images of varying sizes. SPP is modified to preserve the spatial size of the output. Max pooling is applied to sliding kernels of sizes 1×1, 5×5, 9×9, 13×13, and so on. Feature maps from different kernel sizes are then concatenated as the output. Instead of adding adjacent layers together, feature maps are concatenated. SAM (Spatial Attention Mechanism) applies max pooling and average pooling to the input feature maps, resulting in two sets of feature maps. The result is fed into a convolutional layer, followed by a sigmoid function to create spatial attention.

[0080] Third, the head network refers to the network that obtains the network output content and uses the features extracted by the backbone network to make predictions.

[0081] Fourth, Mosaic data enhancement is used to increase detection speed and simplify the network structure. During the network training phase, the system uses Mosaic enhancement, randomly selecting four images each time for operations such as flipping, scaling, and color gamut changes, and rearranging them in four directions to obtain a combination of multiple images and true frames, thereby increasing the training set data. In order to better learn the image features used to determine the target for detection, the input image is divided into an S*S grid. Each grid contains the coordinates, length, width, and confidence prediction. The confidence includes two aspects: the probability that the predicted frame contains the target, and the accuracy of the predicted frame. A convolutional network is used to reduce channels, and then the channels are passed through the convolutional layer and the fully connected layer. A threshold is set to filter out the predicted frames below the threshold, and non-maximum suppression is performed on those above the threshold to obtain the predicted body.

[0082] S203, determining a loss function based on the predicted box and the true box of the output result;

[0083] Among them, for the loss function, first of all, the definition of the IOU loss function is the difference between 1 and the intersection-and-union ratio between the predicted box A and the real box B, which can be expressed as:

[0084]

[0085] in, Represents the IOU loss function, and IOU(A, B) represents the difference between the intersection and union ratio between the predicted box A and the real box B.

[0086] However, this loss function only works when the bounding boxes overlap. If they do not overlap, it will not provide a sliding gradient.

[0087] Therefore, in order to improve the accuracy, the loss function of CIOU is given, which is defined as

[0088]

[0089] Among them, A represents the predicted box, B represents the real box, It represents the coordinates of the center point of the prediction box. It represents the coordinates of the center point of the real frame. Indicates the calculation of the Euclidean distance, C represents the diagonal length of the minimum bounding box of A and B, α represents a positive number, and v is used to indicate the consistency of the measured aspect ratio. The specific definitions are as follows:

[0090]

[0091] in, and Indicates the width and height of the real frame, and h represent the width and height of the prediction box.

[0092] S204, iteratively training the target detection model based on the loss function.

[0093] Among them, through the above training method, a target detection model after training is obtained.

[0094] See Figure 3 Specifically, inputting the first ultrasound image into the trained target detection model to obtain a detection result includes:

[0095] S301, obtaining a rotation angle set, where the rotation angles in the rotation angle set are obtained by gradually decreasing based on a preset rotation angle threshold;

[0096] For the rotation angle, a slightly larger rotation angle is first given, and then the rotation angle is gradually reduced to obtain a set of multiple different rotation angles.

[0097] Based on actual operational experience and performance calculations, the initial preset rotation angle threshold can be set to 30 degrees. When reducing the angle, the angle should be reduced to half the previous rotation angle. For example, if the initial rotation angle is 30 degrees, the first reduction should be 15 degrees, and so on until the rotation angle is reduced to 1 degree.

[0098] S302, performing an image rotation operation on the first ultrasound image according to the rotation angle set to obtain a set of rotated ultrasound images;

[0099] S303, inputting the set of rotated ultrasound images into a target detection model to obtain a set of detection results output by the target detection model;

[0100] S304 , calculating the variance of the image in the prediction frame of the detection result in the detection result set, and screening out the prediction frame corresponding to the minimum variance as the detection result output.

[0101] Among them, for the image in the prediction box of the detection result, the smaller the variance of its graph, the closer the target prediction box is to the object contour, and the smallest variance value of the image in the prediction box indicates that its corresponding prediction box is closest to the object contour.

[0102] Therefore, screening the prediction box corresponding to the minimum variance value is to select the prediction box closest to the lesion contour as the detection result of the target detection model.

[0103] Preferably, if the signal-to-noise ratio is outside the second preset range, the probing pressure of the ultrasound probe is adjusted, and the process proceeds to the step of acquiring the first ultrasound image of the current frame;

[0104] If the signal-to-noise ratio is greater than a second preset range, the probing pressure of the ultrasonic probe is reduced;

[0105] If the signal-to-noise ratio is less than the second preset range, the probing pressure of the ultrasonic probe is increased.

[0106] The signal-to-noise ratio of the first ultrasound image is controlled within a reasonable range by adjusting the probing pressure of the ultrasound probe, so as to control the quality of the acquired ultrasound image.

[0107] S105: If the detection result indicates that there is a lesion in the first ultrasound image and the lesion is located in the center of the image, the first ultrasound image is saved.

[0108] Preferably, if the detection result indicates that there is no lesion in the first ultrasound image or the lesion is not located in the center of the image, the scanning position of the ultrasound probe is adjusted, and the process proceeds to the step of acquiring the first ultrasound image of the current frame;

[0109] If the detection result indicates that no lesion exists in the first ultrasound image, the ultrasound probe is rotated around the breast;

[0110] If the detection result indicates that the lesion in the first ultrasound image is not located at the center of the image, the ultrasound probe is translated toward the lesion.

[0111] The purpose of rotating the ultrasound probe around the breast is to find the lesion.

[0112] Among them, the neural network model is used to judge whether the target detection of breast ultrasound imaging is located in the center position, which increases stability and timeliness and avoids the influence of subjective factors of manual operation as much as possible.

[0113] Specifically, determining whether the lesion is located in the center of the first ultrasound examination image includes:

[0114] calculating coordinates of a first central region of the first ultrasound examination image;

[0115] Calculate the coordinates of the second center area of ​​the prediction box of the detection result;

[0116] The first central region coordinates are compared to the second central region coordinates.

[0117] If the prediction box is not located in the central area of ​​the first ultrasound examination image, the ultrasound probe is translated toward the lesion;

[0118] Preferably, after saving the first ultrasound image, the method further includes rotating the ultrasound probe to obtain and save ultrasound images of the lesion at multiple angles.

[0119] Among them, the ultrasound probe rotation operation here is to rotate the ultrasound probe axially to guide the doctor to collect lesion images at different angles for subsequent diagnosis and treatment.

[0120] In some embodiments, a breast ultrasound image acquisition device is also disclosed, comprising:

[0121] An image acquisition module, the image acquisition module is used to acquire a first ultrasound image of a current frame;

[0122] A similarity calculation module is used to calculate the similarity between the first ultrasonic image and the second ultrasonic image of the previous frame, and compare the similarity with a first preset range;

[0123] a signal-to-noise ratio calculation module, the signal-to-noise ratio calculation module being configured to calculate the signal-to-noise ratio of the first ultrasound image if the similarity is within a first preset range, and compare the signal-to-noise ratio with a second preset range;

[0124] A target detection module, the target detection module is used to input the first ultrasound image into a trained target detection model to obtain a detection result if the signal-to-noise ratio is within a second preset range;

[0125] The image acquisition module is used to save the first ultrasound image if the detection result shows that there is a lesion in the first ultrasound image and the lesion is located in the center of the image.

[0126] In some embodiments, a computer device is also disclosed, including a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor performs the steps of the above-mentioned breast ultrasound image acquisition method.

[0127] The computer device may be a desktop computer, a notebook computer, a PDA, a cloud server, etc. The computer device may interact with the user via a keyboard, a mouse, a remote control, a touchpad, or a voice control device.

[0128] The memory includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or D-interface display memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, magnetic disk, optical disk, etc. In some embodiments, the memory may be an internal storage unit of the computer device, such as the computer device's hard disk or internal memory. In other embodiments, the memory may also be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. Of course, the memory may also include both the internal storage unit of the computer device and its external storage devices. In this embodiment, the memory is often used to store the operating system and various application software installed on the computer device, such as the program code of the breast ultrasound image acquisition method. In addition, the memory may also be used to temporarily store various types of data that have been output or are about to be output.

[0129] In some embodiments, the processor may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor is typically used to control the overall operation of the computer device. In this embodiment, the processor is used to execute program code stored in the memory or process data, such as executing the program code for the breast ultrasound image acquisition method.

[0130] In some embodiments, a computer-readable storage medium is also disclosed, storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the above-mentioned breast ultrasound image acquisition method.

[0131] The computer-readable storage medium stores an interface display program, and the interface display program can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned breast ultrasound image acquisition method.

[0132] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server or network device, etc.) to execute the methods described in each embodiment of the present application.

[0133] The above are embodiments of the present invention. The above embodiments and the specific parameters therein are only for the purpose of clearly describing the verification process of the invention and are not intended to limit the scope of patent protection of the present invention. The scope of patent protection of the present invention shall still be subject to the claims. Any equivalent structural changes made by using the contents of the description and drawings of the present invention shall be included in the scope of protection of the present invention.

Claims

1. A method for acquiring breast ultrasound images, characterized in that: include: Acquire a first ultrasound image of a current frame; Calculating the similarity between the first ultrasonic image and the second ultrasonic image of the previous frame, and comparing the similarity with a first preset range; If the similarity is within the first preset range, calculating the signal-to-noise ratio of the first ultrasound image, and comparing the signal-to-noise ratio with a second preset range; If the signal-to-noise ratio is within the second preset range, inputting the first ultrasound image into the trained target detection model to obtain a detection result; If the detection result indicates that there is a lesion in the first ultrasound image and the lesion is located in the center of the image, the first ultrasound image is saved.

2. The breast ultrasound image acquisition method according to claim 1, characterized in that: If the similarity is outside the first preset range, adjusting the scanning speed of the ultrasound probe, and switching to the step of acquiring the first ultrasound image of the current frame; Wherein, if the similarity is greater than the first preset range, the scanning speed of the ultrasound probe is accelerated; If the similarity is less than the first preset range, the scanning speed of the ultrasound probe is reduced.

3. The breast ultrasound image acquisition method according to claim 1, characterized in that: If the signal-to-noise ratio is outside the second preset range, adjusting the probing pressure of the ultrasonic probe, and proceeding to the step of acquiring a first ultrasonic image of the current frame; Wherein, if the signal-to-noise ratio is greater than the second preset range, reducing the probing pressure of the ultrasonic probe; If the signal-to-noise ratio is less than the second preset range, the probing pressure of the ultrasonic probe is increased.

4. The breast ultrasound image acquisition method according to claim 1, characterized in that: If the detection result indicates that there is no lesion in the first ultrasonic image or the lesion is not located in the center of the image, the scanning position of the ultrasonic probe is adjusted, and the process proceeds to the step of acquiring the first ultrasonic image of the current frame; If the detection result indicates that there is no lesion in the first ultrasound image, the ultrasound probe is rotated around the breast; If the detection result indicates that the lesion in the first ultrasound image is not located at the center of the image, the ultrasound probe is translated toward the lesion.

5. The breast ultrasound image acquisition method according to claim 1, characterized in that: After the first ultrasound image is saved, the method further includes: Rotate the ultrasound probe to obtain and save ultrasound images of the lesion at multiple angles.

6. The breast ultrasound image acquisition method according to claim 1, characterized in that: Training the target detection model includes: Acquire a training set, wherein standard ultrasound images in the training set are annotated with true frames; Inputting the training set into the target detection model to obtain an output result of the target detection model; Determine a loss function based on the predicted box of the output result and the true box; The target detection model is iteratively trained based on the loss function.

7. The breast ultrasound image acquisition method according to claim 1, characterized in that: Inputting the first ultrasound image into the trained target detection model to obtain a detection result includes: Acquire a rotation angle set, wherein the rotation angles in the rotation angle set are obtained by gradually decreasing based on a preset rotation angle threshold; Performing an image rotation operation on the first ultrasound image according to the rotation angle set to obtain a set of rotated ultrasound images; Inputting the set of rotated ultrasound images into the target detection model to obtain a set of detection results output by the target detection model; The variance of the image in the prediction frame of the detection result in the detection result set is calculated, and the prediction frame corresponding to the minimum variance is screened out as the detection result output.

8. A breast ultrasound image acquisition device, characterized in that: include: An image acquisition module, wherein the image acquisition module is used to acquire a first ultrasound image of a current frame; a similarity calculation module, the similarity calculation module being used to calculate the similarity between the first ultrasonic image and the second ultrasonic image of the previous frame, and compare the similarity with a first preset range; a signal-to-noise ratio calculation module, the signal-to-noise ratio calculation module being used to calculate the signal-to-noise ratio of the first ultrasound image if the similarity is within the first preset range, and compare the signal-to-noise ratio with a second preset range; a target detection module, wherein if the signal-to-noise ratio is within the second preset range, the target detection module is used to input the first ultrasound image into a trained target detection model to obtain a detection result; An image acquisition module, wherein the image acquisition module is used to save the first ultrasound image if the detection result indicates that there is a lesion in the first ultrasound image and the lesion is located in the center of the image.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the breast ultrasound image acquisition method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a processor, the processor executes the steps of the breast ultrasound image acquisition method according to any one of claims 1 to 7.