Section recognition method, image generation method, section recognition device and electronic device

Abstract

The invention relates to a section recognition method, an image generation method, a section recognition device and an electronic device, and the method comprises the steps: obtaining original detection data; inputting the original detection data into the trained section recognition model to obtain a section recognition result; and the section identification model is used for indicating a mapping relationship between the original detection data and the section type. According to the invention, the problem of low image processing efficiency is solved, and the automatic section identification method based on non-image data is realized.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a section recognition method, an image generation method, a section recognition device and an electronic device. Background Art

[0002] Ultrasound imaging is an indispensable tool in medical diagnosis, providing real-time visualization of the internal structure of the human body. Ultrasound imaging has become the most popular means of early screening for lesions because of its low price, non-invasive harm to patients, low radiation and high screening efficiency.

[0003] In the field of ultrasound imaging, automatic section or part recognition can effectively improve workflow and imaging efficiency. In related technologies, automatic section recognition methods are all based on image data, which requires the entire imaging process to be processed before section recognition can be performed. This increases the processing time of the ultrasound system in practice, resulting in low image processing efficiency.

[0004] Currently, no effective solution has been proposed for the problem of low efficiency of image processing in related technologies. Summary of the invention

[0005] The embodiments of the present application provide a section recognition method, an image generation method, a section recognition device and an electronic device, so as to at least solve the problem of low efficiency of image processing in the related art.

[0006] In a first aspect, an embodiment of the present application provides a section recognition method, the method comprising:

[0007] Get raw detection data;

[0008] The original detection data is input into the trained section recognition model to obtain the section recognition result; the section recognition model is used to indicate the mapping relationship between the original detection data and the section type.

[0009] In some embodiments, the section recognition model includes a first section recognition model; and inputting the original detection data into the trained section recognition model to obtain a section recognition result includes:

[0010] Performing a first data processing on the original detection data to obtain two-dimensional matrix data;

[0011] The two-dimensional matrix data is input into a first section recognition model to obtain the section recognition result.

[0012] In some embodiments, training the first section recognition model includes:

[0013] Get training probe data;

[0014] Processing the training detection data to obtain training matrix data;

[0015] Performing image generation processing according to the training detection data to obtain a training image, and determining a training section type corresponding to the training image;

[0016] A first training data set is generated according to the training matrix data and the training section type; the first training data set is input into a first initial model for training to generate the first section recognition model.

[0017] In some embodiments, the section recognition model includes a second section recognition model; and inputting the original detection data into the trained section recognition model to obtain the section recognition result includes:

[0018] Performing a second data processing on the original detection data to obtain an original detection sequence;

[0019] The original detection sequence is input into a second section recognition model to obtain the section recognition result.

[0020] In some embodiments, training the second section recognition model includes:

[0021] Get training probe data;

[0022] Processing the training detection data to obtain a plurality of line data, and obtaining a training detection sequence according to the plurality of line data;

[0023] Performing image generation processing according to the training detection data to obtain a training image, and determining a training section type corresponding to the training image;

[0024] A second training data set is generated according to the training detection sequence and the training section type; a plurality of time input points are determined based on the training detection data, and the second training data set is input into a second initial model for training according to the time input points to generate the second section recognition model.

[0025] In some embodiments, after obtaining the section recognition result, the method further includes:

[0026] Sending the section recognition result to an image processing interface for display, and detecting a confirmation instruction for the section recognition result;

[0027] In response to the detected confirmation instruction, a target section result is generated.

[0028] In some embodiments, the step of inputting the raw detection data into a trained section recognition model to obtain a section recognition result includes:

[0029] Performing data preprocessing on the original detection data to obtain post-processing data;

[0030] The post-processing data is input into the section recognition model to obtain the section recognition result.

[0031] In some embodiments, after obtaining the section recognition result, the method further includes:

[0032] Acquiring current acquisition parameters corresponding to the original detection data;

[0033] When it is detected that the section recognition result fails to match the current acquisition parameter, the current acquisition parameter is updated and a new acquisition parameter is obtained; under the new acquisition parameter, corresponding new raw detection data is obtained;

[0034] The new original detection data is input into the section recognition model to obtain a new section recognition result.

[0035] In a second aspect, an embodiment of the present application provides an image generation method, the method comprising:

[0036] Acquire original detection data, and obtain a section recognition result of the original detection data; wherein the section recognition result is obtained according to the section recognition method as described in the first aspect above;

[0037] Based on the section recognition result, image reconstruction processing is performed on the original detection data to generate a target medical image.

[0038] In a third aspect, an embodiment of the present application provides a section recognition device, the device comprising:

[0039] An acquisition module is used to acquire raw detection data;

[0040] The recognition module is used to input the original detection data into the trained section recognition model to obtain the section recognition result; wherein the section recognition model is generated by training based on the training detection data; the section recognition model is used to indicate the mapping relationship between the original detection data and the section type.

[0041] In a fourth aspect, an embodiment of the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the section recognition method as described in the first aspect above or the image generation method as described in the second aspect above is implemented.

[0042] Compared with the related art, the section recognition method, image generation method, section recognition device and electronic device provided in the embodiments of the present application obtain the original detection data; input the original detection data into the trained section recognition model to obtain the section recognition result; wherein the section recognition model is generated by training based on the training detection data; the section recognition model is used to indicate the mapping relationship between the original detection data and the section type. Based on this, the section automatic recognition step is moved forward so that it no longer depends on the image data. Section recognition can be carried out in parallel with the image reconstruction process, which improves the automation level of the system without increasing the time consumption, realizes the automatic section recognition method based on non-image data, and effectively solves the problem of low efficiency in medical image processing.

[0043] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] 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:

[0045] Figure 1 is a hardware structure block diagram of a terminal of a section recognition method according to an embodiment of the present application;

[0046] Figure 2 is a flow chart of a section recognition method according to an embodiment of the present application;

[0047] Figure 3 is a schematic diagram of a first section recognition model training process according to an embodiment of the present application;

[0048] Figure 4 is a schematic diagram of a second section recognition model training process according to an embodiment of the present application;

[0049] Figure 5 is a schematic diagram of another second section recognition model training process according to an embodiment of the present application;

[0050] Figure 6 is a flow chart of a method for displaying section recognition results according to an embodiment of the present application;

[0051] Figure 7 is a flow chart of a method for updating acquisition parameters according to an embodiment of the present application;

[0052] Figure 8 is a flow chart of an image generating method according to an embodiment of the present application;

[0053] Fig. 9is a structural block diagram of a section recognition device according to an embodiment of the present application;

[0054] Fig.10 It is a structural block diagram of another section recognition device according to an embodiment of the present application. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application. In addition, it can also be understood that although the efforts made in this development process may be complex and lengthy, for ordinary technicians in the field related to the contents disclosed in the present application, some changes such as design, manufacturing or production based on the technical contents disclosed in the present application are only conventional technical means, and should not be understood as insufficient contents disclosed in the present application.

[0056] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0057] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to greater than or equal to two. "And / or" describes the association relationship of associated objects, indicating that there can be three relationships, for example, "A and / or B" can represent: A exists alone, A and B exist at the same time, and B exists alone. The terms "first", "second", "third" and the like involved in the present application are merely used to distinguish similar objects and do not represent a specific ordering of the objects.

[0058] The method embodiment provided in this embodiment can be executed in a terminal, a computer or a similar computing device. Taking running on a terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a terminal of a section recognition method according to an embodiment of the present application. Figure 1 As shown, the terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Optionally, the terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0059] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the section recognition method in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0060] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0061] This embodiment provides a section recognition method. Figure 2 is a flow chart of a section recognition method according to an embodiment of the present application. Figure 2 As shown, the process includes the following steps:

[0062] Step S210, obtaining original detection data.

[0063] The above-mentioned original detection data refers to the detection data for the target object detected by ultrasound. For example, an ultrasonic probe is used to transmit ultrasonic pulses into the target object's body, and the echo signals reflected by these sound waves from different tissue interfaces in the body are received. These echo signals are converted into electrical signals and further processed into original detection data in digital form. These data contain information about the structure, morphology, movement and possible pathological changes in the body. Specifically, after the ultrasonic transducer in the ultrasonic probe collects the electrical signal (probe raw data), it will undergo a series of data processing such as analog-to-digital conversion (obtaining RF data), demodulation and filtering (obtaining in-phase and orthogonal phase data), beam synthesis (obtaining line data), filtering and envelope extraction (obtaining envelope data), and finally reconstruct the image. In this embodiment, in order to effectively improve the processing data, all kinds of non-image data between image reconstruction can be used as the above-mentioned original detection data and section recognition can be performed on them. In other words, the above-mentioned original detection data refers to the non-image data collected and not reconstructed in the ultrasonic imaging process. The original detection data may include probe raw data, RF data, in-phase and orthogonal phase data, line data, envelope data, etc.

[0064] Step S202, input the original detection data into the trained section recognition model to obtain the section recognition result; wherein, the section recognition model is generated by training based on the training detection data; the section recognition model is used to indicate the mapping relationship between the original detection data and the section type.

[0065] In this step, the original detection data obtained in step S210 is used as input and passed to a trained section recognition model. The model is trained by machine learning or deep learning algorithms based on a large amount of training detection data (usually a data set with marked section types). The function of the model is to predict or identify the corresponding section type based on the input original detection data. For example, when ultrasound is performed on the patient's liver, the original detection data collected during the liver ultrasound process can be used in the above manner to perform section recognition using the section recognition model. The identified section types may include the left liver fundus longitudinal section, the liver-inferior vena cava longitudinal section, the first hepatic portal level median cross-section, and other types. For another example, when ultrasound is performed on the patient's heart, the section types may include the apical four-chamber heart section, the parasternal left ventricular long axis section, the parasternal short axis section, and other types.

[0066] The above-mentioned section recognition model is a machine learning or deep learning model specifically used to identify and classify different section types in ultrasound images. The section recognition model can adopt various classification models such as Resnet networks, CNN models, and / or target detection models. It should be noted that in order to realize the forward movement of the section recognition step, in this embodiment, the section recognition model is trained by learning and analyzing a large amount of ultrasound data labeled with section types to establish a model of the mapping relationship between the original ultrasound detection data and the section type. Therefore, when new ultrasound original detection data is input, the model can automatically predict and output the most likely section type to provide a diagnostic reference for doctors. Specifically, during the training phase, the section recognition model continuously optimizes the model parameters by learning the mapping relationship between the features in the training data and the section type to improve the accuracy of the prediction. After the training is completed, the section recognition model can receive new original detection data as input and output a section recognition result, that is, the most likely section type predicted based on the input data. The above-mentioned section recognition migration can be performed after processing the original detection data acquired by the ultrasound probe to obtain multiple line data, or it can be directly recognized on the original detection data, RF data or beam-synthesized data collected by the ultrasound probe.

[0067] Among them, the above-mentioned section recognition result is the prediction or judgment given by the model after analyzing the input raw detection data. The section recognition result has important guiding significance for subsequent analysis, decision-making or operation. For example, in medical diagnosis, section recognition can help doctors quickly locate the lesion area. It can be seen that the section recognition method implemented by the above steps does not rely on image data implementation, and does not need to be set after image imaging, thus realizing a truly real-time section recognition method.

[0068] In the traditional medical image processing process, automatic section recognition is usually performed after the original detection data is reconstructed into image data, as part of image reconstruction and analysis. However, with the development of computer technology and artificial intelligence, it has become possible to move the automatic section recognition step forward and make it possible to process it in parallel with the image reconstruction process, which will greatly improve processing efficiency and accuracy. The embodiment of the present application, through the above-mentioned steps S210 to S220, inputs the original detection data into the trained section recognition model to obtain the section recognition result, thereby moving the automatic section recognition step forward, so that it no longer depends on the image data, and the section recognition can be carried out in parallel with the image reconstruction process, while improving the automation level of the system without adding extra time, realizing an automatic section recognition method based on non-image data, and effectively solving the problem of low efficiency in medical image processing.

[0069] In some embodiments, the section recognition model includes a first section recognition model; the step of inputting the original detection data into the trained section recognition model to obtain the section recognition result further includes the following steps:

[0070] Step S310: performing a first data processing on the original detection data to obtain two-dimensional matrix data.

[0071] In this step, the original detection data acquired by the ultrasonic probe is processed to obtain multiple line data, and the multiple line data are directly spliced ​​into two-dimensional matrix data. Specifically, in ultrasonic imaging technology, the ultrasonic system transmits ultrasonic pulses through the probe and receives the reflected echo signals. These echo signals are converted into electrical signals and digitized into original detection data. These data usually exist in the form of time or depth as the horizontal coordinate and echo intensity as the vertical coordinate, forming one-dimensional line data (A-scan) data. Since the subsequent image processing steps are implemented based on a two-dimensional dot matrix, in order to reduce the subsequent data processing steps, in this embodiment, these one-dimensional line data can be converted into two-dimensional matrix data that can be further analyzed, that is, the above-mentioned first data processing process is realized.

[0072] Step S320, inputting the two-dimensional matrix data into the corresponding first section recognition model to obtain a section recognition result.

[0073] It should be noted that, before the two-dimensional matrix data is input into the first section recognition model, some preprocessing may be performed on the two-dimensional matrix data. For example, the two-dimensional matrix data may be normalized, filtered, or otherwise preprocessed, and the preprocessing method may include any of the above methods, or may be a combination of at least two preprocessing methods.

[0074] The first section recognition model is a neural network model used to analyze two-dimensional matrix data. The first section recognition model can be a convolutional neural network (CNN) model or a traditional image classification model (such as a model based on feature extraction and machine learning classifiers), etc. After obtaining the two-dimensional matrix data, the data or the pre-processed data can be input into the pre-trained first section recognition model, and the first section recognition model recognizes a specific anatomical structure or section from the input two-dimensional matrix data.

[0075] Through the above steps S310 to S320, the two-dimensional matrix data obtained after the first data processing is input into the section recognition model, and the corresponding section recognition result is output, which is helpful to reduce subsequent data processing steps and effectively improve the efficiency and accuracy of image processing.

[0076] In some embodiments, training the first section recognition model may further include the following steps:

[0077] Step S301, obtaining training detection data.

[0078] The training detection data refers to the original data used for training the model, which is collected by an ultrasound probe in situations such as experiments. The training detection data can be collected for different target objects, different parts, or under different conditions to ensure that the acquired data has sufficient diversity and representativeness, covering as many section types and anatomical structures as possible. In addition, the type of the training detection data should be consistent with the type of the original detection data; for example, if the type of the original detection data is RF data, the type of the training detection data should also be RF data.

[0079] Step S302: Generate data label pairs including training line data and training slice type according to the training detection data.

[0080] In this step, in order to train the first section recognition model for analyzing the two-dimensional matrix data, it is necessary to process the above-collected training detection data so as to subsequently construct a training set. Specifically, the training detection data is processed to obtain training line data. Image generation processing is performed based on the training detection data to obtain a training image, and the training section type corresponding to the training image is determined; the section type determination process can be: annotating the generated training image to determine the section type corresponding to each image (such as a four-chamber section of the heart, a cross-section of the liver, etc.). Combine the annotated training image (or its corresponding two-dimensional matrix data) with the corresponding section type label to form a data label pair. For example, the background obtains N line data corresponding to each section: (x1, x2, …x n ), as shown in Table 1:

[0081] Table 1 Training line data

[0082] <![CDATA[x1]]> <![CDATA[x2]]> …… <![CDATA[x n ]]> 0 3 …… 0 2 5 …… 4 4 6 …… 7 12 3 …… 11 7 7 …… 124 5 123 …… 168 …… …… …… ……

[0083] x i The length of (i.e., the i-th line data) is L, and L of different data may not be unique, and its physical resolution spacing is known. The classification result corresponding to each slice is C, and the data label pair is composed of (x1, x2…x3, c). It should be understood that it is usually necessary to randomly divide the data label pairs into training sets, validation sets, and test sets in a certain ratio in order to perform performance evaluation and tuning during the model training process; for example, the data label pairs can be divided in a ratio of 7:1:2.

[0084] In addition, for the training process of all automatic section recognition models, the above steps can be used to illustrate that K examples of training data are used as positive samples, and M examples of non-standard section data can also be collected as negative samples.

[0085] Step S303, generating training matrix data according to the training line data; generating a first training data set according to the training matrix data and the data label pair.

[0086] The N data of each slice are spliced ​​into a two-dimensional matrix with a size of L*N. Combined with the corresponding physical resolution spacing, the two-dimensional matrix is ​​resampled to a uniform resolution spacing_new, and finally the corresponding two-dimensional matrix data is obtained, whose size is L new *N new The two-dimensional matrix data, that is, the training matrix data and its corresponding data label pair label c are used as the input and gold standard of the model, that is, the first training data is generated.

[0087] Step S304: input the first training data set into the first initial model for training to generate a first section recognition model.

[0088] Specifically, select an initial model suitable for processing image data, such as a convolutional neural network (CNN), input the first training data set into the initial model for training. During the training process, the model will learn how to extract features from the two-dimensional matrix data and establish a mapping relationship between the features and the section type. Please refer to the image. During the training process, please refer to Figure 3 , input each group of training matrix data such as data 1 and data 2 into the first initial model, and the first initial model outputs the predicted labels such as label 1 and label 2 corresponding to each group of data; based on the data label pair, compare the predicted label with the corresponding true label, and calculate the loss function results of the two; according to the loss function result, calculate the gradient of each parameter in the model through the back propagation algorithm to iteratively optimize the model. In addition, the validation set can be used to evaluate the performance of the model, and the parameters and structure of the model can be adjusted according to the evaluation results. When the model shows good performance on the validation set, stop training to generate a model. Use the test set to evaluate the final model to ensure its accuracy and generalization ability in practical applications. Finally, save the trained first-section recognition model for subsequent use.

[0089] Through the above embodiment, by training the model according to the first training data set generated by the training matrix data and data labels, a fully trained first section recognition model is obtained, which can effectively improve the accuracy of section recognition, thereby facilitating improving the accuracy of image processing.

[0090] In some embodiments, the section recognition model includes a second section recognition model; and the step of inputting the original detection data into the trained section recognition model to obtain the section recognition result may further include the following steps:

[0091] Step S221, performing second data processing on the original detection data to obtain an original detection sequence.

[0092] In this step, the original detection data acquired by the ultrasonic probe is processed to obtain a plurality of line data, and the plurality of line data are pre-processed respectively to obtain n processed data sequences.

[0093] It should be supplemented that, in addition to processing based on line data obtained from the original detection data, the processing in the above step S310 or the above step S221 can also be based on RF data or other types of data.

[0094] Step S222, inputting the original detection sequence into the corresponding second section recognition model to obtain a section recognition result.

[0095] The above-mentioned second section recognition model refers to a neural network model used to analyze data of this type that carries time series characteristics. Specifically, the original detection sequence after the second data processing is input into the second section recognition model. This model is specially designed for section recognition tasks and can be implemented by deep learning models such as Recurrent Neural Network (RNN), Long Short-Term Memory Network (LSTM), or machine learning models such as Support Vector Machine (SVM), Random Forest, etc. The purpose of the second section recognition model is to identify specific sections in images or data (such as organ sections in medical images, stratigraphic sections in geological exploration, etc.) based on the input data sequence. Since the model has learned how to identify sections based on the characteristics of the data during the training process, it can effectively perform section recognition on new input data.

[0096] Through the above steps S221 to S222, the original detection sequence after the second data processing is input into the corresponding second section recognition model, thereby comprehensively considering the time series characteristics of the data and effectively improving the accuracy of image processing.

[0097] In some embodiments, the training process of the second section recognition model includes the following steps:

[0098] Acquire training detection data; process the training detection data to obtain multiple line data, and obtain a training detection sequence based on the multiple line data; perform image generation processing based on the training detection data to obtain a training image, and determine the type of training section corresponding to the training image; generate a second training data set based on the training detection sequence and the training section type. The collected training detection data is processed to extract information useful for training. For example, denoising, filtering, gain adjustment, etc., to improve the quality and signal-to-noise ratio of the data. The processed data is divided into multiple line data according to the scan line or time series. Each line data represents the information of a specific scanning direction or time point. Based on the multiple line data, a training detection sequence is constructed. These sequences may contain data at multiple time points, which can reflect the dynamic changes of the internal structure of the scanned object.

[0099] Based on the training detection sequence, multiple time input points are determined, and the second training data set is input into the second initial model for training according to the time input points to generate a second section recognition model. Figure 4 . Design an LSTM-based RNN network as the second initial model. The LSTM network can handle long-term dependencies in sequence data and is suitable for analyzing time series data. The input layer of the model receives the training probe sequence, and the data at each time point is used as an input feature vector. One or more LSTM layers in the model are used to extract time-dependent features in the sequence data. The LSTM unit contains a forget gate, an input gate, and an output gate, which can control the flow of information and the update of memory. The output layer produces predictions or classification results related to the second aspect. During the training process, forward propagation: input the training probe sequence into the LSTM network, forward propagate through the network, and calculate the hidden state and final output at each time point. Use the loss function (such as cross entropy loss, mean square error, etc.) to calculate the difference between the network output and the true label. The loss value is propagated back to each layer in the network through the backpropagation algorithm, and the gradient is calculated to achieve iterative optimization of the model until the model converges or reaches the training time to generate a second aspect recognition model. Alternatively, refer to Figure 5 , the above-mentioned second initial model can also be a RNN-like network that can support multiple layers. During the training process, the training data is input into the first layer of the model (model layer1) in the order of time steps, and then passed to the second layer (model layer2) layer by layer. At each time step, the RNN unit calculates the new hidden state and output based on the current input and the previous hidden state. The loss (such as cross entropy loss, mean square error, etc.) between the model's prediction result and the true label is calculated in the second layer (or the specified output layer) to iteratively optimize the model.

[0100] Through the above embodiment, by training the model according to the second training data set generated by the training detection sequence and the data label, a fully trained second section recognition model is obtained, which can effectively improve the accuracy of section recognition, thereby facilitating further improving the accuracy of image processing.

[0101] In some embodiments, after obtaining the above section recognition results, please refer to Figure 6 , the above section recognition method also includes the following steps:

[0102] Step S610, sending the section recognition result to the image processing interface for display, and detecting a confirmation instruction for the section recognition result.

[0103] Step S620, in response to the detected confirmation instruction, generating a target section result. The image processing interface may be deployed on a display device of the ultrasound system.

[0104] Specifically, after the recognition is completed, the algorithm will automatically output the section recognition result, that is, the above-mentioned section type. The section recognition result is then sent to a display device to display the original image and the identified section area in a graphical manner on the image processing interface; for example, the section is marked with a highlight box, a color mark or other visual elements on the image. The operator can confirm or modify the section recognition result by interacting with the image processing interface, such as operating buttons, check boxes and other sliders. The system can continuously monitor user interactions and detect confirmation instructions issued by the user. Once the confirmation instruction is detected, the system will make necessary adjustments or confirmations to the initial section recognition result based on the user's confirmation or modification. Finally, the system generates a target section result based on the user's confirmation and possible modifications. This target section result can be which standard section it is currently in, or, if the embodiment allows, it can also be further processed based on the detection result of the standard section to obtain a verified section position, a probe position that matches the section recognition result, and an accurate representation of parameters such as matching corresponding image optimization parameters, or it can be image data for subsequent analysis or processing. It is understandable that the target section result can be stored in a database, or can also be exported as a file for subsequent analysis, reporting or sharing with other systems. Through the above embodiment, through user confirmation and feedback, it is helpful to improve the accuracy and reliability of the automated processing results.

[0105] In some embodiments, after obtaining the above section recognition results, please refer to Figure 7 , the above section recognition method also includes the following steps:

[0106] Step S710, obtaining current acquisition parameters corresponding to the original detection data.

[0107] Step S720, when it is detected that the section recognition result fails to match the current acquisition parameters, the current acquisition parameters are updated and new acquisition parameters are obtained; under the new acquisition parameters, corresponding new original detection data are obtained.

[0108] Step S730, input the new original detection data into the section recognition model to obtain a new section recognition result.

[0109] Specifically, in this embodiment, in order to ensure the accuracy of the section, first, it is possible to further detect whether the current acquisition parameters corresponding to the current original detection data meet the section requirements. This detection method can be implemented based on a parameter dictionary or the like. For example, considering that the parameters of all sections constitute a parameter dictionary, the parameters between different sections may be the same or different, and for a certain section, the acquisition parameters used and the corresponding parameters of the identified section should match. Therefore, if it is detected that the currently used acquisition parameters are A, and the parameters corresponding to the identified current section are B, in other words, it is detected that the current section recognition result fails to match the current acquisition parameters, then it means that the current acquisition parameters do not meet the section requirements; if the parameters corresponding to the identified current section are also A, then it is determined that the section requirements are met.

[0110] Then, based on the above detection results, it is determined whether to continue the subsequent process. Among them, if it is detected that the section identified under the current acquisition parameters meets the section requirements, the subsequent image reconstruction process can be continued based on the current section recognition results. On the contrary, it means that there are still problems such as low section clarity or incorrect position. In this case, the subsequent workflow path can be adjusted to adjust the section results. Alternatively, in order to simplify the calculation process, it is also possible to return to the previous data acquisition step and update the acquisition parameters; in this way, there is no need to adjust the subsequent workflow. For example, dynamically adjust various hardware parameters of the ultrasound equipment, including ultrasound frequency, scanning angle, and gain. In this process, the original detection data acquired synchronously will also change at any time. By inputting these real-time updated original detection data into the section recognition model, new section recognition results can be obtained.

[0111] It should be understood that the above-mentioned new section recognition results can continue to be tested until it is determined that the current acquisition parameters have met the section requirements. At this time, there is no need to repeatedly update the acquisition parameters, and the subsequent image reconstruction process can be performed normally.

[0112] Through the above-mentioned embodiments, by continuously adjusting data acquisition parameters and re-acquiring data until section data that meets the requirements is obtained, the quality and accuracy of the data can be significantly improved, thereby helping to support more accurate analysis and diagnosis.

[0113] In some embodiments, the above-mentioned inputting the original detection data into the trained section recognition model to obtain the section recognition result may also include the following steps: performing data preprocessing on the original detection data to obtain post-processing data; inputting the post-processing data into the section recognition model to obtain the section recognition result. Among them, the above-mentioned data preprocessing method can be at least one of normalization, filtering processing and the like. Through the above-mentioned data preprocessing method, the data to be input into the model is processed, which can improve the efficiency and accuracy of subsequent processing, thereby helping to improve the efficiency and accuracy of image processing.

[0114] This embodiment also provides an image processing method, Figure 8 is a flow chart of an image generation method according to an embodiment of the present application, such as Figure 8 As shown, the process includes the following steps:

[0115] Step S810, obtaining original detection data, and obtaining a section recognition result of the original detection data; wherein the section recognition result is obtained according to the section recognition method described in any of the above embodiments.

[0116] Step S820, based on the section recognition result, the original detection data is reconstructed to generate the target medical image. Key parameters such as depth, gain, frequency, scaling, etc. are matched based on the section recognition result, and these parameters will be used to guide the image reconstruction process. Then, the original detection data is processed using the extracted section parameters and the selected reconstruction algorithm to generate the target medical image.

[0117] Through the above steps S810 to S820, an image generation method that processes the automatic section recognition and image reconstruction process in parallel is realized, which improves the automation level of the ultrasound system without increasing the time consumption, thereby effectively improving the image processing efficiency.

[0118] It should be noted that the steps shown in the above process or the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0119] This embodiment also provides a section recognition device, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated here. As used below, the terms "module", "unit", "subunit", etc. can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0120] Fig. 9is a structural block diagram of a section recognition device according to an embodiment of the present application, such as Fig. 9 As shown, the device includes: an acquisition module 91, used to acquire original detection data; an identification module 92, used to input the original detection data into a trained section recognition model to obtain a section recognition result; wherein the section recognition model is generated by training based on the training detection data; the section recognition model is used to indicate the mapping relationship between the original detection data and the section type.

[0121] In some embodiments, the above-mentioned recognition module 92 is also used to perform a first data processing on the original detection data to obtain two-dimensional matrix data; the recognition module 92 inputs the two-dimensional matrix data into the corresponding first section recognition model to obtain a section recognition result.

[0122] In some embodiments, the above-mentioned section recognition device also includes a training module; the training module is used to obtain training detection data; the training module processes the training detection data to obtain training matrix data; the training module performs image generation processing based on the training detection data to obtain a training image, and determines the training section type corresponding to the training image; the training module generates a first training data set based on the training matrix data and the training section type; the first training data set is input into the first initial model for training to generate a first section recognition model.

[0123] In some embodiments, the identification module 92 is further used to perform a second data processing on the original detection data to obtain an original detection sequence; the identification module 92 inputs the original detection sequence into a corresponding second section recognition model to obtain a section recognition result.

[0124] In some embodiments, the training module is also used to obtain training detection data; the training module processes the training detection data to obtain multiple line data, and obtains a training detection sequence based on the multiple line data; the training module performs image generation processing based on the training detection data to obtain a training image, and determines the training section type corresponding to the training image; the training module generates a second training data set based on the training detection sequence and the training section type; determines multiple time input points based on the training detection sequence, inputs the second training data set into the second initial model for training according to the time input points, and generates a second section recognition model.

[0125] In some embodiments, the above-mentioned section recognition device also includes an interaction module; the interaction module is used to send the section recognition result to the image processing interface for display, and detect a confirmation instruction for the section recognition result; the interaction module generates a target section result in response to the detected confirmation instruction.

[0126] In some of these examples, see Fig.10The above-mentioned section recognition device also includes an updating module 93; the updating module 93 is used to obtain the current acquisition parameters corresponding to the original detection data; when the updating module 93 detects that the section recognition result fails to match the current acquisition parameters, it updates the current acquisition parameters and obtains new acquisition parameters; under the new acquisition parameters, the corresponding new original detection data is obtained; the updating module 93 inputs the new original detection data into the section recognition model to obtain a new section recognition result.

[0127] In some embodiments, the identification module 92 is further used to perform data preprocessing on the original detection data to obtain post-processing data; the identification module 92 inputs the post-processing data into the section recognition model to obtain the section recognition result.

[0128] This embodiment also provides an image generating device, which includes: a detection module; used to obtain original detection data and obtain a section recognition result of the original detection data; wherein the section recognition result is obtained according to the section recognition method described in any of the above embodiments; a reconstruction module, used to perform image reconstruction processing on the original detection data based on the section recognition result to generate a target medical image.

[0129] It should be noted that the above modules can be functional modules or program modules, and can be implemented by software or hardware. For modules implemented by hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0130] This embodiment further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0131] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0132] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0133] S1, obtain the original detection data.

[0134] S2, inputting the original detection data into the trained section recognition model to obtain the section recognition result; wherein, the section recognition model is generated by training based on the training detection data; the section recognition model is used to indicate the mapping relationship between the original detection data and the section type.

[0135] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0136] In addition, in combination with the section recognition method in the above embodiment, the embodiment of the present application can provide a storage medium for implementation. The storage medium stores a computer program; when the computer program is executed by a processor, any section recognition method in the above embodiment is implemented.

[0137] Those skilled in the art should understand that the technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0138] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A section recognition method, characterized in that: The method comprises: Get raw detection data; The original detection data is input into the trained section recognition model to obtain the section recognition result; the section recognition model is used to indicate the mapping relationship between the original detection data and the section type.

2. The section recognition method according to claim 1, characterized in that: The section recognition model includes a first section recognition model; the inputting the original detection data into the trained section recognition model to obtain the section recognition result includes: Performing a first data processing on the original detection data to obtain two-dimensional matrix data; The two-dimensional matrix data is input into a first section recognition model to obtain the section recognition result.

3. The section recognition method according to claim 2, characterized in that: Training the first section recognition model includes: Get training probe data; Generating a data label pair including training line data and training section type according to the training detection data; Generate training matrix data according to the training line data; generate a first training data set according to the training matrix data and the data label pair; The first training data set is input into the first initial model for training to generate the first section recognition model.

4. The section recognition method according to claim 1, characterized in that: The section recognition model includes a second section recognition model; the inputting the original detection data into the trained section recognition model to obtain the section recognition result includes: Performing a second data processing on the original detection data to obtain an original detection sequence; The original detection sequence is input into a second section recognition model to obtain the section recognition result.

5. The section recognition method according to claim 4, characterized in that: Training the second section recognition model includes: Get training probe data; Processing the training detection data to obtain a plurality of line data, and obtaining a training detection sequence according to the plurality of line data; Performing image generation processing according to the training detection data to obtain a training image, and determining a training section type corresponding to the training image; A second training data set is generated according to the training detection sequence and the training section type; a plurality of time input points are determined based on the training detection sequence, and the second training data set is input into a second initial model for training according to the time input points to generate the second section recognition model.

6. The section recognition method according to claim 1, characterized in that: After obtaining the section recognition result, the method further includes: Sending the section recognition result to an image processing interface for display, and detecting a confirmation instruction for the section recognition result; In response to the detected confirmation instruction, a target section result is generated.

7. The section recognition method according to claim 1, characterized in that: After obtaining the section recognition result, the method further includes: Acquiring current acquisition parameters corresponding to the original detection data; When it is detected that the section recognition result fails to match the current acquisition parameter, the current acquisition parameter is updated and a new acquisition parameter is obtained; under the new acquisition parameter, corresponding new raw detection data is obtained; The new original detection data is input into the section recognition model to obtain a new section recognition result.

8. The section recognition method according to any one of claims 1 to 7, characterized in that: The step of inputting the original detection data into the trained section recognition model to obtain the section recognition result includes: Performing data preprocessing on the original detection data to obtain post-processing data; The post-processing data is input into the section recognition model to obtain the section recognition result.

9. An image generation method, characterized in that: The method comprises: Acquire original detection data, and obtain a section recognition result of the original detection data; wherein the section recognition result is obtained according to the section recognition method according to any one of claims 1 to 8; Based on the section recognition result, image reconstruction processing is performed on the original detection data to generate a target medical image.

10. A section recognition device, characterized in that: The device comprises: An acquisition module is used to acquire raw detection data; The recognition module is used to input the original detection data into the trained section recognition model to obtain the section recognition result; the section recognition model is used to indicate the mapping relationship between the original detection data and the section type.

11. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to execute the section recognition method according to any one of claims 1 to 8, or the image generation method according to claim 9.

Images