Method and system for automatic guidance of echocardiographic standard views based on motion capture

Abstract

The application belongs to the field of image processing, and provides a method and system for automatically guiding a standard section of cardiac ultrasound based on motion capture, which captures images of a subject, determines a human posture through acquired image information, and performs motion capture on an ultrasound probe to determine a motion trajectory and posture of the ultrasound probe; real-time ultrasound image information of the subject in a target body position and the ultrasound probe is acquired; trajectory conversion is performed according to the motion trajectory of the ultrasound probe, knowledge learning and feature extraction are performed according to the real-time ultrasound image information, a gap between a current detection section and a standard section is determined based on an extraction result and a conversion result, and a moving angle and a moving distance of the ultrasound probe are predicted; and automatic acquisition of the standard section of cardiac ultrasound and action guidance of the probe can be realized according to a prediction result.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of ultrasound image processing, and relates to a method and system for automatically guiding a standard cardiac ultrasound section based on motion capture. BACKGROUND

[0002] The statements in this section merely provide background information related to the application and do not necessarily constitute prior art.

[0003] Echocardiography has become the most commonly used imaging method for heart examination due to its non-invasiveness, non-radioactivity, portability and real-time dynamic nature. However, the acquisition and analysis of cardiac ultrasound sections are closely related to the subjective operation of doctors, and require a patient in a resting state and appropriate ultrasound detection equipment to complete, thus inevitably causing detection errors. This brings great challenges to the diagnosis of heart disease and heart function, and also puts higher technical requirements on ultrasound doctors.

[0004] The anatomical structure of the heart is highly complex, and although there are detailed echocardiographic quality assessment guidelines, it is still a difficult task for inexperienced ultrasound doctors to acquire all standard cardiac ultrasound sections. In addition, the serious shortage of primary high-level ultrasound doctors, the uneven distribution of medical levels in urban and rural areas, and the large individual differences of patients, etc. Objective problems have caused serious obstacles to the acquisition of standard echocardiograms, which greatly increases the risk of misdiagnosis and delayed treatment of patients. Therefore, the acquisition of standard cardiac ultrasound sections is of great significance for the diagnosis and treatment of heart disease.

[0005] In addition, the cultivation and experience accumulation of high-level ultrasound doctors is a long-term and costly process, and simply relying on professional doctors to collect sections will consume a lot of manpower, material resources and time cost. SUMMARY

[0006] To solve the above problems, the application provides a method and system for automatically guiding a standard cardiac ultrasound section based on motion capture, which uses a motion capture system to obtain the real-time position of the ultrasound doctor's hand probe, combines transfer learning and deep metric learning, realizes knowledge reuse across domains, tasks and distributions, and realizes automatic acquisition of standard cardiac ultrasound sections and probe action guidance.

[0007] According to some embodiments, the application adopts the following technical scheme:

[0008] A method for automatically guiding a standard cardiac ultrasound section based on motion capture, comprising the following steps:

[0009] Image capture of the subject is performed, the human body posture is determined through the acquired image information, and the motion capture of the ultrasound probe is performed to determine the motion trajectory and posture of the ultrasound probe.

[0010] acquire real-time ultrasound image information of the subject in the target body position;

[0011] According to the motion trajectory of the ultrasound probe, trajectory conversion is performed, knowledge learning and feature extraction are performed according to the real-time ultrasound image information, and based on the extraction result and the conversion result, the difference between the current detection section and the standard section is determined, and the moving angle and moving distance of the ultrasound probe are predicted;

[0012] According to the prediction result, the path of the ultrasound probe is guided.

[0013] As an optional implementation, the image of the subject is acquired to obtain the current body position of the subject;

[0014] The marker on the ultrasound probe is captured, and the real-time position of the ultrasound probe is obtained through three-dimensional rigid body transformation criterion according to the capture result;

[0015] According to the conversion of the world coordinate system, the ultrasound probe coordinate system and the subject position coordinate system, the real-time position and angle information of the ultrasound probe is obtained.

[0016] As an optional implementation, if the current body position of the subject is not suitable for heart section acquisition, the subject is guided to change the posture to the target body position.

[0017] As an optional implementation, the standard section feature extraction and confirmation process includes:

[0018] Extracting common deep features from the obtained heart ultrasound image and the heart ultrasound quality evaluation public data set;

[0019] The common deep features are input into the encoder and the decoder for training, the encoded data is obtained after category prediction to obtain the corresponding standard reference image, and the decoder is trained using the standard structure feature, and the quality prediction result is output;

[0020] The prior knowledge of the standard view is trained on the encoded features through the parameter sharing Transformer decoder to obtain the final ultrasound view displacement amount.

[0021] As an optional implementation, the specific process of knowledge learning and feature extraction according to the real-time ultrasound image information includes dividing the obtained image data into image block sequences, and performing position division embedding, which retains spatial information while obtaining spatial correlation of different tissue structures according to pixel distribution distance;

[0022] Then the arranged image data is input into the feature extractor composed of Transformer blocks based on the multi-head attention mechanism to extract deep structure features;

[0023] According to the extracted deep structure features, the displacement amount of the ultrasonic method is predicted.

[0024] As further, the specific process of predicting the displacement amount of the ultrasonic method according to the extracted deep structure features includes measuring the real-time cardiac ultrasound examination section and the corresponding cardiac ultrasound standard section, and calculating the feature difference between the two;

[0025] Map multiple inputs to a single output, and train the network through backpropagation;

[0026] The weights are normalized, and the final six-degree-of-freedom displacement prediction is obtained by summing the weights and the corresponding key values and inverse weighted sum, and the distance information between the real-time section and the standard section is calculated.

[0027] An automatic cardiac ultrasound standard section guiding system based on motion capture, comprising:

[0028] The motion capture unit is used to acquire the image of the subject, determine the human body posture through the image information, and acquire the image of the ultrasonic probe to determine the motion trajectory and posture of the ultrasonic probe;

[0029] The data acquisition module is used to acquire the real-time ultrasonic image information of the ultrasonic probe of the subject in the target body position;

[0030] The depth migration measurement module is used to perform trajectory conversion according to the motion trajectory of the ultrasonic probe, perform knowledge learning and feature extraction according to the real-time ultrasonic image information, determine the gap between the current detection section and the standard section based on the extraction result and the conversion result, and predict the moving angle and moving distance of the ultrasonic probe;

[0031] The result feedback module is used to guide the path of the ultrasonic probe according to the prediction result.

[0032] As an optional implementation, the motion capture unit includes a motion capture device, a camera device, and a processing module, wherein the motion capture device records the trajectory of the ultrasonic probe extension rod in real time, the camera device is arranged in front of the data acquisition module to acquire the image information of the subject, and the processing system is used to determine the motion trajectory, real-time position, and human body position of the ultrasonic probe according to the information acquired by the motion capture device and the camera device.

[0033] As further, the ultrasonic probe is connected with an extension rod, a reflective strip is arranged on the extension rod, and the processing system is used to capture the reflective signal of the reflective strip and acquire the real-time position of the ultrasonic probe through three-dimensional rigid body transformation criteria according to the capture result;

[0034] According to the conversion of the world coordinate system, the ultrasonic probe coordinate system and the subject position coordinate system, the real-time position and angle information of the ultrasonic probe is obtained.

[0035] As an alternative embodiment, a voice module is further included, which is used to broadcast a voice to guide the subject to change the posture and turn to the target position when the current body position of the subject is not suitable for collecting the heart section.

[0036] As an alternative embodiment, the depth migration measurement module comprises:

[0037] The segmentation module is used to segment the obtained image data into an image block sequence and perform position division embedding, so as to obtain the spatial correlation of different tissue structures according to the pixel distribution distance while completely retaining the spatial information.

[0038] The feature extraction module is used to input the arranged image data into a feature extractor composed of a Transformer block based on a multi-head attention mechanism, and extract deep structure features.

[0039] The prediction module is used to predict the displacement amount of the ultrasonic method according to the extracted deep structure features.

[0040] As further, the prediction module comprises:

[0041] The difference calculation module is used to measure the real-time heart ultrasonic examination section and the corresponding heart ultrasonic standard section, and calculate the feature difference between the two.

[0042] The training module is used to map multiple inputs to a single output, and train the network through back propagation.

[0043] The distance calculation module is used to normalize the weight, obtain the final six-degree-of-freedom displacement prediction by summing the weight, the corresponding key value and the inverse weighted sum, and calculate the distance information between the real-time section and the standard section.

[0044] Compared with the prior art, the present application has the following advantages:

[0045] The external camera combined with the motion capture device of the present application does not affect the work of the subject and the ultrasonic probe, does not require a large amount of equipment structure modification of the ultrasonic equipment, has low investment and simple operation.

[0046] The present application can ensure that the subject quickly turns into the best position, and according to the difference between the detection section obtained by the current ultrasonic probe and the standard section, determines the trajectory movement and angle movement of the ultrasonic probe, guides the movement of the ultrasonic probe, and collects the standard section or improves the accuracy and precision of the detection section as much as possible.

[0047] In order to make the above objectives, characteristics and advantages of the present application more apparent, more comprehensible, the following preferred embodiments are specifically described in detail below, together with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0048] The drawings constituting a part of this specification provide further understanding of the present application, the illustrative embodiments of the present application and its description serve the purpose of explanations of the present application, and do not constitute an improper limitation of the present application.

[0049] Figure 1 is a system structure block diagram of the present embodiment;

[0050] Figure 2 is a system layout diagram of the present embodiment;

[0051] Figure 3 is a six-degree-of-freedom schematic diagram of ultrasonic section acquisition of the present embodiment;

[0052] Figure 4 is a structure diagram of the feature extractor based on Transformer of the present embodiment. DETAILED DESCRIPTION

[0053] The present application is further described below in conjunction with the drawings and embodiments.

[0054] It should be noted that the following detailed description is exemplary, and is intended to provide further explanation of the present application. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0055] It should be noted that the terms used herein are only for the purpose of describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and furthermore, it should be understood that when the terms "comprise" and / or "include" are used in the specification, they indicate the presence of a feature, step, operation, device, component and / or combination thereof.

[0056] Embodiment One

[0057] The present embodiment utilizes an external camera combined with a motion capture device to track and judge the posture of the subject, judges the human posture through image information, and guides to the best position through voice prompt, at the best position, the heart section of the subject is collected by the ultrasonic probe, in this process, the trajectory of the ultrasonic probe extension rod is recorded in real time by the motion capture system, and the ultrasonic probe trajectory is converted, at the same time, the real-time video information collected by the ultrasonic probe is input into the deep migration metric network for knowledge learning and feature extraction;

[0058] The deep migration metric network is used to predict the current detection section, the standard section acquisition angle, and the probe movement distance by combining the probe motion trajectory. The recognition results are displayed and broadcast in real time until the standard section of cardiac ultrasound is acquired and the examination results are output.

[0059] The following is a detailed description of the scheme with reference to the attached diagram.

[0060] like Figure 1 As shown, the motion capture-based automatic guidance system for standard cardiac ultrasound sections includes a motion capture module 1, an echocardiogram data acquisition module 2, a depth migration measurement module 3, an ultrasound recognition module 4, and a result feedback module 5.

[0061] Among them, the motion capture module 1 is used to monitor the subject's body position and the movement trajectory of the ultrasound probe.

[0062] Data acquisition module 2 is used to acquire real-time ultrasound image information of the subject.

[0063] The depth migration measurement module 3 is used to extract deep features from the acquired cross-section and perform cross-domain migration of key features. By combining it with the motion trajectory features of the ultrasound probe, the corresponding cross-section quality assessment results and location information are obtained.

[0064] The ultrasound recognition module 4 is used to digitally display and voice broadcast the recognition results and location information.

[0065] The results feedback module 5 is used to guide the ultrasound probe, assisting ultrasound physicians in accurately selecting standard ultrasound sections and guiding the probe.

[0066] like Figure 2 As shown, in this embodiment, the motion capture module 1 uses a motion capture device to record and extract features of the trajectory of the ultrasound probe extension rod in real time, and uses an external camera to capture the patient's position, which is combined with the motion capture device to keep the patient in the optimal examination position.

[0067] In this embodiment, the motion capture device consists of seven visual acquisition cameras, evenly distributed around the perimeter of the roof of the acquisition site. The ultrasonic equipment is located in the center of the room, with no obstructions above, below, or to the sides. In addition, video recording equipment is set up around the perimeter of the site to enable real-time tracking of the acquisition process.

[0068] In the embodiment, the ultrasonic probe extension rod is calibrated by the reflective strip, the processing system of the motion capture device captures the reflective signal, and the real-time position of the ultrasonic probe is obtained through three-dimensional rigid body transformation criteria. The patient position signal is received by the external camera, and when the patient is ready, the whole body image of the patient and the motion trajectory of the ultrasonic probe are collected in real time by each visual acquisition camera in the detection area. According to the conversion of the world coordinate system, the ultrasonic probe coordinate system and the patient position coordinate system, the information of the real-time position and angle (six degrees of freedom) of the ultrasonic probe is obtained.

[0069] The data acquisition module 2 uses an echocardiogram acquisition device (the execution structure is an ultrasonic probe) to collect echocardiogram images in real time. In this process, the six degrees of freedom of the ultrasonic probe are monitored in real time by the motion capture system, as shown in Figure 3 .

[0070] The depth migration measurement module 3 is constructed on the basis of an attention-assisted guided ultrasonic image standard view navigation algorithm and mainly consists of the following two parts: an attention-assisted guided ultrasonic image quality inspection and skill prediction algorithm and an ultrasonic image standard view intelligent navigation.

[0071] Specifically, first, the common deep features of the obtained cardiac ultrasound image and the cardiac ultrasound quality evaluation public data set are extracted by using a feature extractor; second, the deep features are input into an encoder and a decoder for training, the encoded data are subjected to category prediction to obtain the corresponding standard reference image, the decoder is trained by using a standard structure feature (the standard structure feature is extracted by adding standard section prior knowledge in the training process, and the related prior knowledge is applied to real-time standard section selection to obtain the quality prediction result); and finally, the prior knowledge of the standard reference image is used to train the encoded features by a parameter-shared Transformer decoder to obtain the final ultrasonic view displacement amount and apply it to the cardiac image ultrasonic acquisition process.

[0072] The above-mentioned encoder, decoder and feature extractor are composed of two layers of Transformer blocks. The Transformer architecture is composed of a multi-head attention mechanism and a fully connected feedforward network. Each sublayer is connected by a residual link and subjected to layer normalization calculation. The attention mechanism (Attention) allows the dependency of the input and output sequences to be modeled and efficiently parallelized without considering their distance in the sequence. In this way, the network structure composed of the Transformer can obtain the deep features of the ultrasonic image. For example, the Transformer-based feature extractor architecture is as shown in Figure 4 .

[0073] After the ultrasound image input, the input image data is first segmented into a sequence of image blocks, and position division embedding is performed to obtain the spatial correlation of different tissue structures while fully preserving the spatial information according to the pixel distribution distance. Then, the arranged image data is input into a feature extractor composed of Transformer blocks based on a multi-head attention mechanism, the deep structure features are extracted and input into a structure predictor to predict the displacement of the ultrasound method. The attention mechanism can be represented as:

[0074] (1)

[0075] wherein, Q is a query matrix, K is an attention focus content, V is a value matrix.

[0076] Then, the displacement predictor is composed of multiple interactive layers of multi-head attention and multilayer perceptron (MLP) blocks. The multi-head attention block and the multilayer perceptron are part of the original architecture of the transformer network. The attention mechanism can be intuitively explained by using the human visual mechanism, which allows the model to dynamically focus on the structural features that have a positive effect on prediction in the identification task. On this basis, the TIQA, IQA, CNN, and other quality evaluation methods measure the real-time cardiac ultrasound section and the corresponding standard cardiac ultrasound section, calculate the feature difference between the two, and the feature with the smallest second-order statistical difference is the optimal prediction feature. The set is conducive to the quality predictor to better obtain the tissue structure quality prediction result.

[0077] In this embodiment, the quality evaluation result divides the standard degree of the section into 5 levels: excellent, good, qualified, basically qualified, and unqualified.

[0078] The MLP maps multiple inputs to a single output, and the multiple inputs in this embodiment refer to multi-dimensional features. This part converts the multi-dimensional features into a single quality score output, and trains the above overall architecture through backpropagation. The structure finally uses the Softmax function to normalize the weights, and obtains the final six-degree-of-freedom displacement prediction by summing the weights, corresponding key values, and inverse weighting, so as to calculate the distance information between the real-time section and the standard section. The moving direction, moving displacement, and posture of the probe are determined.

[0079] The ultrasound recognition module 4 can output real-time section category information and section quality evaluation results, and comprehensively obtain ultrasound prompt information and perform digital display and voice broadcast. The category refers to the output section name, including the two-chamber heart, three-chamber heart, four-chamber heart, five-chamber heart, and sternal left ventricular long axis, etc. Common ultrasound sections.

[0080] The result feedback module 5 can output the six-degree-of-freedom displacement information of the current cross-section converted standard cross-section, realize real-time feedback interaction of the ultrasonic equipment and the motion capture device, and further improve the accuracy of model training.

[0081] Those skilled in the art will appreciate that embodiments of the application can be supplied as methods, systems, or computer program products. Accordingly, the application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the application can be embodied in the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk memory, CD-ROMs, optical storage media, etc.) having computer usable program code embodied therein.

[0082] The present application is described in reference to the flowcharts and / or block diagrams of the methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, a special purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions, which are executed via the processor of the computer or other programmable data processing apparatus, generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the functions specified in the flowchart

[0083] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer readable memory produce a manufacture product including instruction means, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the functions specified in the flowchart

[0084] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are performed on the computer or other programmable data processing apparatus to produce a computer implemented process, so that the instructions executed on the computer or other programmable data processing apparatus provide a process for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the functions specified in the flowchart

[0085] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0086] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An automatic guidance system for standard sections of cardiac ultrasound based on motion capture, characterized in that, include: The motion capture unit is used to acquire images of the subject, determine the human posture through image information, and acquire images of the ultrasound probe to determine the motion trajectory and posture of the ultrasound probe. The data acquisition module is used to acquire real-time ultrasound image information of the subject in the target position using the ultrasound probe; The depth migration measurement module is used to perform trajectory conversion based on the motion trajectory of the ultrasound probe, perform knowledge learning and feature extraction based on real-time ultrasound image information, determine the difference between the current detection section and the standard section based on the extraction and conversion results, and predict the movement angle and movement distance of the ultrasound probe. The results feedback module is used to guide the ultrasound probe along its path based on the prediction results. The deep migration metric module includes: The segmentation module is used to segment the acquired image data into a sequence of image blocks and perform positional partitioning and embedding. While fully preserving spatial information, it obtains the spatial correlation of different tissue structures based on pixel distribution distance. The feature extraction module is used to input the arranged image data into a feature extractor composed of Transformer blocks based on a multi-head attention mechanism to extract deep structural features; The prediction module is used to predict the displacement of the ultrasonic manipulation based on the extracted deep structural features. The prediction module includes: The difference calculation module is used to measure the real-time cardiac ultrasound examination section and the corresponding standard cardiac ultrasound section, and calculate the characteristic differences between the two. The training module is used to map multiple inputs to a single output and train the network through backpropagation; The distance calculation module is used to normalize the weights, sum the weights and corresponding key values ​​and the inverse weighted sum to obtain the final six-degree-of-freedom displacement prediction, and calculate the distance information between the real-time tangent and the standard tangent.

2. The automatic guidance system for standard cardiac ultrasound sections based on motion capture as described in claim 1, characterized in that, The motion capture unit includes a motion capture device, a camera device, and a processing module. The motion capture device records the trajectory of the ultrasound probe extension rod in real time. The camera device is set in front of the data acquisition module to acquire image information of the subject. The processing module is used to determine the motion trajectory, real-time position, and body position of the ultrasound probe based on the information acquired by the motion capture device and the camera device.

3. The automatic guidance system for standard cardiac ultrasound sections based on motion capture as described in claim 2, characterized in that, The ultrasonic probe is connected to an extension rod, on which a reflective strip is provided. The processing module is used to capture the reflective signal of the reflective strip and, based on the capture result, obtain the real-time position of the ultrasonic probe through a three-dimensional rigid body transformation criterion. Based on the conversion between the world coordinate system, the ultrasound probe coordinate system, and the subject's position coordinate system, the real-time position and angle information of the ultrasound probe are obtained.

4. The automatic guidance system for standard cardiac ultrasound sections based on motion capture as described in claim 1, characterized in that, It also includes a voice module, which is used to broadcast voice instructions to guide the subject to change posture and switch to the target posture when the subject's current position is not suitable for cardiac sectional imaging.

5. A method for automatic guidance of standard cardiac ultrasound sections based on motion capture, applied to the automatic guidance system for standard cardiac ultrasound sections as described in any one of claims 1-4, characterized in that, Includes the following steps: The subject's image is captured, the human posture is determined by the acquired image information, and the motion of the ultrasound probe is captured to determine the motion trajectory and posture of the ultrasound probe. Acquire real-time ultrasound images of the subject in the target position using an ultrasound probe; Based on the motion trajectory of the ultrasound probe, trajectory conversion is performed. Based on real-time ultrasound image information, knowledge learning and feature extraction are performed. Based on the extraction and conversion results, the difference between the current detection section and the standard section is determined, and the movement angle and movement distance of the ultrasound probe are predicted. Based on the prediction results, the ultrasound probe is guided along its path.

6. The method for automatic guidance of standard sections of cardiac ultrasound based on motion capture as described in claim 5, characterized in that, Images of the subject are captured to obtain the subject's current body position; The markers on the ultrasonic probe are captured, and the real-time position of the ultrasonic probe is obtained based on the capture results using the three-dimensional rigid body transformation criterion. Based on the conversion between the world coordinate system, the ultrasound probe coordinate system, and the subject's position coordinate system, the real-time position and angle information of the ultrasound probe are obtained.

7. The method for automatic guidance of standard cardiac ultrasound sections based on motion capture as described in claim 5, characterized in that, The standard section feature extraction and verification process includes: Extract the common deep features between the obtained cardiac ultrasound images and the publicly available cardiac ultrasound quality assessment dataset; The shared deep features are input into the encoder and decoder for training. The encoded data is then used for category prediction to obtain the corresponding standard reference image. The standard structural features are then used to train the decoder and output the quality prediction result. The prior knowledge of the standard view is used to train the encoded features through a parameter-shared Transformer decoder to obtain the final ultrasonic view displacement.

8. The method for automatic guidance of standard sections of cardiac ultrasound based on motion capture as described in claim 5, characterized in that, The specific process of knowledge learning and feature extraction based on real-time ultrasound image information includes segmenting the acquired image data into image block sequences and embedding them at their positions. While fully preserving spatial information, the spatial correlation of different tissue structures is obtained based on the pixel distribution distance. The arranged image data is then input into a feature extractor composed of Transformer blocks based on a multi-head attention mechanism to extract deep structural features. Based on the extracted deep structural features, the displacement of the ultrasonic manipulation is predicted.

9. The method for automatic guidance of standard sections of cardiac ultrasound based on motion capture as described in claim 8, characterized in that, The specific process of predicting the displacement of ultrasound manipulation based on the extracted deep structural features includes measuring the real-time cardiac ultrasound examination section with the corresponding standard cardiac ultrasound section and calculating the feature differences between the two. Multiple inputs are mapped to a single output, and the network is trained using backpropagation; The weights are normalized, and the weights, corresponding key values, and inverse weighted sums are used to obtain the final six-degree-of-freedom displacement prediction, and the distance information between the real-time tangent and the standard tangent is calculated.

Images