Human tissue ultrasonic image processing method, device, equipment and medium

Abstract

The invention discloses a human tissue ultrasonic image processing method and device, equipment and a medium, which are applied to the technical field of medical ultrasonic image processing, and comprise the following steps: acquiring a human tissue ultrasonic image under the current treatment depth; performing feature extraction on the human tissue ultrasonic image through a ResNet network to obtain human tissue feature images of different semantic hierarchies, and performing semantic segmentation based on the human tissue feature images of different semantic hierarchies through a UNet network to obtain pixel position information of different human tissues; according to the method, when it is determined that treatment prohibited tissues exist under the current treatment depth based on pixel position information of different human tissues, injury risk prompting is performed based on the treatment prohibited tissues under the current treatment depth, so that different human tissues can be quickly and accurately identified through a ResNet and UNet mixed model, and the treatment prohibited tissues can be quickly and accurately identified through the ResNet and UNet mixed model. Therefore, the interference between the current treatment depth and the treatment-prohibited tissue can be quickly and accurately judged and prompted, and the treatment safety is improved.

Description

Technical Field

[0001] The present application relates to the technical field of medical ultrasonic image processing, and in particular to a method, device, equipment and medium for processing ultrasonic images of human tissue. Background Art

[0002] Ultrasonic beauty equipment is a device that uses ultrasonic technology for skin treatment. It emits ultrasonic waves of a specific frequency, allowing them to penetrate the skin surface to reach the dermis or even deeper tissue layers, stimulating high-frequency mechanical vibrations in human tissues to achieve different treatment effects. However, at present, during the treatment process, due to the inability to identify or recognize the tissue distribution at the treatment depth, it may cause accidental damage to bones, blood vessels and other prohibited tissues for treatment, which in turn may cause skin bruises and long-term neuralgic pain or numbness, and even cause more serious medical accidents. Summary of the invention

[0003] The present application provides a method, device, equipment and medium for processing ultrasonic images of human tissues, which is used to solve the problem that ultrasonic cosmetic devices in the prior art cannot identify or recognize the distribution of human tissues at the treatment depth, thus affecting the safety of treatment. The technical solutions provided by the present application are as follows: On the one hand, the present application provides a method for processing ultrasonic images of human tissue, which is applied to ultrasonic cosmetic equipment, comprising: Acquire an ultrasonic image of human tissue in a current treatment area detected at a target imaging depth; wherein the target imaging depth is greater than an initial treatment depth of the current treatment area; Based on the human tissue recognition model, human tissue recognition is performed on the human tissue ultrasound image to obtain pixel position information of different human tissues in the human tissue ultrasound image; wherein the human tissue recognition model includes a ResNet network and a UNet network connected in sequence; the human tissue recognition model performs feature extraction processing on the human tissue ultrasound image through the ResNet network to obtain first human tissue feature images of different first semantic levels, performs feature enhancement processing on the first human tissue feature images of each first semantic level through the UNet network to obtain second human tissue feature images of different second semantic levels, and then performs semantic segmentation based on the second human tissue feature images of each second semantic level to obtain pixel position information of different human tissues in the human tissue ultrasound image; When it is determined based on the pixel position information of different human tissues in the human tissue ultrasound image that there is prohibited treatment tissue in the current treatment area at the initial treatment depth, an injury risk warning is given based on the prohibited treatment tissue in the current treatment area at the initial treatment depth.

[0004] Optionally, the ResNet network includes a plurality of downsampling layers connected in sequence; the image resolution output by each downsampling layer decreases exponentially, the number of feature channels output by the first downsampling layer is the same as the number of feature channels output by the second downsampling layer, and the number of feature channels output by the second and subsequent downsampling layers increases exponentially; The UNet network includes an adaptation layer, multiple upsampling layers, and a fully connected layer connected in sequence; the multiple upsampling layers are jump-connected to the multiple downsampling layers of the ResNet network in a one-to-one correspondence; the image resolution output by each upsampling layer increases exponentially and the number of feature channels decreases exponentially.

[0005] Optionally, a ResNet network is used to perform feature extraction processing on the human tissue ultrasound image to obtain first human tissue feature images at different first semantic levels, including: Through the first downsampling layer in the ResNet network, the human tissue ultrasound image is sequentially downsampled and dimensionally reduced to obtain the first human tissue feature image at a high-resolution level; Through the second and subsequent downsampling layers in the ResNet network, the first human tissue feature image at the high-resolution level is iteratively downsampled to obtain the first human tissue feature images at different low-resolution levels; Based on the first human tissue feature image at a high-resolution level and the first human tissue feature images at different low-resolution levels, first human tissue feature images at different first semantic levels are obtained.

[0006] Optionally, after performing feature enhancement processing on the first human tissue feature images of each first semantic level to obtain second human tissue feature images of different second semantic levels through a UNet network, semantic segmentation is performed based on the second human tissue feature images of each second semantic level to obtain pixel position information of different human tissues in the human tissue ultrasound image, including: The first human tissue feature image of the first semantic level output by the last downsampling layer in the ResNet network is adapted through the adaptation layer in the UNet network to obtain the first human tissue feature image of the first semantic level with the target feature channel number; After upsampling the first human tissue feature image of the first semantic level of the target feature channel number through the first upsampling layer in the UNet network, feature fusion processing is performed with the first human tissue feature image of the first semantic level output by the downsampling layer in the corresponding jump-connected ResNet network to obtain a second human tissue feature image of the second semantic level; For the second and subsequent upsampling layers in the UNet network, the second human tissue feature image of the second semantic level output by the previous upsampling layer is upsampled through the upsampling layer, and then feature fused with the first human tissue feature image of the first semantic level output by the downsampling layer in the corresponding jump-connected ResNet network to obtain the second human tissue feature image of the second semantic level; Through the fully connected layer in the UNet network, the second human tissue feature image of the second semantic level output by the last upsampling layer is semantically segmented to obtain the pixel position information of different human tissues in the human tissue ultrasound image.

[0007] Optionally, determining whether there is a prohibited treatment tissue in the current treatment area at the initial treatment depth based on pixel position information of different human tissues in the human tissue ultrasound image includes: Based on the target imaging depth, the pixel position information of different human tissues in the human tissue ultrasound image is converted into the spatial depth information of different human tissues in the current treatment area; Based on the spatial depth information of different human tissues in the current treatment area, when prohibited treatment tissue whose spatial depth information overlaps with the initial treatment depth is detected in the different human tissues, it is determined that prohibited treatment tissue exists in the current treatment area at the initial treatment depth.

[0008] Optionally, when determining that there is a prohibited treatment tissue in the current treatment area at the initial treatment depth based on the pixel position information of different human tissues in the human tissue ultrasound image, the method further includes: Based on the spatial depth information of the prohibited treatment tissue in the current treatment area at the initial treatment depth, the initial treatment depth of the current treatment area is adjusted to obtain the target treatment depth of the current treatment area.

[0009] Optionally, after performing human tissue recognition on the human tissue ultrasound image based on the human tissue recognition model to obtain pixel position information of different human tissues in the human tissue ultrasound image, the method further includes: The pixel position information of different human tissues in the human tissue ultrasound image is labeled and displayed.

[0010] On the other hand, the present application provides a human tissue ultrasonic image processing device, which is applied to ultrasonic cosmetic equipment, comprising: An ultrasonic image acquisition unit, used to acquire an ultrasonic image of human tissue in a current treatment area detected according to a target imaging depth; wherein the target imaging depth is greater than an initial treatment depth of the current treatment area; A human tissue recognition unit is used to perform human tissue recognition on a human tissue ultrasound image based on a human tissue recognition model to obtain pixel position information of different human tissues in the human tissue ultrasound image; wherein the human tissue recognition model includes a ResNet network and a UNet network connected in sequence; the human tissue recognition model performs feature extraction processing on the human tissue ultrasound image through the ResNet network to obtain first human tissue feature images of different first semantic levels, performs feature enhancement processing on the first human tissue feature images of each first semantic level through the UNet network to obtain second human tissue feature images of different second semantic levels, and then performs semantic segmentation based on the second human tissue feature images of each second semantic level to obtain pixel position information of different human tissues in the human tissue ultrasound image; The injury risk warning unit is used to determine, based on the pixel position information of different human tissues in the human tissue ultrasound image, that there is prohibited treatment tissue in the current treatment area at the initial treatment depth, and to provide an injury risk warning based on the prohibited treatment tissue in the current treatment area at the initial treatment depth.

[0011] On the other hand, the present application provides an ultrasonic beauty device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned human tissue ultrasonic image processing method when executing the computer program.

[0012] On the other hand, the present application provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed by a processor, the above-mentioned human tissue ultrasonic image processing method is implemented.

[0013] The beneficial effects of this application are as follows: During the skin treatment process according to the initial treatment line formulated for the human treatment part, the present application performs human tissue recognition on the human tissue ultrasound image of the current treatment area at the initial treatment depth in the initial treatment line by utilizing a human tissue recognition model based on the ResNet network and the UNet network. The application can quickly and accurately identify different human tissues in the human tissue ultrasound image, thereby being able to quickly and accurately determine whether there is prohibited treatment tissue in the current treatment area at the initial treatment depth, and then provide an injury risk warning based on the prohibited treatment tissue in the current treatment area at the initial treatment depth, thereby effectively avoiding problems such as skin bruises, prolonged neuralgic pain or numbness, and more serious medical accidents caused by accidental damage to prohibited treatment tissues such as bones and blood vessels, and then effectively improving the safety of treatment.

[0014] Other features and advantages of the present application will be described in the subsequent description, and in part, will become apparent from the description, or will be understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 Schematic diagram of the general flow of the method for processing ultrasonic images of human tissue in the embodiment of the present application; Figure 2 A schematic diagram of the general structure of the human tissue recognition model in the embodiment of the present application; Figure 3 This is a schematic diagram of the specific structure of the human tissue recognition model in the embodiment of the present application; Figure 4 A schematic diagram of the general flow of the human tissue recognition model training method in the embodiment of the present application; Figure 5 This is a functional structure diagram of a human tissue ultrasonic image processing device in an embodiment of the present application; Figure 6 Schematic diagram of the hardware structure of the ultrasonic beauty device in the embodiment of the present application. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and beneficial effects of this application clearer, the technical solution in the embodiment of this application will be clearly and completely described below in conjunction with the drawings in the embodiment of this application. Obviously, the described embodiment is only a part of the embodiment of this application, not all the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0017] In order to facilitate those skilled in the art to better understand the present application, the technical terms involved in the present application are briefly introduced below.

[0018] The human tissue recognition model is a neural network model for identifying different human tissues and pixel position information of different human tissues in human tissue ultrasound images. The neural network architecture adopted by the model is improved based on the UNet network, and is a hybrid network architecture based on the ResNet network (such as the ResNet34 network) and the UNet network. Among them, the ResNet network, as the front-end network of the human tissue recognition model, that is, as the encoder of the human tissue recognition model, can extract the low-resolution feature information of the human tissue in the human tissue ultrasound image at a deeper level, and will not cause the problems of gradient disappearance and grid degradation, and can also accelerate network convergence; the UNet network, as the back-end network of the human tissue recognition model, that is, as the decoder of the human tissue recognition model, can fully extract the high-resolution feature information of the human tissue in the human tissue ultrasound image (that is, the segmentation detail feature information) through feature splicing and multi-scale fusion, so as to fully integrate the different low-resolution feature information (that is, deep feature information) and different high-resolution feature information (that is, shallow feature information) of the human tissue ultrasound image, and thus can more accurately complete the recognition of different human tissues and pixel position information of different human tissues in the human tissue ultrasound image.

[0019] The target imaging depth is the ultrasonic detection depth used by the ultrasonic beauty device to collect ultrasonic images of human tissue; in the present application, the target imaging depth is a fixed value and is greater than the maximum treatment depth of the ultrasonic beauty device, for example, the target imaging depth is 8 mm.

[0020] The initial treatment depth is one of the treatment depths selected from the different treatment depths equipped with the ultrasonic beauty device according to the patient's treatment needs; in the present application, the different treatment depths equipped with the ultrasonic beauty device include 1.5mm, 3mm, 4.5mm, etc., and the initial treatment depth is one of 1.5mm, 3mm, 4.5mm, etc.

[0021] The target treatment depth is the treatment depth obtained by adjusting the initial treatment depth based on the spatial depth information of the prohibited treatment tissue in the current treatment area at the initial treatment depth identified by the human tissue recognition model.

[0022] After introducing the technical terms involved in this application, the application scenarios and design concepts of this application are briefly introduced.

[0023] The treatment principle of ultrasonic beauty equipment is to focus ultrasound into a certain depth of tissue in the target treatment area to stimulate high-frequency mechanical vibration of human tissue to generate a certain amount of heat, resulting in a temperature rise of about 60~70℃. If this heat is focused on human tissues such as bones or blood vessels, it will cause treatment side effects and should be avoided as much as possible. At present, in order to adapt to the treatment depth required by different treatment projects (that is, the ultrasonic detection depth), although ultrasonic beauty equipment is equipped with ultrasonic probes with different treatment depths, for example, ultrasonic probes with various treatment depths such as 1.5mm, 3mm, and 4.5mm under the skin, during the treatment process using an ultrasonic probe at a certain treatment depth, due to the inability to identify or recognize the tissue distribution at the treatment depth, it may cause accidental damage to prohibited treatment tissues such as bones and blood vessels, and then cause skin bruises and long-term neuralgic pain or numbness, and even cause more serious medical accidents.

[0024] To this end, in the present application, the ultrasonic beauty device is equipped with an ultrasonic probe with a target imaging depth, and the target imaging depth is greater than all treatment depths of the ultrasonic beauty device. For example, the target imaging depth is 8 mm subcutaneously, so that the ultrasonic probe can detect human tissue ultrasonic images that meet all treatment depths. Based on this, when the ultrasonic beauty device is ready to treat the current treatment area of ​​the target patient, the ultrasonic probe can be used to detect the current treatment area according to the target imaging depth to obtain a human tissue ultrasonic image of the current treatment area. Afterwards, the human tissue ultrasonic image is feature extracted through the ResNet network in the human tissue recognition model to obtain first human tissue feature images of different first semantic levels, and the U network in the human tissue recognition model is used to extract the human tissue ultrasonic image. Net network, after performing feature enhancement processing on first human tissue feature images of different first semantic levels to obtain second human tissue feature images of different second semantic levels, semantic segmentation is performed based on the second human tissue feature images of different second semantic levels, so that the position information of different human tissues in the human tissue ultrasound image can be quickly and accurately obtained, and then it can be quickly and accurately determined whether there is prohibited treatment tissue in the current treatment area at the initial treatment depth, and then the injury risk warning can be given based on the prohibited treatment tissue in the current treatment area at the initial treatment depth, so as to effectively avoid problems such as skin ecchymosis, prolonged neuralgic pain or numbness, and more serious medical accidents caused by accidental damage to prohibited treatment tissues such as bones and blood vessels, thereby improving the safety of ultrasound treatment.

[0025] After introducing the application scenarios and design ideas of the present application, the technical solutions provided by the present application are described in detail below.

[0026] The present application embodiment provides a method for processing human tissue ultrasonic images, which is applied to ultrasonic cosmetic equipment. Figure 1As shown, the general process of the human tissue ultrasonic image processing method provided in the embodiment of the present application is as follows: Step 101: Acquire an ultrasonic image of human tissue in a current treatment area detected at a target imaging depth; wherein the target imaging depth is greater than an initial treatment depth of the current treatment area.

[0027] In an embodiment of the present application, the ultrasonic beauty device detects the current treatment area through an ultrasonic probe of a target imaging depth to obtain an ultrasonic image of human tissue in the current treatment area. In one embodiment, the ultrasonic probe of the target imaging depth may be a new ultrasonic probe equipped for the ultrasonic beauty device, and the new ultrasonic probe has a fixed ultrasonic detection depth (i.e., the target imaging depth). In another embodiment, the ultrasonic probe of the target imaging depth may also be one of the original ultrasonic probes of the ultrasonic beauty device, for example, an ultrasonic probe with the largest treatment depth. When collecting the ultrasonic image of human tissue, the ultrasonic detection depth of the ultrasonic probe with the largest treatment depth is increased to the target imaging depth by driving the servo motor corresponding to the ultrasonic probe with the largest treatment depth, so that the current treatment area can be detected according to the target imaging depth to obtain an ultrasonic image of human tissue in the current treatment area. In a specific implementation, when obtaining an ultrasonic image of human tissue in the current treatment area detected according to the target imaging depth, the waveform data obtained by the ultrasonic probe detecting the current treatment area according to the target imaging depth may be first obtained, and then the waveform data may be cleaned, Fourier transformed, and low-pass filtered to obtain the ultrasonic image of human tissue in the current treatment area.

[0028] Step 102: Perform human tissue recognition on the human tissue ultrasound image based on the human tissue recognition model to obtain pixel position information of different human tissues in the human tissue ultrasound image; wherein the human tissue recognition model includes a ResNet network and a UNet network connected in sequence; the human tissue recognition model performs feature extraction processing on the human tissue ultrasound image through the ResNet network to obtain first human tissue feature images of different first semantic levels, performs feature enhancement processing on the first human tissue feature images of each first semantic level through the UNet network to obtain second human tissue feature images of different second semantic levels, and then performs semantic segmentation based on the second human tissue feature images of each second semantic level to obtain pixel position information of different human tissues in the human tissue ultrasound image.

[0029] In the present application, see Figure 2As shown, the human tissue recognition model includes a ResNet network and a UNet network connected in sequence; the ResNet network includes multiple downsampling layers connected in sequence; the image resolution output by each downsampling layer decreases exponentially, the number of feature channels output by the first downsampling layer is the same as the number of feature channels output by the second downsampling layer, and the number of feature channels output by the second and subsequent downsampling layers increases exponentially; the UNet network includes an adaptation layer, multiple upsampling layers, and a fully connected layer connected in sequence; the multiple upsampling layers are jump-connected to the multiple downsampling layers of the ResNet network in a one-to-one correspondence; the image resolution output by each upsampling layer increases exponentially and the number of feature channels decreases exponentially, so that each upsampling layer in the UNet network can gradually restore the resolution level, ensuring that the input and output resolution levels of the human tissue recognition model are the same. Among them, Figure 2 This is only an exemplary human tissue recognition model. The number of downsampling layers included in the ResNet network and the number of upsampling layers included in the UNet network in the human tissue recognition model can be flexibly set according to actual needs. It is only necessary to ensure that the number of downsampling layers in the ResNet network is the same as the number of upsampling layers in the UNet network and have one-to-one corresponding jump connections.

[0030] Based on Figure 2 The human tissue recognition model shown in the figure, in the process of performing human tissue recognition on the human tissue ultrasound image based on the human tissue recognition model, when performing feature extraction processing on the human tissue ultrasound image through the ResNet network to obtain the first human tissue feature images of different first semantic levels, the following methods can be used but are not limited to: First, the human tissue ultrasound image is downsampled and dimensionally reduced in sequence through the first downsampling layer in the ResNet network to obtain the first human tissue feature image at a high-resolution level.

[0031] Then, the first human tissue feature image at the high-resolution level is iteratively downsampled through the second and subsequent downsampling layers in the ResNet network to obtain the first human tissue feature images at different low-resolution levels.

[0032] Finally, based on the first human tissue feature image at the high-resolution level and the first human tissue feature images at different low-resolution levels, first human tissue feature images at different first semantic levels are obtained.

[0033] Based on Figure 2The human tissue recognition model shown in the figure, in the process of human tissue recognition of the human tissue ultrasound image based on the human tissue recognition model, after the first human tissue feature images of each first semantic level are subjected to feature enhancement processing through the UNet network to obtain the second human tissue feature images of different second semantic levels, when the pixel position information of different human tissues in the human tissue ultrasound image is obtained by semantic segmentation based on the second human tissue feature images of each second semantic level, the following methods can be adopted but not limited to: Firstly, the first human tissue feature image of the first semantic level output by the last downsampling layer in the ResNet network is adapted through the adaptation layer in the UNet network to obtain the first human tissue feature image of the first semantic level with the target number of feature channels.

[0034] Next, the first human tissue feature image of the first semantic level with the target feature channel number is upsampled by the first upsampling layer in the UNet network, and then feature fused with the first human tissue feature image of the first semantic level output by the downsampling layer in the corresponding jump-connected ResNet network to obtain the second human tissue feature image of the second semantic level.

[0035] Then, for the second and subsequent upsampling layers in the UNet network, the second human tissue feature image of the second semantic level output by the previous upsampling layer is upsampled through the upsampling layer, and then feature fused with the first human tissue feature image of the first semantic level output by the downsampling layer in the corresponding jump-connected ResNet network to obtain the second human tissue feature image of the second semantic level.

[0036] Finally, the second human tissue feature image of the second semantic level output by the last upsampling layer is semantically segmented through the fully connected layer in the UNet network to obtain the pixel position information of different human tissues in the human tissue ultrasound image; wherein the pixel position information of different human tissues can be represented by a human tissue distribution image, that is, the human tissue recognition model outputs a human tissue distribution image, and the position areas of different human tissues in the human tissue distribution image are represented by different identifiers. For example, the human tissue distribution image includes superficial fascia tissue, deep fascia tissue and vascular tissue, wherein each pixel point in the position area where the superficial fascia tissue is located is represented by an identifier "1", each pixel point in the position area where the deep fascia tissue is located is represented by an identifier "2", and each pixel point in the position area where the vascular tissue is located is represented by an identifier "3".

[0037] Step 103: When it is determined that there is prohibited treatment tissue in the current treatment area at the initial treatment depth based on the pixel position information of different human tissues in the human tissue ultrasound image, an injury risk warning is issued based on the prohibited treatment tissue in the current treatment area at the initial treatment depth.

[0038] In the embodiment of the present application, based on the pixel position information of different human tissues in the human tissue ultrasound image, when determining whether there is a prohibited treatment tissue in the current treatment area at the initial treatment depth, the following methods may be used but are not limited to: Firstly, based on the target imaging depth, the pixel position information of different human tissues in the human tissue ultrasound image is converted into the spatial depth information of the different human tissues in the current treatment area.

[0039] Then, based on the spatial depth information of different human tissues in the current treatment area, when prohibited treatment tissue whose spatial depth information overlaps with the initial treatment depth is detected in different human tissues, it is determined that prohibited treatment tissue exists in the current treatment area at the initial treatment depth.

[0040] In this way, by utilizing the human tissue recognition model based on the ResNet network and the UNet network, human tissue recognition is performed on the human tissue ultrasound image of the current treatment area at the initial treatment depth in the initial treatment line. Different human tissues in the human tissue ultrasound image can be quickly and accurately identified, so that it can be quickly and accurately determined whether there is prohibited treatment tissue in the current treatment area at the initial treatment depth, and then the risk of injury can be warned based on the prohibited treatment tissue in the current treatment area at the initial treatment depth. For example, through pop-up windows and / or voice, the risk of injury can be warned for the prohibited treatment tissue in the current treatment area at the initial treatment depth, thereby playing an auxiliary treatment role, and then the intelligence level of ultrasonic beauty equipment and the skin treatment effect can be improved.

[0041] In addition, in an embodiment of the present application, after determining that there is prohibited treatment tissue in the current treatment area at the initial treatment depth based on the pixel position information of different human tissues in the human tissue ultrasonic image, on the one hand, based on the spatial depth information of the prohibited treatment tissue in the current treatment area at the initial treatment depth, the initial treatment depth of the current treatment area can be adjusted to obtain the target treatment depth of the current treatment area to avoid the depth range where the prohibited treatment tissue is located; on the other hand, the pixel position information and / or spatial depth information of different human tissues can also be labeled and displayed to facilitate the operator to view the distribution of different human tissues in the current treatment area and adjust the initial treatment depth, etc., thereby further improving the intelligence level of ultrasonic beauty equipment and the auxiliary treatment effect.

[0042] The human tissue recognition model and the human tissue ultrasonic image processing method based on the human tissue recognition model are described in detail below. Figure 3 As shown, the human tissue recognition model includes a ResNet network (such as a ResNet34 network) and a UNet network connected in sequence; wherein: The ResNet network is a process of continuously fitting residuals. By learning the residuals between input and output, the network converges faster and has higher recognition accuracy. As the encoder part of the encoder-decoder structure, the ResNet network consists of multiple downsampling layers connected in sequence. Figure 3 In the example, the ResNet network consists of 5 downsampling layers. The first downsampling layer consists of a convolution layer and a pooling layer. The number of convolution kernels in the convolution layer is 64, the size is 7x7, and the step size is 2. The number of pooling windows in the pooling layer is 64, the size is 2 x 2, and the step size is 2; each of the remaining downsampling layers consists of a convolution layer and the number of residual blocks in the convolution layer contained in each of the remaining downsampling layers is the same or different. Specifically, the convolution layer contained in the second downsampling layer consists of 3 residual blocks, and the number of convolution kernels in each residual block is 64, the size is 3x3, and the step size is 1; the convolution layer contained in the third downsampling layer consists of 4 residual blocks, the number of convolution kernels in the first residual block is 128, the size is 3x3, the step size is 2, and the number of convolution kernels in the second residual block to the fourth residual block is 12. 8, size is 3x3, and step size is 1; the convolution layer contained in the 4th downsampling layer consists of 6 residual blocks, the number of convolution kernels in the 1st residual block is 256, the size is 3x3, the step size is 2, and the number of convolution kernels in the 2nd to 6th residual blocks is 256, the size is 3x3, and the step size is 1; the convolution layer contained in the 5th downsampling layer consists of 3 residual blocks, the number of convolution kernels in the 1st residual block is 512, the size is 3x3, the step size is 2, and the number of convolution kernels in the 2nd to 3rd residual blocks is 512, the size is 3x3, and the step size is 1.

[0043] Based on Figure 3 The human tissue recognition model shown in the figure, after the human tissue ultrasound image is input into the human tissue recognition model, the specific process of the ResNet network in the human tissue recognition model to extract features of the human tissue ultrasound image is as follows: First, in the first downsampling layer, the human tissue ultrasound image (taking only the grayscale human tissue ultrasound image of 800x800x1 as an example) is subjected to a convolution operation of 7x7 and a step length of 2 through a convolution layer, and 64 first human tissue feature images with an image dimension of 400x400x1 are obtained. After that, the 64 first human tissue feature images with an image dimension of 400x400x1 are subjected to a dimensionality reduction operation through a pooling layer, and 64 first human tissue feature images with an image dimension of 200x200x1 are output. In this way, by connecting a layer of pooling layers, the data dimension can be reduced, the computational complexity can be reduced, and the risk of overfitting can also be reduced.

[0044] Next, in the second downsampling layer, the 64 first human tissue feature images with an image dimension of 200x200x1 are sequentially passed through the three residual blocks in the convolution layer for a convolution operation with a size of 3x3 and a step size of 1. At the same time, a short connection operation is performed between each residual block to perform residual learning between the residual blocks, and 64 first human tissue feature images with an image dimension of 200x200x1 are output.

[0045] Then, in the third downsampling layer, the 64 first human tissue feature images with an image dimension of 200x200x1 are subjected to a convolution operation with a size of 3x3 and a step size of 2 through the first residual block in the convolution layer, and then sequentially subjected to a convolution operation with a size of 3x3 and a step size of 1 through the second to fourth residual blocks in the convolution layer. At the same time, a short connection operation is performed between each residual block to perform residual learning between the residual blocks, and 128 first human tissue feature images with an image dimension of 100x100x1 are output.

[0046] Secondly, in the 4th downsampling layer, the 128 first human tissue feature images with an image dimension of 100x100x1 are subjected to a convolution operation with a size of 3x3 and a step size of 2 through the 1st residual block in the convolution layer, and then sequentially subjected to a convolution operation with a size of 3x3 and a step size of 1 through the 2nd to 6th residual blocks in the convolution layer. At the same time, a short connection operation is performed between each residual block to perform residual learning between the residual blocks, and 128 first human tissue feature images with an image dimension of 50x50x1 are output.

[0047] Finally, in the fifth downsampling layer, the 128 first human tissue feature images with a dimension of 50x50x1 are convolved with the first residual block in the convolution layer with a size of 3x3 and a step size of 2, and then convolved with the second to third residual blocks in the convolution layer with a size of 3x3 and a step size of 1. At the same time, a short connection operation is performed between each residual block to perform residual learning between residual blocks, and 512 first human tissue feature images with an image dimension of 25x25x1 are output.

[0048] In this way, after the deep convolution operation of the ResNet network, sufficiently macroscopic and abstract human tissue features can be extracted, that is, the first human tissue feature images of different first semantic levels are obtained through the deep convolution operation of the ResNet network.

[0049] The UNet network is a process of continuously splicing features. It gradually restores the image resolution through a series of deconvolutions (also called transposed convolutions or upsampling), while performing feature fusion through a series of jump connections to make the output human tissue distribution more accurate. The UNet network, as the decoder part of the encoder-decoder structure, consists of an adaptation layer, multiple upsampling layers, and a fully connected layer connected in sequence. Figure 3 In this paper, we only take the UNet network consisting of 1 adaptation layer, 5 upsampling layers and 1 fully connected layer as an example. The adaptation layer consists of two convolutional layers, and the number of convolution kernels in the first convolutional layer is one of 512, 768 and 1024 ( Figure 3Only 1024 are taken as an example), the size is 3x3, the stride is 1, the number of convolution kernels in the second convolution layer is 1024, the size is 3x3, and the stride is 1; the first upsampling layer consists of 1 deconvolution layer, 1 splicing layer and 2 convolution layers, and the splicing layer is skipped with the 5th downsampling layer in the ResNet network. The number of convolution kernels in the deconvolution layer is 1024, the size is 2 x 2, the stride is 2, and the number of convolution kernels in the 2 convolution layers is 512, the size is 3x3, and the stride is 1; the second upsampling layer consists of 1 deconvolution layer, 1 splicing layer and 2 convolution layers, and the splicing layer is skipped with the 4th downsampling layer in the ResNet network. The number of convolution kernels in the deconvolution layer is 512, the size is 2 x 2, stride 2, the number of convolution kernels in the 2 convolution layers is 256, the size is 3x3, and the stride is 1; the third upsampling layer consists of 1 deconvolution layer, 1 splicing layer and 2 convolution layers, and the splicing layer is skip-connected to the 3rd downsampling layer in the ResNet network. The number of convolution kernels in the deconvolution layer is 256, the size is 2 x 2, the stride is 2, the number of convolution kernels in the 2 convolution layers is 128, the size is 3x3, and the stride is 1; the fourth upsampling layer consists of 1 deconvolution layer, 1 splicing layer and 2 convolution layers, and the splicing layer is skip-connected to the 2nd downsampling layer in the ResNet network. The number of convolution kernels in the deconvolution layer is 128, the size is 2 x 2, with a stride of 2. The number of convolution kernels in the two convolution layers is 64, with a size of 3x3 and a stride of 1. The fifth upsampling layer consists of 1 deconvolution layer, 1 concatenation layer and 2 convolution layers, and the concatenation layer is jump-connected to the first downsampling layer in the ResNet network. The number of convolution kernels in the deconvolution layer is 64, with a size of 2 x 2 and a stride of 2. The number of convolution kernels in the two convolution layers is 32, with a size of 3x3 and a stride of 1. The fully connected layer consists of 1 point convolution layer and 1 multi-category label mapping layer. The size of the convolution kernel in the point convolution layer is 1x1, with a stride of 1.

[0050] Based on Figure 3 The human tissue recognition model shown in the figure, after the ResNet network in the human tissue recognition model extracts features from the human tissue ultrasound image and outputs first human tissue feature images of different first semantic levels, the UNet network in the human tissue recognition model performs feature enhancement on the first human tissue feature images of different first semantic levels and then performs semantic segmentation. The specific process is as follows: First, in the adaptation layer, the 512 first human tissue feature images with a dimension of 25x25x1 output by the last downsampling layer in the ResNet network are sequentially convolved through two convolutional layers with a size of 3x3 and a step size of 1 to output 1024 first human tissue feature images with a dimension of 25x25x1.

[0051] Next, in the first upsampling layer, the 1024 first human tissue feature images with an image dimension of 25x25x1 are upsampled by a deconvolution layer with a size of 2x2 and a stride of 2 to obtain 1024 human tissue feature images with an image dimension of 50x50x1. They are then fused with the 128 first human tissue feature images with an image dimension of 50x50x1 output by the 5th downsampling layer in the ResNet network through a concatenation layer, and then convolution operations with a size of 3x3 and a stride of 1 are performed in two convolutional layers in sequence to output 512 second human tissue feature images with an image dimension of 50x50x1.

[0052] Then, in the second upsampling layer, 512 second human tissue feature images with an image dimension of 50x50x1 are upsampled by a deconvolution layer with a size of 2 x 2 and a step size of 2 to obtain 512 human tissue feature images with an image dimension of 100x100x1. They are then fused with the 256 first human tissue feature images with an image dimension of 100x100x1 output by the fourth downsampling layer in the ResNet network through a concatenation layer, and then convolution operations with a size of 3x3 and a step size of 1 are performed in two convolutional layers in sequence to output 256 second human tissue feature images with an image dimension of 100x100x1.

[0053] Secondly, in the third upsampling layer, the 256 second human tissue feature images with a dimension of 100x100x1 are upsampled by a deconvolution layer with a size of 2 x 2 and a step size of 2 to obtain 256 human tissue feature images with a dimension of 200x200x1. Then, they are fused with the 128 first human tissue feature images with a dimension of 200x200x1 output by the third downsampling layer in the ResNet network through a concatenation layer, and then convolution operations with a size of 3x3 and a step size of 1 are performed in two convolutional layers in sequence to output 128 second human tissue feature images with a dimension of 200x200x1.

[0054] Afterwards, in the fourth upsampling layer, the 128 second human tissue feature images with an image dimension of 200x200x1 are upsampled by a deconvolution layer with a size of 2 x 2 and a step size of 2 to obtain 128 human tissue feature images with an image dimension of 400x400x1. They are then fused with the 64 first human tissue feature images with an image dimension of 400x400x1 output by the second downsampling layer in the ResNet network through a concatenation layer, and then convolution operations with a size of 3x3 and a step size of 1 are performed through the convolution kernels in the two convolutional layers in turn, outputting 64 second human tissue feature images with an image dimension of 400x400x1.

[0055] Subsequently, in the fifth upsampling layer, the 64 second human tissue feature images with an image dimension of 400x400x1 are upsampled by a deconvolution layer with a size of 2 x 2 and a step size of 2 to obtain 64 human tissue feature images with an image dimension of 800x800x1. They are then fused with the 32 first human tissue feature images with an image dimension of 800x800x1 output by the first downsampling layer in the ResNet network through a concatenation layer, and then convolution operations with a size of 3x3 and a step size of 1 are performed through the convolution kernels in the two convolutional layers in turn, outputting 32 second human tissue feature images with an image dimension of 800x800x1.

[0056] Finally, in the fully connected layer, the 32 second human tissue feature images with an image dimension of 800x800x1 are subjected to a point convolution operation with a size of 1x1 and a step size of 1 through the point convolution layer, and then semantic segmentation is performed through the multi-category label mapping layer to output a human tissue distribution image for representing the pixel position information of different human tissues in the human tissue ultrasound image; wherein the pixel position information of different human tissues in the human tissue distribution image is represented by different identifiers.

[0057] In the embodiment of the present application, after the pixel position information of different human tissues in the human tissue ultrasound image is identified by the human tissue recognition model, based on the target imaging depth, the pixel position information of different human tissues in the human tissue ultrasound image is converted into the spatial depth information of different human tissues in the current treatment area. Based on the spatial depth information of different human tissues in the current treatment area, when the prohibited treatment tissue whose spatial depth information overlaps with the initial treatment depth is detected in the different human tissues, it is determined that the current treatment area has prohibited treatment tissue at the initial treatment depth, and the prohibited treatment tissue in the current treatment area at the initial treatment depth is prompted for the risk of injury through a pop-up window and / or voice; at the same time, the pixel position information and / or spatial depth information of different human tissues are annotated and displayed to facilitate the operator to view the distribution of human tissues in the current treatment area and adjust the initial treatment depth; in addition, after determining that the current treatment area has prohibited treatment tissue at the initial treatment depth, the initial treatment depth of the current treatment area can be adjusted to obtain the target treatment depth of the current treatment area based on the spatial depth information of the prohibited treatment tissue in the current treatment area at the initial treatment depth to avoid the depth range where the prohibited treatment tissue is located, thereby achieving automatic adjustment of the treatment depth and improving the treatment effect and intelligence of the ultrasonic beauty device.

[0058] It is worth mentioning that in the embodiments of the present application, the ultrasonic beauty device usually predetermines an initial treatment line for the target treatment site (such as the face) of the target patient, and the initial treatment line includes different treatment areas of the target treatment site and the initial treatment depth corresponding to each treatment area. The current treatment area mentioned above can be the treatment area currently undergoing skin treatment in the initial treatment line, or it can be the next treatment area in the initial treatment line that is about to undergo skin treatment. That is, while treating one of the treatment areas in the initial treatment line, the next treatment area is subjected to interference detection between the initial treatment depth and the prohibited treatment tissue, thereby improving the treatment effect while also improving the treatment efficiency.

[0059] The following is a detailed description of the human tissue recognition model training method in the embodiment of the present application. Figure 4 As shown, the specific process of the human tissue recognition model training method provided in the embodiment of the present application is as follows: Step 401: Acquire a sample set of a target human body part (e.g., face); wherein the sample set includes a plurality of sample groups, each sample group includes an original human tissue ultrasound image and a standard mask image of the target human body part (e.g., face), the standard mask image being a binary mask image or a grayscale mask image; each pixel value in the binary mask image is 0 or 1, 1 represents the target human tissue, and 0 represents other human tissues except the target human tissue; each pixel value in the grayscale mask image is an integer value between 0 and 255, and different integer values ​​represent different human tissues.

[0060] In specific implementation, when obtaining a sample set of a target human body part (such as a face), the following methods may be used but are not limited to: First, multiple original human tissue ultrasound images are collected; each of which contains relatively rich information on various types of human tissues, such as bones, blood vessels, dermis, etc. under the skin of the target human body part (such as the face).

[0061] Then, Labelme software is used to mark the pixel position information of different human tissues for each original human tissue ultrasound image to generate a standard mask image. Taking the standard mask image as a binary mask image as an example, each pixel point in the position area where the target human tissue is located in the original human tissue ultrasound image is marked as 1, and each pixel point in the position area where the remaining human tissues are located is marked as 0, thereby generating a binary mask image with the same image size as the original human tissue ultrasound image, and adding a target human tissue label to the binary mask image. For example, assuming that the target human tissue is bone, a bone label is added to the binary mask image to facilitate multi-label mapping processing (i.e., semantic segmentation).

[0062] Finally, a plurality of sample groups are formed based on a plurality of original human tissue ultrasound images and a plurality of standard mask images of the original human tissue ultrasound images to obtain a sample set of a target human body part (eg, face).

[0063] Step 402: After performing image enhancement processing on a sample set of a target human body part (eg, face), the sample set after image enhancement processing is randomly divided into a training sample set and a test sample set.

[0064] Step 403: training the human tissue recognition model based on the training sample set.

[0065] In specific implementation, when training the human tissue recognition model based on the training sample set, the following methods may be used but are not limited to: Iteratively perform a training operation on the initial human tissue recognition model based on the training sample set until it is determined that an iteration termination condition is met (for example, the number of iterations is greater than a first threshold and / or the current loss value is less than a second threshold, etc.), and obtain the human tissue recognition model based on the weights of the ResNet network and the UNet network in the initial human tissue recognition model updated when the training operation was last performed; wherein each training operation performed includes: Select a target sample group from the training sample set; Inputting the original human tissue ultrasound image in the target sample group into the initial human tissue recognition model to obtain a human tissue distribution image for representing pixel position information of different human tissues in the original human tissue ultrasound image; calculating a current loss value based on the human tissue distribution image and a standard mask image in the target sample group; Update the weights of the ResNet network and UNet network in the initial human tissue recognition model based on the current loss value.

[0066] Step 404: Evaluate the trained human tissue recognition model based on the test sample set.

[0067] Step 405: determine the human tissue recognition model that has passed the evaluation as the final human tissue recognition model and save it to a computer storage medium.

[0068] Based on the above embodiments, the present application provides a human tissue ultrasonic image processing device, which is applied to ultrasonic beauty equipment. Figure 5 As shown, the human tissue ultrasonic image processing device 500 provided in the embodiment of the present application at least includes: The ultrasonic image acquisition unit 501 is used to acquire an ultrasonic image of human tissue in the current treatment area detected according to the target imaging depth; wherein the target imaging depth is greater than the initial treatment depth of the current treatment area; The human tissue recognition unit 502 is used to perform human tissue recognition on the human tissue ultrasound image based on the human tissue recognition model to obtain pixel position information of different human tissues in the human tissue ultrasound image; wherein the human tissue recognition model includes a ResNet network and a UNet network connected in sequence; the human tissue recognition model performs feature extraction processing on the human tissue ultrasound image through the ResNet network to obtain first human tissue feature images of different first semantic levels, performs feature enhancement processing on the first human tissue feature images of each first semantic level through the UNet network to obtain second human tissue feature images of different second semantic levels, and then performs semantic segmentation based on the second human tissue feature images of each second semantic level to obtain pixel position information of different human tissues in the human tissue ultrasound image; The injury risk warning unit 503 is used to determine, based on the pixel position information of different human tissues in the human tissue ultrasound image, that there is prohibited treatment tissue in the current treatment area at the initial treatment depth, and to provide an injury risk warning based on the prohibited treatment tissue in the current treatment area at the initial treatment depth.

[0069] In a possible implementation, the ResNet network includes a plurality of downsampling layers connected in sequence; the image resolution output by each downsampling layer decreases exponentially, the number of feature channels output by the first downsampling layer is the same as the number of feature channels output by the second downsampling layer, and the number of feature channels output by the second and subsequent downsampling layers increases exponentially; The UNet network includes an adaptation layer, multiple upsampling layers, and a fully connected layer connected in sequence; the multiple upsampling layers are jump-connected to the multiple downsampling layers of the ResNet network in a one-to-one correspondence; the image resolution output by each upsampling layer increases exponentially and the number of feature channels decreases exponentially.

[0070] In a possible implementation, the human tissue recognition unit 502 is used to sequentially downsample and dimensionally reduce the human tissue ultrasound image through the first downsampling layer in the ResNet network to obtain a first human tissue feature image at a high-resolution level; iteratively downsample the first human tissue feature image at a high-resolution level through the second and subsequent downsampling layers in the ResNet network to obtain first human tissue feature images at different low-resolution levels; and based on the first human tissue feature image at the high-resolution level and the first human tissue feature images at different low-resolution levels, obtain first human tissue feature images at different first semantic levels.

[0071] In a possible implementation, the human tissue recognition unit 502 is used to adapt the first human tissue feature image of the first semantic level output by the last downsampling layer in the ResNet network through the adaptation layer in the UNet network to obtain the first human tissue feature image of the first semantic level with the target feature channel number; after upsampling the first human tissue feature image of the first semantic level with the target feature channel number through the first upsampling layer in the UNet network, perform feature fusion processing with the first human tissue feature image of the first semantic level output by the downsampling layer in the corresponding jump-connected ResNet network to obtain the first human tissue feature image of the second semantic level. a second human tissue feature image of the second semantic level output by the previous upsampling layer; for the second and subsequent upsampling layers in the UNet network, the second human tissue feature image of the second semantic level output by the previous upsampling layer is upsampled through the upsampling layer, and then feature fused with the first human tissue feature image of the first semantic level output by the downsampling layer in the corresponding jump-connected ResNet network to obtain a second human tissue feature image of the second semantic level; and the second human tissue feature image of the second semantic level output by the last upsampling layer is semantically segmented through the fully connected layer in the UNet network to obtain pixel position information of different human tissues in the human tissue ultrasound image.

[0072] In a possible implementation, the injury risk warning unit 503 is used to convert the pixel position information of different human tissues in the human tissue ultrasound image into spatial depth information of the different human tissues in the current treatment area based on the target imaging depth; based on the spatial depth information of the different human tissues in the current treatment area, when it is detected that there is prohibited treatment tissue in the different human tissues whose spatial depth information overlaps with the initial treatment depth, it is determined that there is prohibited treatment tissue in the current treatment area at the initial treatment depth.

[0073] In a possible implementation manner, the human tissue ultrasonic image processing device 500 provided in the embodiment of the present application further includes: The treatment line adjustment unit 504 is used to adjust the initial treatment depth of the current treatment area based on the spatial depth information of the prohibited treatment tissue in the current treatment area at the initial treatment depth to obtain the target treatment depth of the current treatment area.

[0074] In a possible implementation manner, the human tissue ultrasonic image processing device 500 provided in the embodiment of the present application further includes: The marking and displaying unit 505 is used to mark and display the pixel position information of different human tissues in the human tissue ultrasound image.

[0075] It should be noted that the principle of solving the technical problem by the above-mentioned human tissue ultrasonic image processing device 500 provided in the embodiment of the present application is similar to the above-mentioned human tissue ultrasonic image processing method provided in the embodiment of the present application. Therefore, the implementation of the above-mentioned human tissue ultrasonic image processing device 500 provided in the embodiment of the present application can refer to the implementation of the above-mentioned human tissue ultrasonic image processing method provided in the embodiment of the present application, and the repeated parts will not be repeated.

[0076] After introducing the human tissue ultrasonic image processing method and device provided in the embodiment of the present application, the ultrasonic beauty equipment provided in the embodiment of the present application is briefly introduced next.

[0077] See also Figure 6 As shown, the ultrasonic beauty device 600 provided in the embodiment of the present application includes at least a processor 601, a memory 602, and a computer program stored in the memory 602 and executable on the processor 601. When the processor 601 executes the computer program, the above-mentioned human tissue ultrasonic image processing method provided in the embodiment of the present application is implemented.

[0078] The ultrasonic beauty device 600 provided in the embodiment of the present application may also include a bus 603 connecting different components (including the processor 601 and the memory 602). The bus 603 represents one or more of several types of bus structures, including a memory bus, a peripheral bus, a local bus, and the like.

[0079] The memory 602 may include a readable medium in the form of a volatile memory, such as a random access memory 6021 and / or a cache memory 6022, and may further include a read-only memory 6023. The memory 602 may also include a program tool 6025 having a set (at least one) of program modules 6024, the program modules 6024 including but not limited to an operating subsystem, one or more application programs, other program modules, and program data, each of which or some combination thereof may include the implementation of a network environment.

[0080] The processor 601 may be a processing element or a collective term for multiple processing elements. For example, the processor 601 may be a microcontroller unit (MCU), a central processing unit (CPU), or one or more integrated circuits configured to implement the above-mentioned human tissue ultrasound image processing method provided in the embodiment of the present application. Specifically, the processor 601 may be a general-purpose processor, including but not limited to a CPU, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.

[0081] The ultrasonic beauty device 600 can also communicate with one or more devices (such as mobile phones, computers, etc.) that allow users to interact with the ultrasonic beauty device 600, and / or with various external devices 604 such as devices (such as routers, modems, etc.) that allow the ultrasonic beauty device 600 to communicate with one or more other ultrasonic beauty devices. Such communication can be performed through the input / output interface 605. In addition, the ultrasonic beauty device 600 can also communicate with one or more networks (such as local area networks (LAN), wide area networks (WAN) and / or public networks, such as the Internet) through a network adapter 606. Figure 6 As shown, the network adapter 606 communicates with other modules of the ultrasonic beauty device 600 via the bus 603. It should be understood that although Figure 6 Not shown, other hardware and / or software modules may be used in conjunction with the ultrasonic cosmetic device 600, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, disk arrays (Redundant Arrays of Independent Disks, RAID) subsystems, tape drives, and data backup storage subsystems.

[0082] It should be noted that Figure 6 The ultrasonic beauty device 600 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0083] In addition, the embodiment of the present application further provides a computer-readable storage medium, which stores computer instructions, and when the computer instructions are executed by the processor, the human tissue ultrasonic image processing method provided in the embodiment of the present application is implemented. Specifically, the computer instructions can be built-in or installed in the processor, so that the processor can implement the above-mentioned human tissue ultrasonic image processing method provided in the embodiment of the present application by executing the built-in or installed computer instructions.

[0084] In addition, the human tissue ultrasonic image processing method provided in the embodiment of the present application can also be implemented as a program product, which includes a program code. When the program code is executed by a processor, the above-mentioned human tissue ultrasonic image processing method provided in the embodiment of the present application is implemented.

[0085] The program product provided in the embodiments of the present application may adopt any combination of one or more readable media, wherein the readable medium may be a readable signal medium or a readable storage medium, and the readable storage medium may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination of the above. Specifically, more specific examples of readable storage media (a non-exhaustive list) include an electrical connection with one or more wires, a portable disk, a hard disk, RAM, ROM, Erasable Programmable Read Only Memory (EPROM), optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0086] The program product provided in the embodiment of the present application may adopt a CD-ROM and include program code, and may also be run on an ultrasonic beauty device. However, the program product provided in the embodiment of the present application is not limited thereto. In the embodiment of the present application, the readable storage medium may be any tangible medium containing or storing a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device.

[0087] It should be noted that, although several units or subunits of the device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided into multiple units to be embodied.

[0088] In addition, although the operations of the method of the present application are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0089] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0090] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for processing ultrasonic images of human tissue, characterized in that: Applied to ultrasonic beauty equipment, including: Acquire an ultrasonic image of human tissue in a current treatment area detected at a target imaging depth; wherein the target imaging depth is greater than an initial treatment depth of the current treatment area; Based on the human tissue recognition model, human tissue recognition is performed on the human tissue ultrasound image to obtain pixel position information of different human tissues in the human tissue ultrasound image; wherein the human tissue recognition model includes a ResNet network and a UNet network connected in sequence; the human tissue recognition model performs feature extraction processing on the human tissue ultrasound image through the ResNet network to obtain first human tissue feature images of different first semantic levels, performs feature enhancement processing on each first human tissue feature image of the first semantic level through the UNet network to obtain second human tissue feature images of different second semantic levels, and then performs semantic segmentation based on each second human tissue feature image of the second semantic level to obtain pixel position information of different human tissues in the human tissue ultrasound image; When it is determined based on the pixel position information of different human tissues in the human tissue ultrasound image that there is prohibited treatment tissue in the current treatment area at the initial treatment depth, an injury risk warning is provided based on the prohibited treatment tissue in the current treatment area at the initial treatment depth.

2. The human tissue ultrasonic image processing method according to claim 1, characterized in that: The ResNet network includes a plurality of downsampling layers connected in sequence; the image resolution output by each of the downsampling layers decreases exponentially, the number of feature channels output by the first downsampling layer is the same as the number of feature channels output by the second downsampling layer, and the number of feature channels output by the second and subsequent downsampling layers increases exponentially; The UNet network includes an adaptation layer, multiple upsampling layers, and a fully connected layer connected in sequence; the multiple upsampling layers are jump-connected to the multiple downsampling layers of the ResNet network in a one-to-one correspondence; the image resolution output by each upsampling layer increases exponentially and the number of feature channels decreases exponentially.

3. The method for processing human tissue ultrasonic images according to claim 2, characterized in that: By using the ResNet network, the human tissue ultrasound image is subjected to feature extraction processing to obtain first human tissue feature images of different first semantic levels, including: The human tissue ultrasound image is sequentially downsampled and dimensionally reduced through the first downsampling layer in the ResNet network to obtain a first human tissue feature image at a high-resolution level; Iteratively downsampling the first human tissue feature image at the high-resolution level through the second and subsequent downsampling layers in the ResNet network to obtain first human tissue feature images at different low-resolution levels; Based on the first human tissue feature image at the high-resolution level and the first human tissue feature images at different low-resolution levels, first human tissue feature images at different first semantic levels are obtained.

4. The method for processing human tissue ultrasonic images according to claim 2, characterized in that: After performing feature enhancement processing on each of the first human tissue feature images at the first semantic level to obtain second human tissue feature images at different second semantic levels through the UNet network, semantic segmentation is performed based on each of the second human tissue feature images at the second semantic level to obtain pixel position information of different human tissues in the human tissue ultrasound image, including: The first human tissue feature image of the first semantic level output by the last downsampling layer in the ResNet network is adapted through the adaptation layer in the UNet network to obtain the first human tissue feature image of the first semantic level with the target feature channel number; After upsampling the first human tissue feature image of the first semantic level of the target feature channel number through the first upsampling layer in the UNet network, feature fusion processing is performed with the first human tissue feature image of the first semantic level output by the downsampling layer in the ResNet network with corresponding jump connection to obtain a second human tissue feature image of the second semantic level; For the second and subsequent upsampling layers in the UNet network, after upsampling the second human tissue feature image of the second semantic level output by the previous upsampling layer, the upsampling layer is used to perform feature fusion processing with the first human tissue feature image of the first semantic level output by the downsampling layer in the ResNet network connected to the corresponding jump layer, to obtain the second human tissue feature image of the second semantic level; Through the fully connected layer in the UNet network, semantic segmentation is performed on the second human tissue feature image of the second semantic level output by the last upsampling layer to obtain pixel position information of different human tissues in the human tissue ultrasound image.

5. The method for processing human tissue ultrasonic images according to any one of claims 1 to 4, characterized in that: Determining, based on pixel position information of different human tissues in the human tissue ultrasound image, that there is a prohibited treatment tissue in the current treatment area at the initial treatment depth, comprises: Based on the target imaging depth, converting pixel position information of different human tissues in the human tissue ultrasound image into spatial depth information of the different human tissues in the current treatment area; Based on the spatial depth information of the different human tissues in the current treatment area, when it is detected that there is prohibited treatment tissue in the different human tissues whose spatial depth information overlaps with the initial treatment depth, it is determined that there is prohibited treatment tissue in the current treatment area at the initial treatment depth.

6. The method for processing human tissue ultrasonic images according to claim 5, characterized in that: When it is determined based on the pixel position information of different human tissues in the human tissue ultrasound image that there is a prohibited treatment tissue in the current treatment area at the initial treatment depth, the method further includes: Based on the spatial depth information of the prohibited treatment tissue of the current treatment area at the initial treatment depth, the initial treatment depth of the current treatment area is adjusted to obtain the target treatment depth of the current treatment area.

7. The method for processing human tissue ultrasonic images according to claim 1, characterized in that: After performing human tissue recognition on the human tissue ultrasonic image based on the human tissue recognition model to obtain pixel position information of different human tissues in the human tissue ultrasonic image, the method further includes: The pixel position information of different human tissues in the human tissue ultrasound image is annotated and displayed.

8. A human tissue ultrasonic image processing device, characterized in that: Applied to ultrasonic beauty equipment, including: An ultrasonic image acquisition unit, used to acquire an ultrasonic image of human tissue in a current treatment area detected according to a target imaging depth; wherein the target imaging depth is greater than an initial treatment depth of the current treatment area; A human tissue recognition unit, configured to perform human tissue recognition on the human tissue ultrasound image based on a human tissue recognition model to obtain pixel position information of different human tissues in the human tissue ultrasound image; wherein the human tissue recognition model comprises a ResNet network and a UNet network connected in sequence; the human tissue recognition model performs feature extraction processing on the human tissue ultrasound image through the ResNet network to obtain first human tissue feature images of different first semantic levels, performs feature enhancement processing on each first human tissue feature image of the first semantic level through the UNet network to obtain second human tissue feature images of different second semantic levels, and then performs semantic segmentation based on each second human tissue feature image of the second semantic level to obtain pixel position information of different human tissues in the human tissue ultrasound image; An injury risk warning unit is used to determine, based on pixel position information of different human tissues in the human tissue ultrasound image, that there is prohibited treatment tissue in the current treatment area at the initial treatment depth, and to provide an injury risk warning based on the prohibited treatment tissue in the current treatment area at the initial treatment depth.

9. An ultrasonic beauty device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for processing human tissue ultrasonic images as claimed in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method for processing human tissue ultrasound images according to any one of claims 1 to 7 is implemented.

Images