Vascular ultrasound data processing method and device, terminal equipment and medium

Abstract

The invention discloses a blood vessel ultrasonic data processing method and device, terminal equipment and a medium, and the method comprises the steps: inputting a plurality of to-be-trained ultrasonic images marked with brachial artery contours into a generative adversarial network for training, and obtaining a contour recognition model, thereby achieving the recognition of the to-be-recognized brachial artery contours when the to-be-recognized brachial artery ultrasonic images are obtained, and achieving the recognition of the to-be-recognized brachial artery contours. Inputting the ultrasonic image into the contour recognition model to obtain a brachial artery positioning map of the brachial artery to be recognized, and performing pixel adjustment on the ultrasonic image based on the diameter data of the artery and the brachial artery positioning map to obtain a target blood vessel map of the brachial artery to be recognized, and finally, inputting the target blood vessel map into a pre-trained blood vessel dilatation screening neural network for blood vessel dilatation screening to obtain a blood vessel dilatation type. Compared with the prior art that brachial artery detection is carried out by applying external pressure, brachial artery detection can be carried out through the brachial artery-based ultrasonic image, tedious external force operation is not needed, and the efficiency of vasodilatation detection is improved.

Description

Technical Field

[0001] The present invention relates to the field of deep learning, and in particular to a method, device, terminal equipment and medium for processing vascular ultrasound data. Background Art

[0002] Endothelial function can be assessed non-invasively using the brachial artery flow-mediated dilation (FMD) examination technique. The principle is to assess the degree of relaxation of vascular smooth muscle by measuring the percentage of change in the brachial artery under shear force. Specifically, an ultrasound probe is used to measure the inner diameter of the brachial artery of the subject in a resting state, and then an inflatable pressure cuff is fixed to the forearm distal to the ultrasound probe. After five minutes of pressure, the pressure is deflated and the change in the inner diameter of the brachial artery is measured.

[0003] In most areas, vasodilation examinations are usually performed manually. Since this operation requires certain skills and experience, there are great differences in the operation of FMD. There may be measurement heterogeneity between experienced ultrasound physicians and residents, which makes the repeatability of FMD worse and makes comparison between studies difficult, resulting in low efficiency of vasodilation detection.

[0004] Therefore, a processing strategy for vascular ultrasound data is urgently needed to solve the problem of low efficiency in vascular dilation detection. Summary of the invention

[0005] The embodiments of the present invention provide a method, an apparatus, a terminal device and a medium for processing vascular ultrasound data to improve the efficiency of vascular dilation detection.

[0006] In order to solve the above problem, an embodiment of the present invention provides a method for processing vascular ultrasound data, comprising:

[0007] Acquire an ultrasound image of the brachial artery to be identified;

[0008] Inputting the ultrasound image into a preset contour recognition model to obtain a brachial artery positioning map of the brachial artery to be identified; wherein the contour recognition model is obtained by inputting a number of ultrasound images to be trained that are marked with brachial artery contours into a generative adversarial network for training;

[0009] Based on the diameter data of the artery and the brachial artery positioning map, the ultrasonic image is pixel-adjusted to obtain a target blood vessel map of the brachial artery to be identified.

[0010] As an improvement of the above scheme, the training method of the vasodilation screening neural network includes:

[0011] Acquire several brachial artery ultrasound images of existing vascular dilation patterns;

[0012] Determine a plurality of blood vessel maps to be trained corresponding to the brachial artery ultrasound images;

[0013] Marking each of the target blood vessel images to be trained with a blood vessel dilation type to obtain a plurality of blood vessel images to be trained with blood vessel dilation types marked;

[0014] Each of the blood vessel images to be trained is input into a convolutional neural network for training to obtain a blood vessel dilation screening neural network.

[0015] As an improvement of the above solution, the training of the contour recognition model includes:

[0016] Acquire a number of ultrasound images to be trained;

[0017] Annotating the contour of the brachial artery for each of the ultrasound images to be trained to obtain an annotated image of each of the ultrasound images to be trained;

[0018] Each of the ultrasound images to be trained and the corresponding annotated images form a first training sample set; wherein the label of the first training sample is Ture;

[0019] According to the plurality of ultrasound images to be trained and the first training sample set, the model training operation is repeatedly performed until the label prediction data output by the discriminator is greater than or equal to the prediction threshold, and the contour recognition model is output.

[0020] As an improvement of the above solution, the generative adversarial network includes: a generator and a discriminator; the model training operation includes:

[0021] Inputting a plurality of ultrasound images to be trained into the generator respectively to predict the contour of the brachial artery to obtain a predicted image of each ultrasound image to be trained;

[0022] Each of the ultrasound images to be trained and the corresponding predicted image form a second training sample; wherein the label of the second training sample is False;

[0023] Inputting the first training sample and the second training sample into a discriminator to obtain a first predicted label of the first training sample and a second predicted label of the second training sample;

[0024] If the first predicted label is True and the second predicted label is False, the generator network parameters for subsequent model training operations are adjusted; otherwise, the discriminator network parameters for subsequent model training operations are adjusted.

[0025] As an improvement of the above solution, the pixel adjustment of the ultrasonic image based on the diameter data of the artery and the brachial artery positioning map to obtain the target blood vessel map of the brachial artery to be identified includes:

[0026] Acquiring blood vessel position data in the ultrasonic image according to the brachial artery positioning map and the ultrasonic image;

[0027] Determine the target blood vessel area according to the arterial diameter data and blood vessel position data;

[0028] In the ultrasonic image, the three-color values ​​of the pixels in the non-target blood vessel area are set as the three-color thresholds to obtain the target blood vessel map of the brachial artery to be identified.

[0029] Accordingly, an embodiment of the present invention further provides a device for processing blood vessel ultrasound data, comprising: a data acquisition module, a model input module, a blood vessel positioning module and a result generation module;

[0030] The data acquisition module is used to acquire an ultrasonic image of the brachial artery to be identified;

[0031] The model input module is used to input the ultrasound image into a preset contour recognition model to obtain a brachial artery positioning map of the brachial artery to be identified; wherein the contour recognition model is obtained by inputting a number of ultrasound image training data marked with brachial artery contours into a generative adversarial network for training;

[0032] The result generation module is used to adjust the pixels of the ultrasonic image based on the diameter data of the artery and the brachial artery positioning map to obtain the target blood vessel map of the brachial artery to be identified;

[0033] The result generation module is used to input the target blood vessel map into a preset vasodilation screening neural network to perform vasodilation screening and obtain the vasodilation type; wherein the vasodilation screening neural network is obtained by inputting the blood vessel map to be trained marked with the vasodilation type into a convolutional neural network for training.

[0034] As an improvement of the above scheme, the training method of the vasodilation screening neural network includes:

[0035] Acquire several brachial artery ultrasound images of existing vascular dilation patterns;

[0036] Determine a plurality of blood vessel maps to be trained corresponding to the brachial artery ultrasound images;

[0037] Marking each of the target blood vessel images to be trained with a blood vessel dilation type to obtain a plurality of blood vessel images to be trained with blood vessel dilation types marked;

[0038] Each of the blood vessel images to be trained is input into a convolutional neural network for training to obtain a blood vessel dilation screening neural network.

[0039] As an improvement of the above solution, the training of the contour recognition model includes:

[0040] Acquire a number of ultrasound images to be trained;

[0041] Annotating the contour of the brachial artery for each of the ultrasound images to be trained to obtain an annotated image of each of the ultrasound images to be trained;

[0042] Each of the ultrasound images to be trained and the corresponding annotated images form a first training sample set; wherein the label of the first training sample is Ture;

[0043] According to the plurality of ultrasound images to be trained and the first training sample set, the model training operation is repeatedly performed until the label prediction data output by the discriminator is greater than or equal to the prediction threshold, and the contour recognition model is output.

[0044] As an improvement of the above solution, the generative adversarial network includes: a generator and a discriminator; the model training operation includes:

[0045] Inputting a plurality of ultrasound images to be trained into the generator respectively to predict the contour of the brachial artery to obtain a predicted image of each ultrasound image to be trained;

[0046] Each of the ultrasound images to be trained and the corresponding predicted image form a second training sample; wherein the label of the second training sample is False;

[0047] Inputting the first training sample and the second training sample into a discriminator to obtain a first predicted label of the first training sample and a second predicted label of the second training sample;

[0048] If the first predicted label is True and the second predicted label is False, the generator network parameters for subsequent model training operations are adjusted; otherwise, the discriminator network parameters for subsequent model training operations are adjusted.

[0049] As an improvement of the above solution, the pixel adjustment of the ultrasonic image based on the diameter data of the artery and the brachial artery positioning map to obtain the target blood vessel map of the brachial artery to be identified includes:

[0050] Acquiring blood vessel position data in the ultrasonic image according to the brachial artery positioning map and the ultrasonic image;

[0051] Determine the target blood vessel area according to the arterial diameter data and blood vessel position data;

[0052] In the ultrasonic image, the three-color values ​​of the pixels in the non-target blood vessel area are set as the three-color thresholds to obtain the target blood vessel map of the brachial artery to be identified.

[0053] Correspondingly, an embodiment of the present invention further provides a computer terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a method for processing vascular ultrasound data as described in the present invention is implemented.

[0054] Correspondingly, an embodiment of the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a method for processing vascular ultrasound data as described in the present invention.

[0055] As can be seen from the above, the present invention has the following beneficial effects:

[0056] The present invention provides a method for processing vascular ultrasound data, by inputting a number of ultrasound images to be trained with brachial artery contours into a generative adversarial network for training, and obtaining a contour recognition model, so that when an ultrasound image of the brachial artery to be identified is obtained, the ultrasound image is input into the contour recognition model to obtain a brachial artery positioning map of the brachial artery to be identified, and based on the diameter data of the artery and the brachial artery positioning map, the ultrasound image is pixel-adjusted to obtain a target blood vessel map of the brachial artery to be identified, and finally the target blood vessel map is input into a pre-trained vascular dilation screening neural network for vascular dilation screening to obtain the type of vascular dilation. The present application identifies the target blood vessel of the brachial artery to be identified by training a generative adversarial network, and determines the type of vascular dilation based on the characteristics of the blood vessel. Compared with the prior art of applying external pressure to detect the brachial artery, the present application can detect the brachial artery by using an ultrasound image based on the brachial artery, without the need for cumbersome external force operations, thereby improving the efficiency of vascular dilation detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a flow chart of a method for processing vascular ultrasound data provided by an embodiment of the present invention;

[0058] Figure 2 is a schematic diagram of the structure of a device for processing vascular ultrasound data provided by an embodiment of the present invention;

[0059] Figure 3 is a flow chart of a method for processing vascular ultrasound data provided by another embodiment of the present invention;

[0060] Figure 4is a schematic diagram of an ultrasonic image provided by an embodiment of the present invention;

[0061] Figure 5 is a schematic diagram of a brachial artery localization map provided by an embodiment of the present invention;

[0062] Figure 6 is a schematic diagram of a target blood vessel map provided by an embodiment of the present invention;

[0063] Figure 7 It is a schematic diagram of the structure of a terminal device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] Embodiment 1

[0066] See also Figure 1 , Figure 1 FIG. 1 is a flow chart of a method for processing vascular ultrasound data provided by an embodiment of the present invention. Figure 1 As shown, this embodiment includes steps 101 to 104, and each step is specifically as follows:

[0067] Step 101: Obtain an ultrasonic image of the brachial artery to be identified.

[0068] In this embodiment,

[0069] Step 102: Input the ultrasound image into a preset contour recognition model to obtain a brachial artery positioning map of the brachial artery to be identified; wherein the contour recognition model is obtained by inputting a number of ultrasound images to be trained that are marked with brachial artery contours into a generative adversarial network for training.

[0070] In this embodiment, the training of the contour recognition model includes:

[0071] Acquire a number of ultrasound images to be trained;

[0072] Annotating the contour of the brachial artery for each of the ultrasound images to be trained to obtain an annotated image of each of the ultrasound images to be trained;

[0073] Each of the ultrasound images to be trained and the corresponding annotated images form a first training sample set; wherein the label of the first training sample is Ture;

[0074] According to the plurality of ultrasound images to be trained and the first training sample set, the model training operation is repeatedly performed until the label prediction data output by the discriminator is greater than or equal to the prediction threshold, and the contour recognition model is output.

[0075] In this embodiment, the generative adversarial network includes: a generator and a discriminator; the model training operation includes:

[0076] Inputting a plurality of ultrasound images to be trained into the generator respectively to predict the contour of the brachial artery to obtain a predicted image of each ultrasound image to be trained;

[0077] Each of the ultrasound images to be trained and the corresponding predicted image form a second training sample; wherein the label of the second training sample is False;

[0078] Inputting the first training sample and the second training sample into a discriminator to obtain a first predicted label of the first training sample and a second predicted label of the second training sample;

[0079] If the first predicted label is True and the second predicted label is False, the generator network parameters for subsequent model training operations are adjusted; otherwise, the discriminator network parameters for subsequent model training operations are adjusted.

[0080] In a specific embodiment, to better illustrate the training of the contour recognition model, please refer to the following description:

[0081] Select a paired brachial artery ultrasound image x (i.e., the ultrasound image to be trained as described in the present invention) and its brachial artery contour image y (i.e., the annotated image as described in the present invention) from the annotated data set to obtain a training sample (x, y) with a label of True (i.e., the first training sample as described in the present invention);

[0082] The brachial artery ultrasound image x is used as the input of the generator (G), and after a series of convolution and deconvolution operations, the corresponding brachial artery contour image y' (i.e., the predicted image described in the present invention) is generated to construct a training sample (x, y') with a label of False (i.e., the second training sample described in the present invention);

[0083] The real sample (x, y) and the generated sample (x, y') are used as the input of the discriminator (D). After operations such as convolution and full connection, the label prediction data is output, and the prediction threshold is compared for the prediction data (where the size of the prediction threshold is adaptively adjusted according to the user's demand for recognition accuracy);

[0084] During the training process, G and D compete with each other to update their respective network parameters. For G, its optimization goal is to induce D to identify the (x, y') samples it generates as True, while the optimization goal of D is to correctly identify (x, y) and (x, y') as True and False.

[0085] After training, you can use G to identify blood vessels in brachial artery ultrasound images.

[0086] In a specific embodiment, the generative adversarial network design can use classic pix2pix, pix2pixHD and SPADE, etc. to achieve the conversion of brachial artery ultrasound image to brachial artery contour map.

[0087] Step 103: Based on the diameter data of the artery and the brachial artery positioning map, the ultrasonic image is pixel-adjusted to obtain a target blood vessel map of the brachial artery to be identified.

[0088] In this embodiment, the pixel adjustment of the ultrasound image based on the artery diameter data and the brachial artery positioning map to obtain the target blood vessel map of the brachial artery to be identified includes:

[0089] Acquiring blood vessel position data in the ultrasonic image according to the brachial artery positioning map and the ultrasonic image;

[0090] Determine the target blood vessel area according to the arterial diameter data and blood vessel position data;

[0091] In the ultrasonic image, the three-color values ​​of the pixels in the non-target blood vessel area are set as the three-color thresholds to obtain the target blood vessel map of the brachial artery to be identified.

[0092] In a specific embodiment, the brachial artery ultrasound image x and its brachial artery contour image y' are aligned to locate the blood vessel in x;

[0093] Eliminate the regional information between the blood vessels in x, specifically by setting the RGB value of the pixel therein to (255, 255, 255) (i.e., the three-color threshold of the present invention);

[0094] The regional information of the blood vessels in x that extend upward and downward to the blood vessel diameter is retained, and the information of other regions is eliminated, that is, the RGB values ​​of the pixels in these regions are set to (255, 255, 255);

[0095] By saving the changed image x, a cropped ROI image (ie, the target blood vessel image described in the present invention) can be obtained.

[0096] In a specific embodiment, (region of interest) ROI image cropping, combined with OpenCV image processing methods and medical knowledge, uses paired brachial artery ultrasound images and brachial artery contour maps to outline a region of interest that is roughly equal to the diameter range of the brachial artery within the ultrasound field of view, which helps to learn more accurate FMD pathological characteristics, thereby improving the model's prediction capabilities.

[0097] Step 104: inputting the target vascular image into a preset vascular dilation screening neural network to perform vascular dilation screening to obtain a vascular dilation type; wherein the vascular dilation screening neural network is obtained by inputting a vascular image to be trained that is labeled with a vascular dilation type into a convolutional neural network for training.

[0098] In this embodiment, the training method of the vasodilation screening neural network includes:

[0099] Acquire several brachial artery ultrasound images of existing vascular dilation patterns;

[0100] Determine a plurality of blood vessel maps to be trained corresponding to the brachial artery ultrasound images;

[0101] Marking each of the target blood vessel images to be trained with a blood vessel dilation type to obtain a plurality of blood vessel images to be trained with blood vessel dilation types marked;

[0102] Each of the blood vessel images to be trained is input into a convolutional neural network for training to obtain a blood vessel dilation screening neural network.

[0103] In a specific embodiment, a number of brachial artery ultrasound images of existing vascular dilation types are obtained, specifically: a brachial artery vascular dilation test is performed on the patient corresponding to each brachial artery ultrasound image, and the vascular dilation type corresponding to each test result is recorded and associated with the corresponding brachial artery ultrasound image.

[0104] In a specific embodiment, the vasodilation screening neural network can predict the state of endothelial function by identifying the image features of the brachial artery and its surrounding connective tissue when the brachial artery is not pressurized, thereby achieving the purpose of FMD screening.

[0105] In a specific embodiment, the convolutional neural network CNN classification can adopt classic VGG, ResNet and Efficientnet, etc.

[0106] In a specific embodiment, the model framework is as follows Figure 3As shown in the figure, in the first stage, the input is a high-resolution brachial artery ultrasound image, and the brachial artery contour map paired with it is obtained through a generative adversarial network; in the second stage, the ROI area is cropped out based on the brachial artery contour map and medical prior knowledge through OpebCV; in the third stage, CNN is used for the ROI area to output the FMD classification prediction result.

[0107] In a specific embodiment, see Figure 4 , 5 , 6, the input is a high-resolution brachial artery ultrasound image ( Figure 4 ), and the paired brachial artery contour map is obtained through the generative adversarial network ( Figure 5 ), combined with the brachial artery contour map and medical prior knowledge, the ROI area is cut out ( Figure 6 ).

[0108] See also Figure 2 , Figure 2 2 is a schematic diagram of the structure of a device for processing vascular ultrasound data provided by an embodiment of the present invention, comprising: a data acquisition module 201, a model input module 202, a vascular positioning module 203 and a result generation module 204;

[0109] The data acquisition module is used to acquire an ultrasonic image of the brachial artery to be identified;

[0110] The model input module is used to input the ultrasound image into a preset contour recognition model to obtain a brachial artery positioning map of the brachial artery to be identified; wherein the contour recognition model is obtained by inputting a number of ultrasound image training data marked with brachial artery contours into a generative adversarial network for training;

[0111] The result generation module is used to adjust the pixels of the ultrasonic image based on the diameter data of the artery and the brachial artery positioning map to obtain the target blood vessel map of the brachial artery to be identified;

[0112] The result generation module is used to input the target blood vessel map into a preset vasodilation screening neural network to perform vasodilation screening and obtain the vasodilation type; wherein the vasodilation screening neural network is obtained by inputting the blood vessel map to be trained marked with the vasodilation type into a convolutional neural network for training.

[0113] As an improvement of the above scheme, the training method of the vasodilation screening neural network includes:

[0114] Acquire several brachial artery ultrasound images of existing vascular dilation patterns;

[0115] Determine a plurality of blood vessel maps to be trained corresponding to the brachial artery ultrasound images;

[0116] Marking each of the target blood vessel images to be trained with a blood vessel dilation type to obtain a plurality of blood vessel images to be trained with blood vessel dilation types marked;

[0117] Each of the blood vessel images to be trained is input into a convolutional neural network for training to obtain a blood vessel dilation screening neural network.

[0118] As an improvement of the above solution, the training of the contour recognition model includes:

[0119] Acquire a number of ultrasound images to be trained;

[0120] Annotating the contour of the brachial artery for each of the ultrasound images to be trained to obtain an annotated image of each of the ultrasound images to be trained;

[0121] Each of the ultrasound images to be trained and the corresponding annotated images form a first training sample set; wherein the label of the first training sample is Ture;

[0122] According to the plurality of ultrasound images to be trained and the first training sample set, the model training operation is repeatedly performed until the label prediction data output by the discriminator is greater than or equal to the prediction threshold, and the contour recognition model is output.

[0123] As an improvement of the above solution, the generative adversarial network includes: a generator and a discriminator; the model training operation includes:

[0124] Inputting a plurality of ultrasound images to be trained into the generator respectively to predict the contour of the brachial artery to obtain a predicted image of each ultrasound image to be trained;

[0125] Each of the ultrasound images to be trained and the corresponding predicted image form a second training sample; wherein the label of the second training sample is False;

[0126] Inputting the first training sample and the second training sample into a discriminator to obtain a first predicted label of the first training sample and a second predicted label of the second training sample;

[0127] If the first predicted label is True and the second predicted label is False, the generator network parameters for subsequent model training operations are adjusted; otherwise, the discriminator network parameters for subsequent model training operations are adjusted.

[0128] As an improvement of the above solution, the pixel adjustment of the ultrasonic image based on the diameter data of the artery and the brachial artery positioning map to obtain the target blood vessel map of the brachial artery to be identified includes:

[0129] Acquiring blood vessel position data in the ultrasonic image according to the brachial artery positioning map and the ultrasonic image;

[0130] Determine the target blood vessel area according to the arterial diameter data and blood vessel position data;

[0131] In the ultrasonic image, the three-color values ​​of the pixels in the non-target blood vessel area are set as the three-color thresholds to obtain the target blood vessel map of the brachial artery to be identified.

[0132] In this embodiment, a plurality of ultrasound images to be trained with the contour of the brachial artery are input into a generative adversarial network for training to obtain a contour recognition model. When the ultrasound image of the brachial artery to be identified is obtained, the ultrasound image is input into the contour recognition model to obtain a brachial artery localization map of the brachial artery to be identified, and the ultrasound image is pixel-adjusted based on the diameter data of the artery and the brachial artery localization map to obtain a target vessel map of the brachial artery to be identified. Finally, the target vessel map is input into a pre-trained vascular dilation screening neural network for vascular dilation screening to obtain the type of vascular dilation. The present application identifies the target vessel of the brachial artery to be identified by training a generative adversarial network, and determines the type of vascular dilation based on the characteristics of the vessel. Compared with the prior art that performs brachial artery detection by applying external pressure, the present application can perform brachial artery detection based on the ultrasound image of the brachial artery, without the need for cumbersome external force operations, thereby improving the efficiency of vascular dilation detection.

[0133] Embodiment 2

[0134] See also Figure 7 , Figure 7 It is a schematic diagram of the structure of a terminal device provided in one embodiment of the present invention.

[0135] A terminal device of this embodiment includes: a processor 701, a memory 702, and a computer program stored in the memory 702 and executable on the processor 701. When the processor 701 executes the computer program, the steps of the above-mentioned method for processing vascular ultrasound data in the embodiment are implemented, for example Figure 1 Alternatively, when the processor executes the computer program, the functions of each module in the above-mentioned device embodiments are realized, for example: Figure 2 All modules of the device for processing vascular ultrasound data are shown.

[0136] In addition, an embodiment of the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method for processing vascular ultrasound data as described in any of the above embodiments.

[0137] Those skilled in the art will understand that the schematic diagram is merely an example of a terminal device and does not constitute a limitation on the terminal device. The terminal device may include more or fewer components than shown in the diagram, or a combination of certain components, or different components. For example, the terminal device may also include input and output devices, network access devices, buses, etc.

[0138] The processor 701 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor 701 is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0139] The memory 702 can be used to store the computer program and / or module, and the processor 701 implements various functions of the terminal device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory 702. The memory 702 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0140] Wherein, if the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0141] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0142] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for processing vascular ultrasound data, characterized in that: include: Acquire an ultrasound image of the brachial artery to be identified; Inputting the ultrasound image into a preset contour recognition model to obtain a brachial artery positioning map of the brachial artery to be identified; wherein the contour recognition model is obtained by inputting a number of ultrasound images to be trained that are marked with brachial artery contours into a generative adversarial network for training; Based on the diameter data of the artery and the brachial artery positioning map, pixel adjustment is performed on the ultrasound image to obtain a target blood vessel map of the brachial artery to be identified; The target blood vessel map is input into a preset vasodilation screening neural network for vasodilation screening to obtain the vasodilation type; wherein the vasodilation screening neural network is obtained by inputting a blood vessel map to be trained marked with vasodilation types into a convolutional neural network for training.

2. The method for processing vascular ultrasound data according to claim 1, characterized in that: The training method of the vasodilation screening neural network comprises: Acquire several brachial artery ultrasound images of existing vascular dilation patterns; Determine a plurality of blood vessel maps to be trained corresponding to the brachial artery ultrasound images; Marking each of the target blood vessel images to be trained with a blood vessel dilation type to obtain a plurality of blood vessel images to be trained with blood vessel dilation types marked; Each of the blood vessel images to be trained is input into a convolutional neural network for training to obtain a blood vessel dilation screening neural network.

3. The method for processing vascular ultrasound data according to claim 2, characterized in that: The training of the contour recognition model includes: Acquire a number of ultrasound images to be trained; Annotating the contour of the brachial artery for each of the ultrasound images to be trained to obtain an annotated image of each of the ultrasound images to be trained; Each of the ultrasound images to be trained and the corresponding annotated images form a first training sample set; wherein the label of the first training sample is Ture; According to the plurality of ultrasound images to be trained and the first training sample set, the model training operation is repeatedly performed until the label prediction data output by the discriminator is greater than or equal to the prediction threshold, and the contour recognition model is output.

4. The method for processing vascular ultrasound data according to claim 3, characterized in that: The generative adversarial network includes: a generator and a discriminator; the model training operation includes: Inputting a plurality of ultrasound images to be trained into the generator respectively to predict the contour of the brachial artery to obtain a predicted image of each ultrasound image to be trained; Each of the ultrasound images to be trained and the corresponding predicted image form a second training sample; wherein the label of the second training sample is False; Inputting the first training sample and the second training sample into a discriminator to obtain a first predicted label of the first training sample and a second predicted label of the second training sample; If the first predicted label is True and the second predicted label is False, the generator network parameters for subsequent model training operations are adjusted; otherwise, the discriminator network parameters for subsequent model training operations are adjusted.

5. The method for processing vascular ultrasound data according to claim 3, characterized in that: The step of adjusting the pixels of the ultrasound image based on the diameter data of the artery and the brachial artery positioning map to obtain the target blood vessel map of the brachial artery to be identified includes: Acquiring blood vessel position data in the ultrasonic image according to the brachial artery positioning map and the ultrasonic image; Determine the target blood vessel area according to the arterial diameter data and the blood vessel position data; In the ultrasonic image, the three-color values ​​of the pixels in the non-target blood vessel area are set as the three-color thresholds to obtain the target blood vessel map of the brachial artery to be identified.

6. A device for processing vascular ultrasound data, characterized in that: include: Data acquisition module, model input module, blood vessel positioning module and result generation module; The data acquisition module is used to acquire an ultrasonic image of the brachial artery to be identified; The model input module is used to input the ultrasound image into a preset contour recognition model to obtain a brachial artery positioning map of the brachial artery to be identified; wherein the contour recognition model is obtained by inputting a number of ultrasound image training data marked with brachial artery contours into a generative adversarial network for training; The result generation module is used to adjust the pixels of the ultrasonic image based on the diameter data of the artery and the brachial artery positioning map to obtain the target blood vessel map of the brachial artery to be identified; The result generation module is used to input the target blood vessel map into a preset vasodilation screening neural network to perform vasodilation screening and obtain the vasodilation type; wherein the vasodilation screening neural network is obtained by inputting the blood vessel map to be trained marked with the vasodilation type into a convolutional neural network for training.

7. The device for processing blood vessel ultrasound data according to claim 6, characterized in that: The training method of the vasodilation screening neural network comprises: Acquire several brachial artery ultrasound images of existing vascular dilation patterns; Determine a plurality of blood vessel maps to be trained corresponding to the brachial artery ultrasound images; Marking each of the target blood vessel images to be trained with a blood vessel dilation type to obtain a plurality of blood vessel images to be trained with blood vessel dilation types marked; Each of the blood vessel images to be trained is input into a convolutional neural network for training to obtain a blood vessel dilation screening neural network.

8. The device for processing blood vessel ultrasound data according to claim 7, characterized in that: The training of the contour recognition model includes: Acquire a number of ultrasound images to be trained; Annotating the contour of the brachial artery for each of the ultrasound images to be trained to obtain an annotated image of each of the ultrasound images to be trained; Each of the ultrasound images to be trained and the corresponding annotated images form a first training sample set; wherein the label of the first training sample is Ture; The model training operation is repeatedly performed according to the plurality of ultrasound images to be trained and the first training sample set until the label prediction data output by the discriminator is greater than or equal to the prediction threshold, and the contour recognition model is output.

9. A computer terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for processing vascular ultrasound data according to any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for processing vascular ultrasound data according to any one of claims 1 to 5.

Images