Ultrasonic inspection method, system, equipment, storage medium and program product

Abstract

The invention relates to the technical field of artificial intelligence, and provides an ultrasonic inspection method, system and device, a storage medium and a program product. On the basis of an ultrasonic inspection model, intention recognition is conducted on the instruction text, an interface function corresponding to the instruction text is obtained, and the ultrasonic inspection model is obtained through training on the basis of a large language model by applying the sample instruction text and the sample interface function corresponding to the sample instruction text; performing parameter analysis on the interface function to obtain execution parameters, and controlling the ultrasonic robot to execute ultrasonic inspection based on the execution parameters. According to the ultrasonic inspection method, system and equipment, the storage medium and the program product, the ultrasonic inspection model is obtained through training on the basis of the large language model, the trained ultrasonic inspection model is used as a dominant intelligent system, the ultrasonic robot for ultrasonic inspection is driven, and the ultrasonic inspection efficiency is improved. And a more accurate, comfortable and widely applicable ultrasonic scanning scheme is provided in an automatic, intelligent and convenient manner.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an ultrasonic inspection method, system, equipment, storage medium and program product. Background Art

[0002] Ultrasound is a widely used imaging technique in the medical field, especially in the examination of carotid artery vessels. Currently, ultrasound examination is usually performed manually by professional medical personnel, relying on their professional knowledge and skills to obtain and interpret images. Handheld ultrasound instruments require manual operation by the operator, which may be affected by the operator's skills and experience, resulting in subjectivity and errors.

[0003] In recent years, with the advancement of technology, robot-assisted ultrasound systems have emerged, aiming to improve the accuracy and efficiency of scanning. However, most of these systems rely on preset programs and have limited adaptive capabilities. Summary of the invention

[0004] The present invention provides an ultrasonic examination method, system, device, storage medium and program product, which are used to solve the defects in the prior art that medical personnel manually perform ultrasonic examinations, resulting in subjective errors, and the robot-assisted ultrasonic system has poor adaptability.

[0005] The present invention provides an ultrasonic inspection method, comprising: Get the command text; Based on the ultrasonic inspection model, the instruction text is subjected to intent recognition to obtain the interface function corresponding to the instruction text, wherein the ultrasonic inspection model is obtained by training on the basis of a large language model using sample instruction texts and their corresponding sample interface functions; The interface function is parameter parsed to obtain execution parameters, and the ultrasonic robot is controlled to perform ultrasonic inspection based on the execution parameters.

[0006] According to the ultrasonic inspection method provided by the present invention, the training step of the ultrasonic inspection model includes: Acquire corpus text related to the ultrasonic robot, and pre-train the large language model based on the corpus text to obtain a pre-trained ultrasonic examination model; Determine a sample instruction text and a sample interface function corresponding thereto based on the corpus text; Inputting the sample instruction text into the pre-trained ultrasonic inspection model to obtain a prediction interface function output by the pre-trained ultrasonic inspection model; Based on the difference between the sample interface function and the predicted interface function, the pre-trained ultrasound examination model is fine-tuned to obtain the ultrasound examination model.

[0007] According to the ultrasonic inspection method provided by the present invention, the step of determining the sample instruction text and its corresponding sample interface function based on the corpus text includes: The corpus text is standardized, and based on the standardized corpus text, an initial sample instruction text and its corresponding initial sample interface function are determined; The format of the initial sample instruction text and the corresponding initial sample interface function are converted to obtain the sample instruction text and the corresponding sample interface function.

[0008] According to the ultrasonic inspection method provided by the present invention, the instruction text is subjected to intention recognition based on the ultrasonic inspection model to obtain the interface function corresponding to the instruction text, including: Based on the knowledge base, relevant information retrieval is performed on the instruction text to obtain retrieval results; The instruction text and the search result are input into the ultrasonic inspection model to obtain the corresponding interface function output by the ultrasonic inspection model.

[0009] According to the ultrasonic inspection method provided by the present invention, the controlling the ultrasonic robot to perform ultrasonic inspection based on the execution parameters includes: Controlling the ultrasonic robot to perform ultrasonic inspection based on the execution parameters to obtain the current inspection result; The current inspection result is returned to the ultrasound inspection model so that the ultrasound inspection model makes a decision based on the current inspection result.

[0010] According to the ultrasonic inspection method provided by the present invention, the method further comprises: The instruction text and ultrasound examination results are displayed on the user interface.

[0011] The present invention also provides an ultrasonic inspection system, comprising: A text acquisition unit, used for acquiring instruction text; An interface determination unit, configured to perform intent recognition on the instruction text based on an ultrasonic inspection model to obtain an interface function corresponding to the instruction text, wherein the ultrasonic inspection model is obtained by training on the basis of a large language model using sample instruction texts and their corresponding sample interface functions; The ultrasonic inspection unit is used to perform parameter analysis on the interface function to obtain execution parameters, and control the ultrasonic robot to perform ultrasonic inspection based on the execution parameters.

[0012] The present invention also 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 program, any one of the ultrasonic inspection methods described above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the ultrasonic inspection method described in any one of the above is implemented.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the ultrasonic inspection method described above is implemented.

[0015] The ultrasonic inspection method, system, device, storage medium and program product provided by the present invention train an ultrasonic inspection model based on a large language model, and use the trained ultrasonic inspection model as a leading intelligent system to drive an ultrasonic robot for ultrasonic inspection. The ultrasonic robot, as a carrier for executing the intelligent decision made by the ultrasonic inspection model, provides a more accurate, comfortable and widely applicable ultrasonic scanning method in an automated, intelligent and convenient manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 This is one of the flow charts of the ultrasonic inspection method provided by the present invention.

[0018] Figure 2 This is the second flow chart of the ultrasonic inspection method provided by the present invention.

[0019] Figure 3 It is a structural schematic diagram of the ultrasonic inspection system provided by the present invention.

[0020] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Existing methods of neck vascular ultrasound examination, whether manual or robot-assisted, have some degree of accuracy and consistency problems. Manual operation is limited by the operator's skills and experience, and the quality of ultrasound examination depends largely on the operator's professional skills and experience; while the robot-assisted system has difficulty adapting to complex medical scenarios due to its lack of advanced cognitive capabilities.

[0023] In addition, traditional ultrasound scanning equipment has limitations in operational flexibility and adaptability, especially when facing patients with complex anatomical structures or significant individual differences, resulting in poor patient comfort and satisfaction.

[0024] In response to the above problems, an embodiment of the present invention provides an ultrasonic inspection method. In the method, an instruction text is first obtained, and then the intent of the instruction text is recognized based on an ultrasonic inspection model to obtain an interface function corresponding to the instruction text. The ultrasonic inspection model is obtained by training on the basis of a large language model using sample instruction texts and their corresponding sample interface functions; the interface function is parameter parsed to obtain execution parameters, and the ultrasonic robot is controlled to perform ultrasonic inspection based on the execution parameters.

[0025] The ultrasonic inspection method provided by the embodiment of the present invention trains an ultrasonic inspection model based on a large language model, and uses the trained ultrasonic inspection model as a leading intelligent system to drive an ultrasonic robot for ultrasonic inspection. The ultrasonic robot, as a carrier for executing the intelligent decision made by the ultrasonic inspection model, provides a more accurate, comfortable and widely applicable ultrasonic scanning method in an automated, intelligent and convenient manner.

[0026] The embodiments of the present invention can be applied to scenarios where ultrasonic examination is required, such as whole-body ultrasonic examination, such as medical imaging and diagnosis scenarios such as carotid artery, abdomen, musculoskeletal system, etc.; it can also be applied to telemedicine services, by remotely controlling ultrasonic robots to perform ultrasonic examinations; it can also be applied to medical education and professional training to help students and doctors learn ultrasonic examination techniques and interpret results; it can also be used during surgery to provide surgeons with real-time ultrasonic imaging support to increase the safety and success rate of surgery. The execution subject of this method can be an electronic device such as a terminal device, a computer, a server, a server cluster, or a specially designed ultrasonic examination device, or it can be an ultrasonic examination system set in the electronic device, and the ultrasonic examination system can be implemented by a combination of software and hardware.

[0027] Figure 1 It is one of the flow charts of the ultrasonic inspection method provided by the present invention, such as Figure 1 As shown, the ultrasonic inspection method comprises the following steps: Step 110, obtaining the instruction text.

[0028] Specifically, the instruction text refers to text that can express user needs, and can be an instruction in a natural language form input by the user.

[0029] The instruction text may be text directly input by the user; it may also be obtained by the user inputting voice and transcribing the collected audio; it may also be obtained by the user inputting an image and performing OCR (Optical Character Recognition) or intent recognition on the image, and the embodiments of the present invention do not specifically limit this.

[0030] The instruction text here may be in any language, and the ultrasound examination method may support multiple languages, such as Chinese, English, Russian, and French, etc. For example, the instruction text may be “Please perform a carotid - ultrasound scan of the neck” or “Please perform a carotid - ultrasound scan of the neck”, etc.

[0031] It should be noted that the requirement text may be a deterministic instruction or an uncertain instruction, and the embodiment of the present invention does not specifically limit this.

[0032] Step 120, based on the ultrasonic inspection model, the instruction text is subjected to intent recognition to obtain the interface function corresponding to the instruction text. The ultrasonic inspection model is obtained by training on the basis of a large language model using sample instruction texts and their corresponding sample interface functions.

[0033] Specifically, the embodiment of the present invention uses the trained ultrasound inspection model as the leading intelligent system to drive the ultrasound robot for ultrasound inspection. Therefore, after obtaining the instruction text, it is necessary to perform intent recognition on the instruction text to obtain the interface function corresponding to the instruction text.

[0034] It should be noted that the ultrasonic inspection model may be trained before executing step 120. In this embodiment, the ultrasonic inspection model is obtained by training on the basis of a large language model using sample instruction texts and their corresponding sample interface functions.

[0035] Large Language Model (LLM), also known as Large Model, refers to a Natural Language Processing (NLP) model with a large number of parameters. The model processes large amounts of text data during training and has the ability to understand and generate natural language. For example, Large Language Models can include ChatGLM-6B and its derivative models.

[0036] In some embodiments, the steps of training the ultrasound inspection model include: Acquire corpus text related to the ultrasonic robot, and pre-train a large language model based on the corpus text to obtain a pre-trained ultrasonic examination model; Determine a sample instruction text and its corresponding sample interface function based on the corpus text; Inputting the sample instruction text into the pre-trained ultrasound inspection model to obtain the prediction interface function output by the pre-trained ultrasound inspection model; Based on the difference between the sample interface function and the predicted interface function, the pre-trained ultrasound inspection model is fine-tuned to obtain the ultrasound inspection model.

[0037] Specifically, the large language model is pre-trained and fine-tuned for the specific needs of the ultrasonic robot. The second pre-training here refers to a transitional stage between pre-training and supervised fine-tuning, which retains the extensiveness of pre-training while adding specialization for specific tasks.

[0038] For the ultrasonic robot performing ultrasound examination scenario, we first collect text corpus related to the operation of the ultrasonic robot, which may include the robot's operating manual, code repository, medical terms, operating instructions, and the robot's code.

[0039] Based on the corpus text, the large language model is pre-trained to obtain a pre-trained ultrasound examination model. Using the collected corpus text, the data is combined in the following format: [

[0040] {

[0041] "text": "xxx"

[0042] },

[0043] {

[0044] "text": "xxx"

[0045] }, ... ]

[0048] According to the "next word prediction" method, the second pre-training is performed. The loss calculated for each text here will be used for gradient backpropagation. The loss calculation method can be cross entropy loss.

[0049] Then, based on the corpus text, the sample instruction text and its corresponding sample interface function are determined to perform supervised fine-tuning on the pre-trained ultrasound inspection model. Supervised fine-tuning here refers to the use of supervised learning methods to enhance the ability of the pre-trained ultrasound inspection model to follow specific format instructions in this scenario. The sample instruction text and its corresponding sample interface function should cover various instructions and corresponding operations that may be encountered in ultrasound inspection.

[0050] The corpus text may be first standardized, including data cleaning, word segmentation, and removal of irrelevant data to ensure data quality and relevance. Based on the standardized corpus text, the initial sample instruction text and its corresponding initial sample interface function are determined. The initial sample instruction text and its corresponding initial sample interface function may be obtained by combining the standardized corpus text according to a preset format.

[0051] On this basis, the format of the initial sample instruction text and the corresponding initial sample interface function are converted to obtain the sample instruction text and the corresponding sample interface function.

[0052] For the ultrasound inspection scenario, first determine the data format. Then convert the "initial sample instruction text-corresponding initial sample interface function" data into the corresponding data format. Then perform training according to the "next word prediction" method. Note that only the output loss is used for gradient return, and the loss calculation method can be the cross entropy loss.

[0053] The data for supervised fine-tuning is combined in the following format: [

[0054] {

[0055] "instruction": "xxxx", "input": "xxxx", "output": "xxxx"

[0056] },

[0057] {

[0058] "instruction": "xxxx", "input": "xxxx", "output": "xxxx"

[0059] }, ... ]

[0062] The ultrasound inspection model trained by the above method can realize efficient natural language understanding and processing, and generate the API (Application Programming Interface) of the ultrasound robot that needs to be called. That is, the interface function, which can be the best interface calling method.

[0063] For example, the instruction text is "Please perform carotid artery ultrasound scanning", and the interface function corresponding to the instruction text may be "Run robot".

[0064] In other embodiments, based on the ultrasonic inspection model, the instruction text is subjected to intent recognition to obtain the interface function corresponding to the instruction text, including: Based on the knowledge base, relevant information is retrieved for the instruction text to obtain the retrieval results; The instruction text and the search results are input into the ultrasonic inspection model to obtain the corresponding interface function output by the ultrasonic inspection model.

[0065] Specifically, in order to enable the ultrasonic detection model to more accurately identify user intentions, ensure accurate parsing of instructions, and make the generated interface functions more in line with user intentions, this embodiment first performs relevant information retrieval on the instruction text based on the knowledge base to obtain retrieval results, and performs data retrieval through multiple knowledge bases to provide comprehensive information support.

[0066] On this basis, the instruction text and the search results are input into the ultrasonic inspection model to obtain the corresponding interface function output by the ultrasonic inspection model.

[0067] Step 130 , performing parameter parsing on the interface function to obtain execution parameters, and controlling the ultrasonic robot to perform ultrasonic inspection based on the execution parameters.

[0068] Specifically, the ultrasonic inspection model and the ultrasonic robot can achieve effective docking and collaborative work. After the ultrasonic inspection model generates an interface function, the interface function can be further parameter parsed to obtain execution parameters. The ultrasonic robot performs specific ultrasonic scanning operations according to the execution parameters obtained by parsing.

[0069] In some other embodiments, controlling the ultrasonic robot to perform ultrasonic inspection based on the execution parameters includes: Control the ultrasonic robot to perform ultrasonic inspection based on the execution parameters and obtain the current inspection result; The current examination result is returned to the ultrasound examination model so that the ultrasound examination model makes a decision based on the current examination result.

[0070] Specifically, after each ultrasonic inspection, the ultrasonic robot obtains the current inspection result and returns the current inspection result to the ultrasonic inspection model for the ultrasonic inspection model to make a decision based on the current inspection result. The decision here may be, for example, to perform the operation again or to end the operation.

[0071] If the ultrasonic inspection model determines that the operation should be performed again, the ultrasonic robot is controlled to perform the ultrasonic inspection again; if the ultrasonic inspection model determines that the operation should be ended, the ultrasonic robot is controlled to end the ultrasonic inspection.

[0072] Based on any of the above embodiments, the ultrasonic inspection method further includes: The user interface displays instruction texts, interface functions and ultrasound examination results.

[0073] Specifically, in order to enable medical professionals to easily monitor and guide the operation of the robot, an intuitive user interface can also be designed. The user interface can interact with the user by displaying instruction text, interface functions, and ultrasound examination results.

[0074] Based on any of the above embodiments, Figure 2 FIG. 2 is a flow chart of the ultrasonic inspection method provided by the present invention. Figure 2 As shown, the method includes: S1, Data preparation and preprocessing.

[0075] Using efficient data collection tools, we focus on collecting text corpus related to ultrasound robot operation, including medical terms, operating instructions, etc.

[0076] The corpus text is standardized, and based on the standardized corpus text, the initial sample instruction text and its corresponding initial sample interface function are determined; the format of the initial sample instruction text and its corresponding initial sample interface function is converted to obtain the sample instruction text and its corresponding sample interface function, so as to prepare for subsequent model training.

[0077] S2, model training.

[0078] Secondary pre-training of the large language model: Based on the general large language model, secondary pre-training is performed using specific ultrasonic robot operation corpus to enhance the model's professionalism in this field.

[0079] Supervised fine-tuning of large language models: The model is further fine-tuned through structured data to precisely adapt to the specific needs of ultrasound scanning.

[0080] S3, model evaluation and optimization.

[0081] Thoroughly test the model with the validation dataset and evaluate its performance on key metrics such as accuracy, recall, and F1 score. Adjust model parameters based on the evaluation results and perform detailed optimization to improve the overall performance and accuracy of the model. Implement a continuous training cycle to adapt to new data and the changing medical environment.

[0082] S4, model deployment.

[0083] Export the trained model to a format compatible with the actual application environment. Integrate the exported model into the ultrasound robot system to ensure seamless collaboration between the model and the robot. When the model receives natural language instructions, it can process them in real time and provide corresponding operation guidance.

[0084] S5, combined with embodied systems.

[0085] Realize the effective connection between the large language model and the robot system to ensure the accurate transmission of instructions. Optimize the natural language input mechanism to improve the accuracy and efficiency of language understanding. Optimize the interface calling process to reduce latency and increase response speed. Further optimize the system based on actual application performance to ensure efficient and stable operation.

[0086] S6, operation execution and feedback.

[0087] Combine multiple knowledge bases for data retrieval and provide comprehensive information support. Use the large language model to identify user intent and ensure accurate parsing of instructions. Determine the best interface calling method through the large language model. The embodied system parses the parameters provided by the large language model and prepares to perform operations. The robot performs specific ultrasound scanning operations based on the parsed parameters. After the robot completes the operation, it returns the results to the large language model intention recognition module for the large language model to make decisions. Complete the entire ultrasound scanning process and provide necessary diagnostic information.

[0088] The present invention provides an innovative solution for ultrasound examination by combining the technology of large language model and ultrasound robot. This integration brings significant performance improvement, especially in terms of accuracy, efficiency, adaptability and user experience. By utilizing the advanced data processing and analysis capabilities of large language model and the precise and flexible operation of robot, the present invention can optimize the ultrasound examination process and provide more accurate and reliable diagnosis results.

[0089] Specifically, automated robotic operation reduces the time for manual setup and adjustment, speeds up the scanning process, and improves overall examination efficiency. The system can automatically adjust to the specific conditions of different patients and adapt to various anatomical structures and specific medical needs. It reduces the reliance on the operator's advanced ultrasound operation skills, enabling more medical staff to perform high-quality ultrasound examinations. By reducing examination time and improving diagnostic accuracy, patients' comfort and satisfaction with medical services are enhanced. The continuous learning ability of the large language model ensures that the performance of the system will continue to improve over time and adapt to new data sets. Using large language models to process and parse large amounts of data provides doctors with strong data-based support to assist them in making more accurate diagnostic decisions.

[0090] The method provided by the embodiment of the present invention can provide a fast and non-invasive examination, reduce the pressure on the body of elderly patients, and is particularly suitable for elderly patients. At the same time, as a research case of the application of artificial intelligence in the medical field, it can promote the application and exploration of artificial intelligence in more medical fields.

[0091] Based on any of the above embodiments, Figure 3 Schematic diagram of the structure of the ultrasonic inspection system provided by the present invention. Figure 3 As shown, the ultrasound inspection system includes: A text acquisition unit 310, used to acquire instruction text; The interface determination unit 320 is used to perform intent recognition on the instruction text based on an ultrasonic inspection model to obtain an interface function corresponding to the instruction text, wherein the ultrasonic inspection model is obtained by training on the basis of a large language model using sample instruction texts and their corresponding sample interface functions; The ultrasonic inspection unit 330 is used to perform parameter analysis on the interface function to obtain execution parameters, and control the ultrasonic robot to perform ultrasonic inspection based on the execution parameters.

[0092] The system provided by the embodiment of the present invention trains an ultrasound examination model based on a large language model, and uses the trained ultrasound examination model as a leading intelligent system to drive an ultrasound robot for ultrasound examination. The ultrasound robot, as a carrier for executing the intelligent decisions made by the ultrasound examination model, provides a more accurate, comfortable and widely applicable ultrasound scanning solution in an automated, intelligent and convenient manner.

[0093] Based on any of the above embodiments, it also includes a model training unit, which is used to: Acquire corpus text related to the ultrasonic robot, and pre-train the large language model based on the corpus text to obtain a pre-trained ultrasonic examination model; Determine a sample instruction text and a sample interface function corresponding thereto based on the corpus text; Inputting the sample instruction text into the pre-trained ultrasonic inspection model to obtain a prediction interface function output by the pre-trained ultrasonic inspection model; Based on the difference between the sample interface function and the predicted interface function, the pre-trained ultrasound examination model is fine-tuned to obtain the ultrasound examination model.

[0094] Based on any of the above embodiments, the model training unit is specifically used for: The corpus text is standardized, and based on the standardized corpus text, an initial sample instruction text and its corresponding initial sample interface function are determined; The format of the initial sample instruction text and the corresponding initial sample interface function are converted to obtain the sample instruction text and the corresponding sample interface function.

[0095] Based on any of the foregoing embodiments, the interface determination unit is specifically configured to: Based on the knowledge base, relevant information retrieval is performed on the instruction text to obtain retrieval results; The instruction text and the search result are input into the ultrasonic inspection model to obtain the corresponding interface function output by the ultrasonic inspection model.

[0096] Based on any of the above embodiments, the ultrasonic inspection unit is specifically used for: Controlling the ultrasonic robot to perform ultrasonic inspection based on the execution parameters to obtain the current inspection result; The current inspection result is returned to the ultrasound inspection model so that the ultrasound inspection model makes a decision based on the current inspection result.

[0097] Based on any of the above embodiments, it further includes a display unit, which is used for: The instruction text and ultrasound examination results are displayed on the user interface.

[0098] Based on any of the above embodiments, an ultrasonic inspection system is provided, comprising: Data processing layer: responsible for collecting, cleaning, and preprocessing corpus data related to ultrasonic robots, including medical terms, operating instructions, etc.

[0099] Model training layer: Use the cleaned and preprocessed data to perform secondary pre-training and supervised fine-tuning on the large language model, specifically for ultrasonic robot application scenarios.

[0100] Model evaluation and optimization layer: Perform performance evaluation on the trained model, including indicators such as accuracy, recall rate, and F1 score, and make optimization adjustments based on the evaluation results.

[0101] Model deployment layer: Integrate the optimized model into the ultrasound robot system to ensure the model's real-time processing and natural language understanding capabilities.

[0102] Embodied system integration layer: realize the interface connection between the large language model and the ultrasonic robot, optimize the interface call link, and perform performance optimization.

[0103] Operation execution and feedback layer: including knowledge base merge retrieval, intent recognition, interface function determination, parameter parsing, robot operation execution and return of operation results.

[0104] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the ultrasonic inspection method, which includes: obtaining the instruction text; based on the ultrasonic inspection model, performing intent recognition on the instruction text to obtain the interface function corresponding to the instruction text, the ultrasonic inspection model is obtained by applying the sample instruction text and the corresponding sample interface function, and training on the basis of the large language model; performing parameter parsing on the interface function to obtain the execution parameters, and controlling the ultrasonic robot to perform the ultrasonic inspection based on the execution parameters.

[0105] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0106] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the ultrasonic inspection method provided by the above-mentioned methods, and the method includes: obtaining an instruction text; based on an ultrasonic inspection model, performing intent recognition on the instruction text to obtain an interface function corresponding to the instruction text, and the ultrasonic inspection model is obtained by training on the basis of a large language model using sample instruction texts and their corresponding sample interface functions; performing parameter parsing on the interface function to obtain execution parameters, and controlling the ultrasonic robot to perform ultrasonic inspection based on the execution parameters.

[0107] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the ultrasonic inspection method provided by the above-mentioned methods, the method comprising: obtaining an instruction text; based on an ultrasonic inspection model, performing intent recognition on the instruction text to obtain an interface function corresponding to the instruction text, the ultrasonic inspection model being trained on the basis of a large language model by applying sample instruction texts and their corresponding sample interface functions; performing parameter parsing on the interface function to obtain execution parameters, and controlling the ultrasonic robot to perform ultrasonic inspection based on the execution parameters.

[0108] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0109] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An ultrasonic inspection method, characterized in that: include: Get the command text; Based on the ultrasonic inspection model, the instruction text is subjected to intent recognition to obtain the interface function corresponding to the instruction text, wherein the ultrasonic inspection model is obtained by training on the basis of a large language model using sample instruction texts and their corresponding sample interface functions; The interface function is parameter parsed to obtain execution parameters, and the ultrasonic robot is controlled to perform ultrasonic inspection based on the execution parameters.

2. The ultrasonic inspection method according to claim 1, characterized in that: The training steps of the ultrasound examination model include: Acquire corpus text related to the ultrasonic robot, and pre-train the large language model based on the corpus text to obtain a pre-trained ultrasonic examination model; Determine a sample instruction text and a sample interface function corresponding thereto based on the corpus text; Inputting the sample instruction text into the pre-trained ultrasonic inspection model to obtain a prediction interface function output by the pre-trained ultrasonic inspection model; Based on the difference between the sample interface function and the predicted interface function, the pre-trained ultrasound examination model is fine-tuned to obtain the ultrasound examination model.

3. The ultrasonic inspection method according to claim 2, characterized in that: The step of determining a sample instruction text and a sample interface function corresponding to the sample instruction text based on the corpus text includes: The corpus text is standardized, and based on the standardized corpus text, an initial sample instruction text and its corresponding initial sample interface function are determined; The format of the initial sample instruction text and the corresponding initial sample interface function are converted to obtain the sample instruction text and the corresponding sample interface function.

4. The ultrasonic inspection method according to claim 1, characterized in that: The method of performing intent recognition on the instruction text based on the ultrasonic inspection model to obtain an interface function corresponding to the instruction text includes: Based on the knowledge base, relevant information retrieval is performed on the instruction text to obtain retrieval results; The instruction text and the search result are input into the ultrasonic inspection model to obtain the corresponding interface function output by the ultrasonic inspection model.

5. The ultrasonic inspection method according to claim 1, characterized in that: The controlling the ultrasonic robot to perform ultrasonic inspection based on the execution parameters comprises: Controlling the ultrasonic robot to perform ultrasonic inspection based on the execution parameters to obtain the current inspection result; The current inspection result is returned to the ultrasound inspection model so that the ultrasound inspection model makes a decision based on the current inspection result.

6. The ultrasonic inspection method according to claim 1, characterized in that: The method further comprises: The instruction text and ultrasound examination results are displayed on the user interface.

7. An ultrasonic inspection system, characterized in that: include: A text acquisition unit, used for acquiring instruction text; An interface determination unit, configured to perform intent recognition on the instruction text based on an ultrasonic inspection model to obtain an interface function corresponding to the instruction text, wherein the ultrasonic inspection model is obtained by training on the basis of a large language model using sample instruction texts and their corresponding sample interface functions; The ultrasonic inspection unit is used to perform parameter analysis on the interface function to obtain execution parameters, and control the ultrasonic robot to perform ultrasonic inspection based on the execution parameters.

8. 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 program, the ultrasonic inspection method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the ultrasonic inspection method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the ultrasonic inspection method according to any one of claims 1 to 6 is implemented.

Images