Data processing method and device based on dynamic routing mechanism and computer equipment

Abstract

The invention relates to a data processing method and device based on a dynamic routing mechanism, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring physical examination data of a target object, and extracting symptoms of a target patient from the physical examination data; querying a plurality of candidate potential diseases structurally associated with the target patient symptom from a structured knowledge graph, wherein the structured knowledge graph comprises the plurality of potential diseases, the patient symptom and a membership structure relationship between each potential disease and the patient symptom; activating diagnosis sub-models corresponding to the plurality of candidate potential diseases in the diagnosis model; and based on the diagnosis result of the diagnosis sub-model, determining a target potential disease corresponding to the physical examination data in the plurality of candidate potential diseases. By adopting the method, general disease classification can be remarkably improved, knowledge and experience from experts in different fields are combined, and the accuracy and reliability of a diagnosis conclusion are improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and particularly to a data processing method, apparatus, computer device, computer-readable storage medium, and computer program product based on a dynamic routing mechanism. Background Art

[0002] With the rapid development of the large model field and the continuous iterative update of related technologies, the role of large models in the medical field covers multiple aspects such as diagnostic support, personalized treatment recommendations, patient education, drug research and development, and data analysis.

[0003] In many regions, especially in rural or remote areas, diagnostic equipment and technical resources are limited, and doctors can only rely on basic tools or referrals, making it easier to have misdiagnosis or missed diagnosis. Therefore, current artificial intelligence technologies represented by large models can play a very important role in this field, but there is currently no model diagnosis method that can significantly improve the classification of general diseases. Summary of the Invention

[0004] Based on this, in view of the above technical problems, it is necessary to provide a data processing method, apparatus, computer device, computer-readable storage medium, and computer program product based on a dynamic routing mechanism that can significantly improve the classification of general diseases.

[0005] In a first aspect, this application provides a data processing method based on a dynamic routing mechanism, including:

[0006] Obtain the physical examination data of the target object, and extract the target patient symptoms from the physical examination data;

[0007] Query multiple candidate potential diseases that are structurally associated with the target patient symptoms from a structured knowledge graph, where the structured knowledge graph includes multiple potential diseases, patient symptoms, and the membership structure relationship between each potential disease and patient symptom;

[0008] Activate the respective diagnostic sub-models corresponding to the multiple candidate potential diseases in the diagnostic model;

[0009] Based on the diagnostic results of the diagnostic sub-models, determine the target potential disease corresponding to the physical examination data among the multiple candidate potential diseases.

[0010] In one embodiment, querying multiple candidate potential diseases that are structurally associated with the target patient symptoms from a structured knowledge graph includes:

[0011] Query the preliminary potential diseases including the target patient's symptoms from the structured knowledge graph; based on the structured knowledge graph, determine all the patient symptoms of the preliminary potential diseases; if the target patient's symptoms are the same as all the patient symptoms, then use the preliminary potential disease as the candidate potential disease.

[0012] In one embodiment, based on the diagnosis result of the diagnosis sub-model, determining the target potential disease corresponding to the physical examination data among the multiple candidate potential diseases includes:

[0013] For each candidate potential disease, use the diagnosis sub-model corresponding to the candidate potential disease to predict the physical examination data, and obtain the confidence score that the target object belongs to the candidate potential disease; based on the confidence scores that the target object belongs to each candidate potential disease, determine the target potential disease corresponding to the physical examination data.

[0014] In one embodiment, for each candidate potential disease, using the diagnosis sub-model corresponding to the candidate potential disease to predict the physical examination data, and obtaining the confidence score that the target object belongs to the candidate potential disease includes:

[0015] Use the diagnosis sub-model corresponding to the candidate potential disease to extract features from the physical examination data to obtain a healthy feature representation; convert the healthy feature representation into a disease probability distribution feature vector; based on the disease probability distribution feature vector, determine the confidence score that the target object belongs to the candidate potential disease.

[0016] In one embodiment, based on the confidence scores that the target object belongs to each candidate potential disease, determining the target potential disease corresponding to the physical examination data includes:

[0017] Obtain the rank relationship of each of the multiple candidate potential diseases from the structured knowledge graph; based on the rank relationship, determine the upper-level potential diseases, lower-level potential diseases, and isolated potential diseases among the multiple candidate potential diseases; based on the confidence scores of the upper-level potential diseases, the lower-level potential diseases, and the isolated potential diseases respectively, determine the target potential disease corresponding to the physical examination data.

[0018] In one embodiment, based on the confidence scores of the upper-level potential diseases, the lower-level potential diseases, and the isolated potential diseases respectively, determining the target potential disease corresponding to the physical examination data includes:

[0019] Compare the confidence score of the upper-level potential disease with the first confidence threshold; if the confidence score of the upper-level potential disease exceeds the first confidence threshold, adjust the confidence scores of the lower-level potential diseases that are at the same hierarchical level as the upper-level potential disease to zero; compare the confidence score of the upper-level potential disease and the confidence score of the isolated potential disease with the second confidence threshold respectively; based on the comparison results, determine the target potential disease corresponding to the physical examination data.

[0020] In one embodiment, query treatment suggestions related to the target potential disease from the structured knowledge graph; perform risk screening on the treatment suggestions through pre-imported medical specification data; if there are risk items in the treatment suggestions, adjust the risk items in the treatment suggestions based on the medical specification data; send the adjusted treatment suggestions to the target object.

[0021] In a second aspect, the present application also provides a data processing device based on a dynamic routing mechanism, including:

[0022] An extraction module, configured to obtain the physical examination data of a target object and extract target patient symptoms from the physical examination data;

[0023] A query module, configured to query multiple candidate potential diseases structurally associated with the target patient symptoms from a structured knowledge graph, where the structured knowledge graph includes multiple potential diseases, patient symptoms, and the membership structure relationship between each potential disease and patient symptom;

[0024] An activation module, configured to activate the respective diagnostic sub-models of the multiple candidate potential diseases in the diagnostic model;

[0025] A determination module, configured to determine the target potential disease corresponding to the physical examination data among the multiple candidate potential diseases based on the diagnostic results of the diagnostic sub-models.

[0026] In a third aspect, the present application also provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0027] Obtain the physical examination data of a target object and extract target patient symptoms from the physical examination data;

[0028] Query multiple candidate potential diseases structurally associated with the target patient symptoms from a structured knowledge graph, where the structured knowledge graph includes multiple potential diseases, patient symptoms, and the membership structure relationship between each potential disease and patient symptom;

[0029] Activate the respective diagnostic sub-models corresponding to the multiple candidate potential diseases in the diagnostic model;

[0030] Based on the diagnostic results of the diagnostic sub-models, determine the target potential disease corresponding to the physical examination data among the multiple candidate potential diseases.

[0031] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0032] Obtain the physical examination data of a target object, and extract target patient symptoms from the physical examination data;

[0033] Query, from a structured knowledge graph, multiple candidate potential diseases that are structurally associated with the target patient symptoms. The structured knowledge graph includes multiple potential diseases, patient symptoms, and the subordinate structural relationships between each potential disease and patient symptom;

[0034] Activate the respective diagnostic sub-models corresponding to the multiple candidate potential diseases in the diagnostic model;

[0035] Based on the diagnostic results of the diagnostic sub-models, determine the target potential disease corresponding to the physical examination data among the multiple candidate potential diseases.

[0036] In a fifth aspect, the present application also provides a computer program product, including a computer program. When the computer program is executed by a processor, the following steps are implemented:

[0037] Obtain the physical examination data of a target object, and extract target patient symptoms from the physical examination data;

[0038] Query, from a structured knowledge graph, multiple candidate potential diseases that are structurally associated with the target patient symptoms. The structured knowledge graph includes multiple potential diseases, patient symptoms, and the subordinate structural relationships between each potential disease and patient symptom;

[0039] Activate the respective diagnostic sub-models corresponding to the multiple candidate potential diseases in the diagnostic model;

[0040] Based on the diagnostic results of the diagnostic sub-models, determine the target potential disease corresponding to the physical examination data among the multiple candidate potential diseases.

[0041] The above data processing method, device, computer device, computer-readable storage medium, and computer program product based on a dynamic routing mechanism ensure the comprehensiveness and accuracy of information by obtaining the physical examination data of a target object and extracting target patient symptoms from the physical examination data. This is particularly important for regions lacking advanced diagnostic equipment and technical resources.

[0042] Query multiple candidate potential diseases that are structurally associated with the target patient's symptoms from a structured knowledge graph, which includes multiple potential diseases, patient symptoms, and the membership structure relationships between each potential disease and patient symptoms. Based on the patient's symptoms, quickly locate multiple potentially relevant diseases. This not only accelerates the process of preliminary diagnosis but also provides a direction for subsequent precise diagnosis, helps narrow the scope of further examinations, and reduces the risks of unnecessary referrals and misdiagnosis.

[0043] Activate the respective diagnostic sub-models corresponding to multiple candidate potential diseases in the diagnostic model. Adopt a dynamic routing mechanism to select the "expert" diagnostic sub-model that is most suitable for handling a specific disease condition. This approach enables the system to provide more professional and personalized diagnostic suggestions according to the patient's specific condition. Compared with a single generalized model, this method can handle complex and diverse medical problems more effectively and improve the diagnostic accuracy.

[0044] Based on the diagnostic results of the diagnostic sub-models, determine the target potential disease corresponding to the physical examination data among multiple candidate potential diseases. Finally, integrate the results of each diagnostic sub-model to determine the most likely disease. This process combines the knowledge and experience of experts from different fields and improves the accuracy and reliability of the diagnostic conclusion. Description of the Drawings

[0045] To more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments of the present application or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0046] Figure 1 It is an application environment diagram of a data processing method based on a dynamic routing mechanism in an embodiment;

[0047] Figure 2 It is a schematic flowchart of a data processing method based on a dynamic routing mechanism in an embodiment;

[0048] Figure 3 It is a schematic architecture diagram of a data processing method based on a dynamic routing mechanism in an embodiment;

[0049] Figure 4 It is a structural block diagram of a data processing device based on a dynamic routing mechanism in an embodiment;

[0050] Figure 5 It is a structural block diagram of a data processing device based on a dynamic routing mechanism in another embodiment;

[0051] Figure 6 It is the internal structure diagram of a computer device in an embodiment. Specific implementation manners

[0052] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be further described in detail below with reference to 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 used to limit the present application.

[0053] The data processing method based on a dynamic routing mechanism provided by an embodiment of the present application can be applied to, for example Figure 1 the application environment shown. Among them, the terminal 102 communicates with the server 104 through a network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or can be placed in the cloud or on other network servers. The terminal 102 is used to generate a data processing request based on a dynamic routing mechanism, and send the data processing request based on the dynamic routing mechanism to the server 104, so that the server 104 determines the target potential disease corresponding to the physical examination data among multiple candidate potential diseases based on the diagnosis result of the diagnosis sub-model. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, projection devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0054] In an exemplary embodiment, as Figure 2 shown, a data processing method based on a dynamic routing mechanism is provided. Taking the method applied to the Figure 1 server 104 in as an example for description, it includes the following steps 202 to step 208. Among them:

[0055] Step 202, obtain the physical examination data of the target object, and extract the target patient symptoms from the physical examination data.

[0056] Among them, the data processing method based on a dynamic routing mechanism in the present application is applied to a diagnosis model, and the diagnosis model includes multiple diagnosis sub-models for different potential diseases.

[0057] A diagnostic model refers to a large model system based on a hybrid model architecture, which is used for the classification of general diseases and treatment recommendations. The diagnostic model contains multiple diagnostic sub-models for different potential diseases, and selects the most appropriate sub-model to handle specific diseases through a dynamic routing mechanism.

[0058] Potential diseases refer to various possible disease types defined in the knowledge graph. These diseases have a subordinate relationship with specific symptoms, signs, and examination results, which are the basis for the system to classify diseases.

[0059] A diagnostic sub-model refers to an independent neural network module that focuses on a specific category of diseases (such as cardiovascular diseases, respiratory diseases, etc.). After specialized training, it can provide accurate diagnostic support for specific types of diseases.

[0060] The dynamic routing mechanism is one of the key features in the hybrid model architecture, which allows the automatic determination of which experts (i.e., diagnostic sub-models) should be activated to participate in the decision-making process based on the input data characteristics. This mechanism enables the system to use resources more flexibly and efficiently, improving the prediction accuracy.

[0061] The target object refers to the patient who is being diagnosed. Understandably, in this application, the term "target object" is used to distinguish the patient data being currently processed from other data.

[0062] Physical examination data refers to information including but not limited to the patient's symptom description, medical history, family history, various examination results (such as blood tests, imaging reports), etc.

[0063] Target patient symptoms refer to the specific symptom information extracted from the physical examination data of the target object. Understandably, target patient symptoms are the basis for querying the structured knowledge graph, identifying candidate potential diseases, and activating the corresponding diagnostic sub-models.

[0064] Specifically, first extract all medical data related to a specific patient from multiple tables or databases. Ensure the correct matching and collection of all relevant information belonging to the same patient (i.e., the target object) from different sources (such as hospital information systems, electronic health records, etc.). Further clean the original data, including handling missing values, abnormal data, duplicate data, and irrelevant information. Ensure the consistency and accuracy of the output results to prepare for the next data standardization. Adjust the format of data from different tables, such as unifying the time format, units, naming rules, etc., to make the data structure consistent.

[0065] Finally, natural language processing techniques are used to parse text descriptions (such as chief complaints and medical histories), examination results (such as blood test reports and imaging data), and demographic information (such as age and gender). These pieces of information are then integrated to generate a natural language description, namely the symptoms of the target patient, which covers key information such as the patient's symptoms, signs, and examination results. Based on the comprehensive description, specific symptoms that may affect disease classification and treatment recommendations are automatically identified and extracted. These symptoms will be used as inputs to query the knowledge graph to find potential diseases related to them and their relevant information.

[0066] Step 204: Query multiple candidate potential diseases structurally related to the symptoms of the target patient from the structured knowledge graph. The structured knowledge graph includes multiple potential diseases, patient symptoms, and the subordination structural relationships between each potential disease and patient symptom.

[0067] Among them, the structured knowledge graph refers to a special semantic network that stores medical knowledge related to general diseases in a structured manner. In this structured knowledge graph, nodes represent entities (such as diseases, symptoms, treatment plans, etc.), and edges represent the relationships between these entities. Specifically, in this application, the structured knowledge graph contains multiple potential diseases, patient symptoms, and the subordination structural relationships between them. Through this structured organizational form, complex medical information can be better understood and processed to support diagnosis.

[0068] Candidate potential diseases refer to a series of diseases that may be related to these symptoms identified after querying the structured knowledge graph based on the symptoms of the target patient. Understandably, when the system receives the specific symptoms of a patient, it will use the knowledge graph to find diseases directly or indirectly related to these symptoms as preliminary diagnostic hypotheses. These identified diseases are the "candidate potential diseases", which form the basis for further detailed analysis.

[0069] The subordination structural relationship refers to the specific connection between different entities (such as symptoms and diseases) in the structured knowledge graph. For example, a certain symptom may appear in multiple different diseases, but its severity, occurrence frequency, or other related characteristics may vary, which forms a complex subordination relationship between the symptom and the disease. Similarly, treatment plans may also vary depending on the disease. By clarifying these subordination structural relationships, the diseases that may be hidden behind the symptoms can be understood more accurately, and reasonable diagnosis and treatment recommendations can be made accordingly.

[0070] Specifically, the parsed patient symptom information, i.e., the target patient symptoms, is used as a query condition to search for matching potential diseases and their related information in the knowledge graph. For example, based on the subordination structure relationship between each potential disease and the patient symptoms in the knowledge graph, it is determined which diseases are most likely to match the patient's symptoms. This step may involve complex pattern matching algorithms to ensure high-precision identification of relevant diseases.

[0071] Optionally, based on the query results, a list of candidate potential diseases is generated. Each disease is attached with a corresponding confidence score, indicating the strength of the correlation between the disease and the patient symptoms. This list of candidate potential diseases is then used to activate the corresponding diagnostic submodels for further analysis.

[0072] Step 206, activate the diagnostic submodels corresponding to multiple candidate potential diseases in the diagnostic model.

[0073] Among them, the diagnostic submodel refers to an independent neural network module designed specifically for a specific type or category of potential diseases. Each diagnostic submodel is like an "expert" focusing on a certain field (such as cardiovascular diseases, respiratory diseases, etc.), and after being specially trained, it can provide precise diagnostic support for specific types of diseases.

[0074] Specifically, first, according to the nature of each candidate potential disease, the corresponding diagnostic submodel is dynamically selected. These submodels are professional modules trained specifically for a certain category of diseases (such as cardiovascular diseases, respiratory diseases, etc.). It can be understood that this step relies on a dynamic routing mechanism, which can automatically determine which experts (i.e., diagnostic submodels) should be activated to handle a specific case based on the input data characteristics.

[0075] Then, for each candidate potential disease, its corresponding diagnostic submodel is activated. It can be understood that only those submodels considered most suitable for handling the current case will be activated, while other irrelevant submodels remain in a silent state. Optionally, the activation process involves loading the relevant parameters and configurations of the specific submodel so that they can start processing the input data.

[0076] Finally, each activated diagnostic submodel conducts a detailed analysis of the comprehensive diagnostic input of the patient based on the knowledge and patterns learned during its training. The diagnostic submodel will compare the specific condition of the patient with the similarity between the known disease patterns to infer the most likely disease type.

[0077] Step 208, based on the diagnostic results of the diagnostic submodels, determine the target potential disease corresponding to the physical examination data among multiple candidate potential diseases.

[0078] Among them, the diagnostic result refers to the specific conclusion obtained after each activated diagnostic sub-model analyzes the input data (i.e., the patient's physical examination data and symptom information). These conclusions usually include but are not limited to: for each candidate potential disease, the diagnostic sub-model will evaluate the likelihood of its occurrence and give a confidence score.

[0079] The target potential disease refers to one or more specific diseases that are finally determined to be most likely to match the patient's current symptoms based on the diagnostic results of all diagnostic sub-models. Understandably, after considering the diagnostic results of all candidate potential diseases, those diseases with the highest confidence scores or meeting other preset criteria will be selected as the target for the final diagnosis.

[0080] Specifically, first, all candidate potential diseases are sorted according to the confidence scores given by each diagnostic sub-model and other relevant indicators (such as symptom matching degree, complication risk, etc.). Determine which diseases have the highest possibility of becoming the target potential diseases.

[0081] Then, based on the results of the comprehensive evaluation, select those diseases with the highest confidence and most in line with the actual situation of the patient as the target potential diseases. Optionally, if the possibilities of multiple diseases are similar, they can be considered as possible target potential diseases and corresponding explanations can be provided.

[0082] In one embodiment, query the preliminary potential diseases including the target patient's symptoms from the structured knowledge graph; based on the structured knowledge graph, determine all the patient symptoms of the preliminary potential diseases; if the target patient's symptoms are the same as all the patient symptoms, then use the preliminary potential diseases as candidate potential diseases.

[0083] Among them, the preliminary potential diseases refer to a group of diseases that are preliminarily screened out in the structured knowledge graph based on the symptoms of the target patient. These diseases are identified according to the symptoms shown by the patient in the preliminary query stage and serve as the basis for further detailed analysis. Understandably, when the system inputs the patient's symptoms, it will search the knowledge graph for all diseases related to these symptoms and mark them as preliminary potential diseases.

[0084] All the patient symptoms refer to the set of all possible symptoms associated with each preliminary potential disease. In the knowledge graph, each disease is connected to its typical symptoms, signs, and other relevant information. Therefore, "all the patient symptoms" refers to all the symptoms that a specific disease usually shows, which includes typical symptoms and some less common but still related symptoms.

[0085] Specifically, first, use the target patient's symptoms as query conditions to search for all potential diseases directly or indirectly associated with these symptoms in a pre-constructed general medical knowledge graph. These initially identified diseases are called "preliminary potential diseases" and form the basis for further analysis. Then, for each preliminary potential disease, extract all possible symptom sets associated with this disease from the knowledge graph. This includes typical symptoms and atypical but possible symptoms, forming the "symptoms of all patients" for this disease. Finally, compare the symptom list of the target patient with the symptom list of all patients for each preliminary potential disease. Analyze the similarities and differences between the two to determine which preliminary potential diseases best match the actual situation of the target patient. If the symptoms of the target patient are highly consistent or exactly the same as the symptoms of all patients for a certain preliminary potential disease, then consider this preliminary potential disease as a valid "candidate potential disease".

[0086] Since in the face of complex cases, a single symptom may correspond to multiple different potential diseases. This step can effectively identify all possible disease options and provide detailed symptom comparison results. By querying the preliminary potential diseases related to the target patient's symptoms from the structured knowledge graph and further comparing the symptoms of all patients with these diseases with the symptoms of the target patient, accurate matching of potential diseases can be achieved. This helps to improve the accuracy of diagnosis and reduce the possibility of misdiagnosis or missed diagnosis.

[0087] In one embodiment, for each candidate potential disease, use the corresponding diagnostic sub-model to predict the physical examination data to obtain the confidence score that the target object belongs to the candidate potential disease being targeted; based on the confidence scores that the target object belongs to each candidate potential disease, determine the target potential disease corresponding to the physical examination data.

[0088] Among them, the confidence score refers to a numerical score given by each diagnostic sub-model after analyzing the physical examination data of the target object (i.e., the patient) for a specific candidate potential disease. This score reflects the degree of certainty or probability estimate that the patient belongs to this specific potential disease. For example, the higher the score, the more certain the model is that the target object has this disease; conversely, a lower score indicates a lower likelihood of having the disease or a higher degree of uncertainty.

[0089] Specifically, first, for each candidate potential disease, use its corresponding diagnostic sub-model to analyze the physical examination data of the target object (including symptoms, signs, examination results, etc.). Each diagnostic sub-model will output a confidence score indicating the probability or credibility that the patient has this specific potential disease. Further collect the confidence scores of each candidate potential disease and record them.

[0090] Then, compare the confidence scores of all candidate potential diseases to identify the disease with the highest score. Understandably, the disease with the highest score is considered the target potential disease that best matches the patient's symptoms. If the confidence scores of multiple diseases are very close, further analysis or clinical judgment may be required to determine the final selection. Based on the comparison results, select one or several diseases with the highest confidence scores as the target potential diseases.

[0091] Since the probability of a patient having a certain disease can be quantitatively evaluated by predicting each candidate potential disease using a dedicated diagnostic sub-model and obtaining a confidence score. This method enables the system to more accurately identify the most likely target potential disease, thereby improving the accuracy of the overall diagnosis.

[0092] In one embodiment, feature extraction is performed on the physical examination data through the diagnostic sub-model corresponding to the candidate potential disease to obtain a health feature representation; the health feature representation is converted into a disease probability distribution feature vector; based on the disease probability distribution feature vector, the confidence score of the target object belonging to the candidate potential disease is determined.

[0093] Among them, the health feature representation refers to a set of key features extracted from the physical examination data of the target object (i.e., the patient) that can reflect their health status. Understandably, the original physical examination data is converted into a structured form, which is convenient for subsequent analysis and processing. In short, the health feature representation is a digital abstraction of the patient's health status to facilitate understanding and processing by computer models.

[0094] The disease probability distribution feature vector refers to a mathematical expression form further processed based on the health feature representation, which reflects the probability distribution of the patient having different potential diseases. For example, each element in the disease probability distribution feature vector represents the probability or likelihood of the patient having a specific potential disease. These probability values are usually calculated by the diagnostic sub-model based on the health feature representation, and the sum of the probabilities of all elements is usually 1 (if it is a probability distribution), or it can be an unnormalized score value.

[0095] Specifically, first, according to the multiple candidate potential diseases identified from the structured knowledge graph before, select the dedicated diagnostic sub-model corresponding to each disease. Understandably, these sub-models are trained to handle specific types of diseases. Then use the selected diagnostic sub-model to perform feature extraction on the physical examination data of the target patient. The extracted health feature representation is a digital abstraction of the patient's health status, which is convenient for subsequent calculations.

[0096] Then, the health feature representation is input into the corresponding diagnostic sub-model. Based on its internal algorithm and the knowledge learned during training, the model calculates the probability or score of the patient having each potential disease. The output is a disease probability distribution feature vector, where each element represents the probability or score value of a potential disease. For each candidate potential disease, the confidence score is determined according to its score in the disease probability distribution feature vector. The confidence score reflects the degree of credibility that the model believes the patient has the disease. The higher the score, the greater the likelihood of having the disease.

[0097] Finally, compare the confidence scores of all candidate potential diseases and identify the disease or group of diseases with the highest score as the most likely target potential disease. If the confidence score of a certain disease is significantly higher than that of other diseases, it can be more confidently identified as the target potential disease; if there are multiple groups of diseases with similar scores, further analysis or clinical judgment is required.

[0098] Since the feature extraction of physical examination data is carried out through a specially designed diagnostic sub-model, it can effectively identify the health feature representation related to specific diseases. And converting the health feature representation into a disease probability distribution feature vector enables the system to quantitatively evaluate the probability of the patient having each potential disease. This not only improves the objectivity and scientific nature of the diagnostic results but also reduces the risk of misdiagnosis or missed diagnosis.

[0099] In one embodiment, the rank relationships of multiple candidate potential diseases are obtained from the structured knowledge graph; based on the rank relationships, the upper-level potential diseases, lower-level potential diseases, and isolated potential diseases are determined among the multiple candidate potential diseases; based on the confidence scores of the upper-level potential diseases, lower-level potential diseases, and isolated potential diseases respectively, the target potential disease corresponding to the physical examination data is determined.

[0100] Among them, the rank relationship refers to the hierarchical or subordinate relationship between different potential diseases in the structured knowledge graph. This relationship reflects the possible inclusion, causal, or other relevance between diseases in the medical field. For example, some diseases may be subsets of other more general or basic diseases (such as a specific type of pneumonia relative to general respiratory infections).

[0101] The upper-level potential disease refers to the disease at a higher level in the rank relationship, usually having more extensive symptom manifestations or covering more specific diseases. Such diseases are often more common or basic health problems, and may contain multiple specific subtypes or complications below. For example, in respiratory diseases, "chronic obstructive pulmonary disease" can be regarded as the upper-level potential disease of "chronic bronchitis" and "emphysema".

[0102] Subordinate potential diseases refer to diseases at a lower level in the hierarchy. They are usually more specific or special diseases and often exist as subcategories of a superordinate potential disease. For example, "acute asthma attack" can be regarded as a subordinate potential disease under the broader concept of "asthma".

[0103] Isolated potential diseases are those that do not form a direct hierarchical relationship with other diseases. These diseases exist independently and do not have a clear inclusion or inclusion relationship with other potential diseases. Isolated potential diseases may not be simply classified into a certain category due to their unique etiology, pathological mechanism or clinical manifestations.

[0104] Specifically, we first use the multiple candidate potential diseases identified in the previous step to query the structured knowledge graph to obtain the hierarchical relationships between these diseases, including which diseases are superordinate potential diseases (broader diseases), which are subordinate potential diseases (specific subtypes or complications), and which are isolated potential diseases (not directly related to other diseases).

[0105] Then, based on the hierarchical relationship information provided by the structured knowledge graph, the candidate potential diseases are classified into three categories: superordinate potential diseases (representing a wider range of underlying diseases); subordinate potential diseases (representing specific subtypes or complications); and isolated potential diseases (existing independently without a clear inclusion or being included relationship with other diseases).

[0106] Finally, the confidence scores of all candidate potential diseases and their positions in the hierarchy are comprehensively considered. Optionally, if the confidence score of a subordinate potential disease is significantly higher than that of other diseases and highly matches the patient's specific symptoms, the subordinate potential disease is selected as the target potential disease. If the confidence score of the isolated potential disease is the highest, the isolated potential disease is selected as the target potential disease. In some cases, clinical judgment may be required to determine the final selection, especially when the confidence scores of multiple diseases are similar.

[0107] By obtaining the hierarchical relationships of candidate potential diseases from the structured knowledge graph and classifying them into superordinate potential diseases, subordinate potential diseases, and isolated potential diseases, we can understand the correlation between different diseases in more detail. This method helps to exclude irrelevant diseases and focus on the most likely diseases, thereby improving the accuracy of diagnosis. This is difficult to handle effectively in traditional single disease diagnosis models. By clarifying these relationships, we can better deal with complex medical cases and ensure that important diagnostic clues are not missed.

[0108] In one embodiment, the confidence score of the upper-level potential disease is compared with the first confidence threshold; if the confidence score of the upper-level potential disease exceeds the first confidence threshold, the confidence scores of the lower-level potential diseases that are in the same hierarchical relationship as the upper-level potential disease are adjusted to zero; the confidence scores of the upper-level potential disease and the isolated potential disease are respectively compared with the second confidence threshold; based on the comparison results, the target potential disease corresponding to the physical examination data is determined.

[0109] Among them, the first confidence threshold refers to a preset numerical standard used to determine whether the confidence score of the upper-level potential disease is high enough to consider that the upper-level potential disease is a credible target potential disease. If the confidence score of a certain upper-level potential disease exceeds this threshold, it is considered that the disease has a high possibility of being the correct diagnosis result.

[0110] The same hierarchical relationship means all the diseases in the structured knowledge graph that are at the same level as a specific upper-level potential disease. For example, if "chronic obstructive pulmonary disease" is an upper-level potential disease, then those in the same hierarchical relationship with it may include "chronic bronchitis" and "emphysema". When an upper-level potential disease is confirmed to have a high confidence level, the confidence scores of its lower-level potential diseases at the same level are adjusted to zero to avoid repeated or unnecessary consideration.

[0111] The second confidence threshold refers to another preset numerical standard used to evaluate whether the confidence scores of the upper-level potential disease and the isolated potential disease are high enough to finally determine their possibility as the target potential disease. This threshold is usually used to further screen the diseases after the initial screening to ensure that only those diseases with high confidence scores are selected as the target potential diseases.

[0112] The comparison result refers to the result obtained by comparing the confidence scores of each potential disease with the corresponding confidence threshold. Based on these comparison results, it can be determined which diseases are most likely to be the target potential diseases.

[0113] Specifically, first, the confidence score of each upper-level potential disease is compared with the first confidence threshold. If the confidence score of a certain upper-level potential disease exceeds the first confidence threshold, it is considered that the upper-level potential disease has a high confidence level. The confidence scores of all lower-level potential diseases in the same hierarchical relationship with it are set to zero, and these lower-level potential diseases are no longer considered as target potential diseases.

[0114] Then, the confidence scores of all the remaining upper-level potential diseases and isolated potential diseases are compared with the second confidence threshold. If the confidence score of a certain disease exceeds the second confidence threshold, it is regarded as the final target potential disease.

[0115] Finally, combine all comparison results and select those diseases with scores exceeding the corresponding thresholds as the final target potential diseases. If multiple diseases exceed the threshold, further analysis or clinical judgment needs to be combined to determine the final selection. For example, if the confidence score of an isolated potential disease is significantly higher than that of other diseases, it may be the final target potential disease; if the scores of multiple parent potential diseases are close, further evaluation by a doctor may be required.

[0116] By setting the first confidence threshold and the second confidence threshold and comparing the confidence scores of different types of potential diseases according to these thresholds, the most likely target potential diseases can be screened out more precisely. This method helps reduce the possibility of misdiagnosis or missed diagnosis and improves the accuracy of diagnosis. And when the confidence score of a parent potential disease exceeds the first confidence threshold, the confidence scores of its subordinate potential diseases at the same level are set to zero, which can avoid confusion and incorrect judgment caused by considering too many related subordinate potential diseases.

[0117] In one embodiment, query treatment suggestions related to the target potential disease from the structured knowledge graph; perform risk screening on the treatment suggestions through pre-imported medical standard data; if there are risk items in the treatment suggestions, adjust the risk items in the treatment suggestions based on the medical standard data; send the adjusted treatment suggestions to the target object.

[0118] Among them, treatment suggestions refer to recommended measures on how to treat the disease queried from the structured knowledge graph based on the diagnosis result (i.e., the target potential disease). These include but are not limited to: drug treatment such as the types of drugs recommended, dosage, and usage methods. Non-drug therapies such as physical therapy, psychotherapy, etc. Lifestyle adjustments such as dietary advice, exercise plans, etc. Surgery or other medical procedures.

[0119] Medical standard data refers to a set of standardized guidelines or rules pre-imported into the system, which are formulated based on the latest medical research results, clinical guidelines, and laws and regulations. These data help ensure that the provided treatment suggestions comply with current medical standards and can maximize the safety of patients. Medical standard data includes but is not limited to: medication guidelines, i.e., information on the safe use of drugs, contraindications, side effect management, etc. Treatment process specifications, i.e., standardized treatment paths or processes for different diseases.

[0120] A risky item refers to some potential problems or inappropriateness that may exist in treatment recommendations. If these problems are not taken into account, they may have an adverse impact on the patient's health. Risky items can cover multiple aspects. For example, drug interactions mean that certain drug combinations may cause adverse reactions. Allergic reactions mean that the patient may have a history of allergies to certain ingredients. Inappropriate treatment methods mean that certain treatment methods may not be suitable for patients of a specific age group or with a specific medical history.

[0121] Specifically, first use the imported medical standard data to conduct a comprehensive risk screening of the retrieved treatment recommendations. The screening process should identify any potential risky items, such as drug interactions, allergic reactions, inappropriate treatment methods, etc. If any risky items are found in the treatment recommendations, mark them. For example, mark that a specific drug may have side effects or contraindications, or that certain treatment methods are not applicable to a specific patient group.

[0122] Then, for the marked risky items, make necessary adjustments based on the medical standard data. This includes but is not limited to: replacing high-risk drugs with safer options. Modifying the treatment process to avoid known risks. Adding additional preventive measures or precautions. Integrate all the adjusted information to generate the final version of the treatment recommendations. Ensure that these recommendations not only conform to the actual situation of the patient but also comply with the current medical norms and standards. And send the adjusted and confirmed treatment recommendations to the target object (i.e., the patient). Optionally, achieve information transmission through an electronic medical record system, a mobile application, or other communication methods.

[0123] Since risk screening of treatment recommendations is carried out by using pre-imported medical standard data, potential risky items such as drug interactions, allergic reactions, or treatment methods not suitable for a specific patient group can be identified. This helps to avoid medical accidents caused by ignoring these risks, thereby improving the safety of treatment.

[0124] In one embodiment, refer to Figure 3 , which shows three main modules in the data processing architecture based on a dynamic routing mechanism: a data integration module, an intelligent decision-making module, and a security fence module.

[0125] Among them, the data integration module is responsible for data collection, cleaning, standardization, and integration to ensure data quality. The intelligent decision-making module uses a knowledge graph and machine learning models for disease classification and treatment recommendation output, and provides referral prompts. The security fence module ensures the security and compliance of the output content, and guarantees the reliability of the final output through detection, inspection, filtering, and adjustment. Through the collaborative work of these three modules, the system can provide efficient, accurate, and secure medical decision-making support.

[0126] It should be understood that although the steps in the flowcharts involved in the above embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise clearly stated in this article, there is no strict order limit for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turns with at least a part of the steps or stages in other steps or other steps.

[0127] Based on the same inventive concept, an embodiment of the present application further provides a data processing device based on a dynamic routing mechanism for implementing the data processing method based on a dynamic routing mechanism described above. The implementation solutions provided by this device to solve problems are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the data processing device based on a dynamic routing mechanism provided below can refer to the limitations on the data processing method based on a dynamic routing mechanism in the foregoing text, and will not be elaborated here.

[0128] In an exemplary embodiment, as Figure 4 shown, a data processing device 400 based on a dynamic routing mechanism is provided, including: an extraction module 402, a query module 404, an activation module 406, and a determination module 408, where:

[0129] The extraction module 402 is configured to obtain the physical examination data of the target object and extract the target patient symptoms from the physical examination data;

[0130] The query module 404 is configured to query a plurality of candidate potential diseases structurally associated with the target patient symptoms from a structured knowledge graph, where the structured knowledge graph includes a plurality of potential diseases, patient symptoms, and the membership structure relationship between each potential disease and patient symptom;

[0131] The activation module 406 is configured to activate the respective diagnostic sub-models of the plurality of candidate potential diseases in a diagnostic model;

[0132] The determination module 408 is configured to determine the target potential disease corresponding to the physical examination data among the plurality of candidate potential diseases based on the diagnostic results of the diagnostic sub-models.

[0133] In one embodiment, the query module 404 is configured to query, from the structured knowledge graph, preliminary potential diseases including the target patient's symptoms; determine all patient symptoms of the preliminary potential diseases based on the structured knowledge graph; and if the target patient's symptoms are the same as all patient symptoms, use the preliminary potential diseases as candidate potential diseases.

[0134] In one embodiment, the determination module 408 is configured to, for each candidate potential disease, predict the physical examination data through the diagnostic sub-model corresponding to the candidate potential disease targeted, to obtain the confidence score that the target object belongs to the candidate potential disease targeted; and determine the target potential disease corresponding to the physical examination data based on the confidence scores that the target object belongs to each candidate potential disease.

[0135] In one embodiment, the determination module 408 is configured to extract features from the physical examination data through the diagnostic sub-model corresponding to the candidate potential disease targeted, to obtain a healthy feature representation; convert the healthy feature representation into a disease probability distribution feature vector; and determine the confidence score that the target object belongs to the candidate potential disease targeted according to the disease probability distribution feature vector.

[0136] In one embodiment, the determination module 408 is configured to obtain the rank relationships of multiple candidate potential diseases respectively from the structured knowledge graph; determine the upper-level potential diseases, lower-level potential diseases, and isolated potential diseases among the multiple candidate potential diseases based on the rank relationships; and determine the target potential disease corresponding to the physical examination data based on the confidence scores of the upper-level potential diseases, lower-level potential diseases, and isolated potential diseases respectively.

[0137] In one embodiment, the determination module 408 is configured to compare the confidence score of the upper-level potential disease with a first confidence threshold; if the confidence score of the upper-level potential disease exceeds the first confidence threshold, adjust the confidence score of the lower-level potential disease that has the same rank relationship as the upper-level potential disease to zero; compare the confidence score of the upper-level potential disease and the confidence score of the isolated potential disease with a second confidence threshold respectively; and determine the target potential disease corresponding to the physical examination data based on the comparison results.

[0138] In one embodiment, the data processing device based on the dynamic routing mechanism further includes a suggestion module 410, configured to query treatment suggestions related to the target potential disease from the structured knowledge graph; perform risk screening on the treatment suggestions through the pre-imported medical standard data; if there are risk items in the treatment suggestions, adjust the risk items in the treatment suggestions based on the medical standard data; and send the adjusted treatment suggestions to the target object.

[0139] In another embodiment, as Figure 5 shown Figure 5FIG. 400 is a structural block diagram of a data processing device based on a dynamic routing mechanism in another embodiment, including: an extraction module 402, a query module 404, an activation module 406, a determination module 408, and a recommendation module 410. Among them, the recommendation module 410 is configured to query treatment recommendations related to a target potential disease from a structured knowledge graph; perform risk screening on the treatment recommendations through pre-imported medical specification data; if there are risk items in the treatment recommendations, adjust the risk items in the treatment recommendations based on the medical specification data; and send the adjusted treatment recommendations to the target object.

[0140] Each module in the above data processing device based on a dynamic routing mechanism can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in or independent of a processor in a computer device in the form of hardware, or stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0141] In an exemplary embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 6 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data related to data processing based on a dynamic routing mechanism. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a data processing method based on a dynamic routing mechanism.

[0142] Those skilled in the art can understand that Figure 6 the structure shown in

[0143] In one embodiment, a computer device is further provided, which includes a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0144] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0145] In one embodiment, a computer program product is provided, which includes a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0146] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties. Moreover, the collection, use, and processing of relevant data need to comply with relevant regulations.

[0147] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., without limitation.

[0148] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this application.

[0149] The above embodiments only represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A data processing method based on a dynamic routing mechanism, characterized in that Applied to a diagnostic model, the diagnostic model includes multiple diagnostic sub-models for different potential diseases, and the method includes: Obtain the physical examination data of the target object, and extract the target patient symptoms from the physical examination data; Query multiple candidate potential diseases structurally associated with the target patient symptoms from a structured knowledge graph, where the structured knowledge graph includes multiple potential diseases, patient symptoms, and the membership structure relationship between each potential disease and patient symptom; Activate the diagnostic sub-models corresponding to the multiple candidate potential diseases in the diagnostic model; Based on the diagnostic results of the diagnostic sub-models, determine the target potential disease corresponding to the physical examination data among the multiple candidate potential diseases.

2. The method according to claim 1, wherein The querying of multiple candidate potential diseases structurally associated with the target patient symptoms from the structured knowledge graph includes: Query the preliminary potential diseases including the target patient symptoms from the structured knowledge graph; Based on the structured knowledge graph, determine all the patient symptoms of the preliminary potential diseases; If the target patient symptoms are the same as all the patient symptoms, use the preliminary potential disease as the candidate potential disease.

3. The method according to claim 1, wherein The determining of the target potential disease corresponding to the physical examination data among the multiple candidate potential diseases based on the diagnostic results of the diagnostic sub-models includes: For each candidate potential disease, predict the physical examination data through the diagnostic sub-model corresponding to the candidate potential disease to obtain the confidence score that the target object belongs to the candidate potential disease; Based on the confidence scores that the target object belongs to each candidate potential disease, determine the target potential disease corresponding to the physical examination data.

4. The method according to claim 3, characterized in that The predicting of the physical examination data through the diagnostic sub-model corresponding to each candidate potential disease to obtain the confidence score that the target object belongs to the candidate potential disease includes: Extract features from the physical examination data through the diagnostic sub-model corresponding to the candidate potential disease to obtain a healthy feature representation; Convert the healthy feature representation into a disease probability distribution feature vector; Based on the disease probability distribution feature vector, determine the confidence score that the target object belongs to the candidate potential disease.

5. The method according to claim 3, wherein The determining of the target potential disease corresponding to the physical examination data based on the confidence scores that the target object belongs to each candidate potential disease includes: Obtain the rank relationships of the multiple candidate potential diseases from the structured knowledge graph; Based on the rank relationships, determine the upper-level potential diseases, lower-level potential diseases, and isolated potential diseases among the multiple candidate potential diseases; Based on the confidence scores of the upper-level potential diseases, the lower-level potential diseases, and the isolated potential diseases respectively, determine the target potential disease corresponding to the physical examination data.

6. The method according to claim 5, wherein The determining of the target potential disease corresponding to the physical examination data based on the confidence scores of the upper-level potential diseases, the lower-level potential diseases, and the isolated potential diseases respectively includes: Compare the confidence score of the upper-level potential disease with a first confidence threshold; If the confidence score of the upper-level potential disease exceeds the first confidence threshold, adjust the confidence scores of the lower-level potential diseases that are in the same hierarchical relationship as the upper-level potential disease to zero; Compare the confidence score of the upper-level potential disease and the confidence score of the isolated potential disease with the second confidence threshold respectively; Based on the comparison results, determine the target potential disease corresponding to the physical examination data.

7. The method according to any one of claims 1 to 6, characterized in that The method further includes: Query treatment suggestions related to the target potential disease from the structured knowledge graph; Perform risk screening on the treatment suggestions through pre-imported medical standard data; If there are risk items in the treatment suggestions, adjust the risk items in the treatment suggestions based on the medical standard data; Send the adjusted treatment suggestions to the target object.

8. A data processing device based on a dynamic routing mechanism, characterized in that, The device includes: An extraction module, configured to obtain the physical examination data of the target object and extract the target patient symptoms from the physical examination data; A query module, configured to query a plurality of candidate potential diseases that are structurally associated with the target patient symptoms from the structured knowledge graph, where the structured knowledge graph includes a plurality of potential diseases, patient symptoms, and the subordination structure relationship between each potential disease and patient symptom; An activation module, configured to activate the respective diagnostic sub-models corresponding to the plurality of candidate potential diseases in the diagnostic model; A determination module, configured to determine the target potential disease corresponding to the physical examination data among the plurality of candidate potential diseases based on the diagnostic results of the diagnostic sub-models.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images