Disease name matching method, device and equipment for medical insurance and storage medium

Abstract

The application provides a disease name comparison method, device and equipment for medical insurance and a storage medium. The method comprises the following steps: obtaining disease names to be compared; comparing the disease names to be compared based on a basic positive sample pair and an enhanced negative sample pair to obtain a comparison result; the basic positive sample pair comprises multiple disease names corresponding to the same disease; the enhanced negative sample pair comprises disease names corresponding to different diseases; and the enhanced negative sample pair is obtained by replacing elements in the disease names in the basic positive sample pair. The disease name comparison method, device and equipment for medical insurance and the storage medium provided by the application can accurately determine the relationship between the disease names to be compared based on the basic positive sample pair and the enhanced negative sample pair, and then accurately obtain the comparison result.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, device, and storage medium for comparing disease names for medical insurance purposes. Background Technology

[0002] In the field of medical insurance, it is usually necessary to determine whether the diseases corresponding to two disease names are similar diseases. For example, when doctors write medical records, there may be a problem that the disease names in the description part of the medical record are inconsistent with the disease names in the conclusion part of the medical record. In this case, it is necessary to compare the disease names in the description part of the medical record with the disease names in the conclusion part of the medical record to determine whether there is any omission in the disease names in the conclusion part of the medical record.

[0003] Currently, disease name comparison is mostly performed using character comparison methods, such as comparing the edit distance between disease names to determine if any disease names are missing. However, these methods are prone to misjudging two disease names that have similar characters but actually correspond to different diseases. For example, the disease names "hyperkalemia" and "hyperlipidemia" have similar characters but actually correspond to two completely different diseases, leading to comparison errors. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and storage medium for comparing disease names for medical insurance purposes, in order to address the shortcomings of low accuracy in disease name comparison in the prior art.

[0005] This invention provides a method for comparing disease names for medical insurance purposes, comprising:

[0006] Obtain the name of the disease to be compared;

[0007] Based on the basic positive sample pairs and the enhanced negative sample pairs, the names of the diseases to be compared are compared to obtain the comparison results;

[0008] The basic positive sample pair includes multiple disease names corresponding to the same disease, and the enhanced negative sample pair includes disease names corresponding to different diseases. The enhanced negative sample pair is obtained by replacing elements in the disease names in the basic positive sample pair.

[0009] According to the disease name comparison method for medical insurance provided by the present invention, the enhanced negative sample pairs are determined based on the following steps:

[0010] The disease names in the basic positive sample pairs are replaced with elements to obtain candidate negative sample pairs;

[0011] If the disease name of the candidate negative sample pair is the same as the disease name in any base positive sample pair, the candidate negative sample pair is used as the enhanced negative sample pair.

[0012] According to the present invention, a disease name comparison method for medical insurance is provided, wherein the comparison of the disease names to be compared is based on basic positive sample pairs and enhanced negative sample pairs to obtain comparison results, including:

[0013] Based on the base negative sample pairs, the base positive sample pairs, and the enhanced negative sample pairs, the names of the diseases to be compared are compared to obtain the comparison results;

[0014] The base positive sample pairs and the base negative sample pairs are determined based on the medical record disease names in the medical record text and / or the standard disease names in the disease system.

[0015] According to a disease name comparison method for medical insurance provided by the present invention, the comparison of the disease names to be compared is performed based on the basic negative sample pairs, the basic positive sample pairs, and the enhanced negative sample pairs to obtain the comparison results, including:

[0016] Based on the disease name comparison model, the disease names to be compared are compared to obtain the comparison results;

[0017] The disease name comparison model is trained sequentially based on a first type of sample, a second type of sample, and a third type of sample. The first type of sample includes basic positive sample pairs and basic negative sample pairs determined based on the disease name in the medical record. The second type of sample includes enhanced negative sample pairs and basic positive sample pairs determined based on the disease name in the medical record. The third type of sample includes basic positive sample pairs and basic negative sample pairs determined based on the standard disease name.

[0018] According to a disease name comparison method for medical insurance provided by the present invention, the basic positive sample pairs and the basic negative sample pairs are determined based on the following steps:

[0019] Determine the hierarchy of each standard disease name within the disease system;

[0020] Based on the hierarchical relationship between the standard disease names, positive and negative examples are divided into positive and negative examples for each standard disease name, resulting in the basic positive sample pairs and the basic negative sample pairs.

[0021] According to the present invention, a method for comparing disease names for medical insurance, wherein the method involves dividing each standard disease name into positive and negative examples based on the hierarchical relationship between the standard disease names to obtain the basic positive sample pairs and the basic negative sample pairs, includes:

[0022] If one of the two standard disease names is a higher-level standard disease name than the other, the two standard disease names are assigned to the basic positive sample pair.

[0023] If the two standard disease names are at the same level, the two standard disease names are assigned to the base negative sample pair.

[0024] According to the present invention, a disease name comparison method for medical insurance is provided, wherein the comparison of the disease names to be compared is based on basic positive sample pairs and enhanced negative sample pairs to obtain comparison results, including:

[0025] Based on the encoding parameters, the names of each disease to be compared are encoded to obtain the encoding features of each disease name to be compared.

[0026] Based on the interaction parameters, the coding features of each disease name to be compared are interacted to obtain the relationship between each disease name to be compared.

[0027] Based on the classification parameters and the relationship between the disease names to be compared, the comparison results between the disease names to be compared are determined.

[0028] The encoding parameters, the interaction parameters, and the classification parameters are determined based on the base positive sample pairs and the enhanced negative sample pairs.

[0029] According to the present invention, a method for comparing disease names for medical insurance purposes includes encoding each disease name to be compared based on encoding parameters to obtain the encoding features of each disease name to be compared, comprising:

[0030] Disease element analysis is performed on each disease name to be compared to obtain the elements of each disease name to be compared.

[0031] Based on the encoding parameters, the elements of each disease name to be compared are encoded to obtain the encoding features of each disease name to be compared.

[0032] The present invention also provides a disease name comparison device for medical insurance, comprising:

[0033] The acquisition unit is used to acquire the name of the disease to be compared.

[0034] The comparison unit is used to compare the disease names to be compared based on the basic positive sample pairs and the enhanced negative sample pairs, and obtain the comparison results;

[0035] The basic positive sample pair includes multiple disease names corresponding to the same disease, and the enhanced negative sample pair includes disease names corresponding to different diseases. The enhanced negative sample pair is obtained by replacing elements in the disease names in the basic positive sample pair.

[0036] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the disease name comparison method for medical insurance as described above.

[0037] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the disease name comparison method for medical insurance as described above.

[0038] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the disease name comparison method for medical insurance as described above.

[0039] The disease name comparison method, apparatus, equipment, and storage medium provided by this invention for medical insurance can accurately determine the relationship between disease names to be compared when comparing disease names based on basic positive sample pairs and enhanced negative sample pairs, because basic positive sample pairs can characterize the inherent rules of overall names between disease names of the same disease, and enhanced negative sample pairs can characterize the inherent rules of detailed elements between disease names of different diseases. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0041] Figure 1 This is a flowchart illustrating the disease name comparison method for medical insurance provided by the present invention.

[0042] Figure 2 This is a flowchart illustrating the enhanced negative sample pair determination method provided by the present invention;

[0043] Figure 3 This is a flowchart illustrating the method for determining the basic sample pair provided by the present invention;

[0044] Figure 4 This is a flowchart illustrating the implementation of step 320 in the method for determining the basic sample pair provided by the present invention.

[0045] Figure 5 This is a flowchart illustrating an implementation of step 120 in the disease name comparison method for medical insurance provided by the present invention.

[0046] Figure 6This is a schematic diagram of the structure of the disease name comparison model provided by the present invention;

[0047] Figure 7 This is a flowchart illustrating an implementation of step 121 in the disease name comparison method for medical insurance provided by the present invention.

[0048] Figure 8 This is a schematic diagram of the disease name comparison device for medical insurance provided by the present invention;

[0049] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0051] Currently, disease name comparison is mostly performed using character comparison methods, such as comparing the edit distance between disease names to determine if any disease names are missing. However, these methods are prone to misjudging two disease names that have similar characters but actually correspond to different diseases. For example, the disease names "hyperkalemia" and "hyperlipidemia" have similar characters but actually correspond to two completely different diseases, leading to comparison errors.

[0052] In addition, character-level detection based on knowledge bases is used to compare disease names. However, due to the non-standard elements of various disease names, the accuracy of character-level detection is low. Furthermore, there are methods that use text matching for disease name comparison, but this method only encodes at the character level, and it is also prone to misclassification for two disease names with similar characters but corresponding to different diseases.

[0053] In response, this invention provides a method for comparing disease names for use in medical insurance. Figure 1 This is a flowchart illustrating the disease name comparison method for medical insurance provided by the present invention, as shown below. Figure 1 As shown, the method includes the following steps:

[0054] Step 110: Obtain the name of the disease to be compared.

[0055] Here, the disease names to be compared are those that need to be compared for similarity in disease categories. For example, "sequelae of poliomyelitis" and "polio" are the same disease; "chronic bronchitis" and "COPD" are different diseases.

[0056] It should be noted that the disease names to be compared here are multiple disease names, that is, there can be two disease names or more than two disease names. This embodiment of the invention does not make a specific limitation in this regard.

[0057] Step 120: Based on the basic positive sample pairs and the enhanced negative sample pairs, compare the disease names to be compared to obtain the comparison results;

[0058] The basic positive sample pair includes multiple disease names corresponding to the same disease, and the enhanced negative sample pair includes disease names corresponding to different diseases. The enhanced negative sample pair is obtained by replacing elements in the disease names in the basic positive sample pair.

[0059] Specifically, the basic positive sample pair includes multiple disease names corresponding to the same disease. For example, "sequelae of poliomyelitis" and "polio" are disease names corresponding to the same disease, thus {sequelae of poliomyelitis, polio} can constitute a basic positive sample pair. The basic positive sample pair can be determined from the disease names in the medical record text, or it can be determined from the standard disease names in the disease system, such as those based on the standard disease names in the disease system obtained through the International Classification of Diseases (ICD). This embodiment of the invention does not specifically limit this.

[0060] Enhanced negative sample pairs consist of disease names corresponding to different diseases, obtained by replacing elements in the disease names of the base positive sample pairs. The elements corresponding to the disease names in the base positive sample pairs describe the corresponding diseases and can include clinical manifestations, location, etiology, pathology, etc., which can be obtained by performing disease parsing on the disease names. For example, after parsing the disease name "pneumonia" in the base positive sample pairs, its corresponding elements are {lung infection, pneumonia}. If "lung" is replaced with "stomach", the corresponding disease name element is {stomach infection, gastritis}. That is, the disease name corresponding to this element is "gastritis," which corresponds to two different diseases, namely {pneumonia, gastritis}, forming an enhanced negative sample pair.

[0061] Furthermore, basic positive sample pairs can characterize the inherent patterns among disease names of the same disease. Based on these patterns, the disease names to be compared can be analyzed to determine if they conform to these patterns. If so, it indicates a high probability that the disease names to be compared are the same disease. Similarly, enhanced negative sample pairs can characterize the inherent patterns among disease names of different diseases. Based on these patterns, the disease names to be compared can be analyzed to determine if they conform to these patterns. If so, it indicates a high probability that the disease names to be compared are different diseases.

[0062] Furthermore, enhanced negative sample pairs are obtained by replacing elements of disease names in basic positive sample pairs. Thus, enhanced negative sample pairs not only contain overall pattern information of disease names, but also detailed pattern information of the corresponding elements of disease names. This allows for a more accurate characterization of the inherent patterns between different disease names. Consequently, based on enhanced negative sample pairs, it is possible to accurately identify whether the disease names to be compared correspond to different diseases, and thus obtain the corresponding comparison results.

[0063] Optionally, embodiments of the present invention can train a disease name comparison model based on basic positive sample pairs and enhanced negative sample pairs. This allows the trained disease name comparison model to learn the inherent patterns between disease names of the same disease from basic positive sample pairs, and also to learn the inherent patterns between disease names of different diseases from enhanced negative sample pairs. In this way, the comparison results of the disease names to be compared can be accurately obtained based on the disease name comparison model.

[0064] The disease name comparison method provided in this invention can accurately determine the relationship between the disease names to be compared when comparing disease names based on the basic positive sample pairs and the enhanced negative sample pairs, since the basic positive sample pairs can characterize the inherent rules of the overall names of the same disease and the enhanced negative sample pairs can characterize the inherent rules of the detailed elements of the disease names of different diseases. Thus, the comparison results can be accurately obtained.

[0065] Based on the above embodiments, Figure 2 This is a flowchart illustrating the enhanced negative sample pair determination method provided by the present invention, as shown below. Figure 2 As shown, the methods for determining enhanced negative sample pairs include:

[0066] Step 210: Replace the disease names in the basic positive sample pairs with relevant elements to obtain candidate negative sample pairs;

[0067] Step 220: If the disease name of a candidate negative sample pair is the same as the disease name in any basic positive sample pair, the candidate negative sample pair is used as an enhanced negative sample pair.

[0068] Specifically, the elements of the disease name in the basic positive sample pair are used to describe the corresponding disease. These elements may include clinical manifestations, location, etiology, pathology, and a leading term. The leading term can be any one of the clinical manifestations, location, etiology, and pathology, or other elements may be selected as the leading term according to the actual situation. This embodiment of the invention does not impose specific limitations on this. For example, the elements of the disease name in the basic positive sample pair can be represented as {lung infection, pneumonia}, where “lung infection” is the location and “pneumonia” is the leading term.

[0069] Furthermore, the elements of a disease name can be obtained by parsing the disease name. For example, the disease name can be input into a named entity recognition model, which will then perform named entity recognition to obtain the elements of the disease name. The named entity recognition model can be based on a BERT+LSTM+CRF model, or it can be based on other models; this embodiment of the invention does not specifically limit its application.

[0070] After determining the elements of the disease names in the baseline positive sample pairs, any one or more elements can be replaced to obtain candidate negative sample pairs. However, since candidate negative sample pairs are obtained by randomly replacing the elements of the disease names in the baseline positive sample pairs, the disease names determined based on the elements in the candidate negative sample pairs may not have corresponding clinical diseases, while the disease names in the baseline positive sample pairs are usually obtained from medical record texts or medical record systems, and therefore the disease names in the baseline positive sample pairs correspond to clinical diseases.

[0071] In this embodiment of the invention, after obtaining a candidate negative sample pair, if the disease name of the candidate negative sample pair is the same as the disease name in any basic positive sample pair, it indicates that the disease name of the candidate negative sample pair corresponds to a clinical disease, that is, the candidate negative sample pair is real and reliable, and thus the candidate negative sample pair can be used as an enhanced negative sample pair.

[0072] For example, in a basic positive sample pair, the disease name element is {lung infection, pneumonia}. If “lung” is replaced with “stomach”, the corresponding disease name element becomes {stomach infection, gastritis}. That is, the disease name corresponding to this element is “gastritis”, and “gastritis” corresponds to a clinical disease. Therefore, {pneumonia, gastritis} can constitute an enhanced negative sample pair.

[0073] Therefore, in the embodiments of the present invention, when the disease name of the candidate negative sample pair is the same as the disease name in any basic positive sample pair, the candidate negative sample pair is used as the enhanced negative sample pair, ensuring that the disease name of the enhanced negative sample pair corresponds to the corresponding clinical disease, thereby ensuring the authenticity and reliability of the enhanced negative sample pair data, and thus accurately obtaining the comparison results based on the authentic and reliable enhanced negative sample pair.

[0074] Based on any of the above embodiments, step 120 includes:

[0075] Based on the basic negative sample pairs, the basic positive sample pairs, and the enhanced negative sample pairs, the disease names to be compared are compared to obtain the comparison results;

[0076] The base positive sample pairs and base negative sample pairs are determined based on the disease names in the medical record text and / or the standard disease names in the disease system.

[0077] Specifically, the basic positive sample pairs include multiple disease names corresponding to the same disease, while the basic negative sample pairs include disease names corresponding to different diseases. The difference between basic and enhanced negative sample pairs is that basic negative sample pairs are determined based on the disease names in the medical record text and / or the standard disease names in the disease system, while enhanced negative sample pairs are obtained by replacing elements in the disease names of the basic positive sample pairs. In other words, basic negative sample pairs can represent the overall pattern information of different disease names, while enhanced negative sample pairs can represent both the overall pattern information of different disease names and the detailed pattern information of the elements corresponding to different disease names.

[0078] Since basic positive sample pairs can characterize the inherent patterns among disease names of the same disease, the comparison of disease names can be analyzed based on these patterns to determine whether they conform to the inherent patterns among disease names of the same disease. If so, it indicates a high probability that the comparison of disease names is the same disease. Basic negative sample pairs can characterize the inherent patterns among disease names of different diseases. Therefore, the comparison of disease names can be analyzed based on these inherent patterns to determine whether they conform to the inherent patterns among disease names of different diseases. If so, it indicates a high probability that the comparison of disease names is different disease. Enhanced negative sample pairs can characterize the overall pattern information of different disease names as well as the detailed pattern information of the corresponding elements of different disease names. Therefore, the comparison of disease names can be analyzed from both overall and detailed pattern information perspectives to determine whether they conform to the inherent patterns among disease names of different diseases. If so, it indicates a high probability that the comparison of disease names is different disease.

[0079] Optionally, embodiments of the present invention can train a disease name comparison model based on basic positive sample pairs, basic negative sample pairs, and enhanced negative sample pairs. This allows the trained model to learn the inherent patterns between disease names of the same disease from basic positive sample pairs, the inherent patterns between disease names of different diseases from basic negative sample pairs, and the overall and detailed patterns between disease names of different diseases from enhanced negative sample pairs. This enables the model to accurately obtain the comparison results of the disease names to be compared. Specifically, when training based on basic positive sample pairs, basic negative sample pairs, and enhanced negative sample pairs, a learning-by-course approach can be used, i.e., training from easy to difficult. For example, the initial model can be trained first using basic positive and negative sample pairs, enabling it to learn the inherent patterns between the same and different diseases at the overall level. Then, enhanced negative sample pairs are used to train the initial model, allowing the resulting disease name comparison model to learn the inherent patterns between different diseases at both the overall and detailed levels, thereby accurately obtaining the comparison results.

[0080] As an optional embodiment, the basic positive sample pairs and basic negative sample pairs can be determined based on the disease names in the medical record text. For example, the disease names and their corresponding diagnostic codes (such as ICD codes) can be obtained from the medical record text, and the standard disease names corresponding to the diagnostic codes can be obtained based on the mapping relationship between the diagnostic codes and the standard disease names. Then, different disease names in the medical records are used as basic negative sample pairs, and the disease names in the medical records and their corresponding standard disease names are used as basic positive sample pairs.

[0081] As an optional implementation, the basic positive sample pairs and basic negative sample pairs can be determined based on the standard disease names in the disease system. For example, if the disease system (such as ICD-10) includes standard disease names at different levels, the standard disease names at the same level can be used as the basic negative sample pairs, and any standard disease name and the standard disease name at the next higher level can be used as the basic positive sample pairs.

[0082] As an optional embodiment, the basic positive sample pairs and basic negative sample pairs can be determined based on the medical record disease names in the medical record text and the standard disease names in the disease system. For example, different medical record disease names in the medical record text can be used as basic negative sample pairs, medical record disease names and their corresponding standard disease names can be used as basic positive sample pairs, and standard disease names at the same level can be used as basic negative sample pairs, and any standard disease name and the standard disease name at the next higher level can be used as basic positive sample pairs.

[0083] Based on any of the above embodiments, in step 120, the disease names to be compared are compared based on the base negative sample pairs, the base positive sample pairs, and the enhanced negative sample pairs to obtain the comparison results, including:

[0084] Based on the disease name comparison model, the disease names to be compared are compared to obtain the comparison results;

[0085] The disease name comparison model is trained on the first type of samples, the second type of samples, and the third type of samples in sequence. The first type of samples includes basic positive sample pairs and basic negative sample pairs determined based on the disease names in the medical records. The second type of samples includes enhanced negative sample pairs and basic positive sample pairs determined based on the disease names in the medical records. The third type of samples includes basic positive sample pairs and basic negative sample pairs determined based on the standard disease names.

[0086] Specifically, the training phase of the disease name comparison model includes three stages: the first stage is based on the first type of samples, the second stage is based on the first stage and the second type of samples, and the third stage is based on the second stage and the third type of samples.

[0087] Among them, the basic positive sample pairs and basic negative sample pairs in the first type of samples are determined based on the disease names in the medical records. Therefore, the first type of samples has the largest number and the lowest learning difficulty. The basic positive sample pairs in the second type of samples are determined based on the disease names in the medical records, but their negative sample pairs are enhanced negative sample pairs. That is, the second type of samples has the same positive sample pairs as the first type of samples, but the negative sample pairs are different. Since the enhanced negative sample pairs are obtained by replacing elements of the disease names in the basic positive sample pairs, the second type of samples is more difficult to learn than the first type of samples. The basic positive sample pairs and basic negative sample pairs in the third type of samples are determined based on the standard disease names. The standard disease names are more standardized and rigorous than the disease names in the medical records, and involve more professional domain knowledge. Therefore, the third type of samples has the highest learning difficulty.

[0088] In this embodiment of the invention, the disease name comparison model is trained sequentially using first-class samples, second-class samples, and third-class samples. This allows the model to learn and train gradually from easy to difficult during the training process, thereby obtaining a disease name comparison model with better performance. This model can then accurately compare the disease names to be compared and obtain the comparison results.

[0089] Based on any of the above embodiments Figure 3 This is a flowchart illustrating the method for determining the basic sample pair provided by this invention, as shown below. Figure 3 As shown, the base positive sample pairs and base negative sample pairs are determined based on the following steps:

[0090] Step 310: Determine the hierarchy of each standard disease name within the disease system;

[0091] Step 320: Based on the hierarchical relationship between the standard disease names, divide each standard disease name into positive and negative examples to obtain basic positive sample pairs and basic negative sample pairs.

[0092] Specifically, the hierarchy of each standard disease name within the disease system is used to characterize its hierarchical position within the system. This hierarchy can be determined based on the coding of each standard disease name. The hierarchical relationship between standard disease names is used to characterize their subordinate relationships. Taking the ICD-10 disease system as an example, a standard disease name with a 3-digit ICD code is at the next higher level than a standard disease name with a 4-digit code.

[0093] Because the disease system considers the relationship between the corresponding diseases of each standard disease name (such as the same disease or different diseases) when classifying the standard disease names into hierarchical levels, the relationship between the corresponding diseases of each standard disease name can be known based on the hierarchical relationship between the standard disease names. Then, based on the relationship between the corresponding diseases, the standard disease names can be divided into positive and negative examples to obtain basic positive sample pairs and basic negative sample pairs.

[0094] It should be noted that the disease system is quite extensive, covering almost all existing clinical diseases. Furthermore, the hierarchical classification of standard disease names within the system is strictly based on the corresponding disease type. This means that when classifying standard disease names into hierarchical levels, the classification of names with similar characters is taken into account. For example, "hyperkalemia" and "hyperlipidemia" actually correspond to different diseases. Because medical record names are written by doctors based on experience, they cannot distinguish between these two similar-sounding names, leading to the mistaken assumption that the two disease names correspond to the same disease. The disease system, however, strictly classifies such similar-sounding disease names into hierarchical levels, thereby accurately determining the baseline positive and negative sample pairs based on the disease system to obtain accurate comparison results.

[0095] Based on any of the above embodiments Figure 4 This is a flowchart illustrating the implementation of step 320 in the method for determining the basic sample pair provided by the present invention, as shown below. Figure 4 As shown, step 320 includes:

[0096] Step 321: If one of the two standard disease names is a higher-level standard disease name than the other, classify the two standard disease names into a basic positive sample pair.

[0097] Step 322: If two standard disease names are at the same level, classify the two standard disease names into the base negative sample pair.

[0098] Specifically, the hierarchical relationship between standard disease names is used to characterize the subordinate relationship between standard disease names, and then based on the subordinate relationship, it can be determined whether the diseases corresponding to each standard disease name are the same disease or different diseases.

[0099] When one of the two standard disease names is a higher-level standard disease name than the other, according to the hierarchical classification rules of the disease system, the two standard disease names correspond to the same disease, and are therefore assigned to the basic positive sample pair. When the two standard disease names are at the same level, according to the hierarchical classification rules of the disease system, the two standard disease names correspond to different diseases, and are therefore assigned to the basic negative sample pair.

[0100] Taking the ICD-10 disease system as an example, the standard disease name 1 with a 3-digit ICD code is one level above the standard disease name 2 and standard disease name 3 with 4-digit codes. Therefore, standard disease name 1 and standard disease name 2, and standard disease name 1 and standard disease name 3 are basic positive sample pairs. However, after performing Cartesian calculations on standard disease name 2 and standard disease name 3, standard disease name 2 and standard disease name 3 are basic negative samples. The 3-digit ICD code also corresponds to standard disease name 4 and standard disease name 5. Therefore, standard disease name 1, standard disease name 4, and standard disease name 5 are also basic negative sample pairs.

[0101] Based on any of the above embodiments Figure 5 This is a flowchart illustrating an implementation of step 120 in the disease name comparison method for medical insurance provided by the present invention, as shown below. Figure 5 As shown, step 120 includes:

[0102] Step 121: Based on the encoding parameters, encode each disease name to be compared to obtain the encoding features of each disease name to be compared.

[0103] Step 122: Based on the interaction parameters, interact with the encoding features of each disease name to be compared to obtain the relationship between each disease name to be compared.

[0104] Step 123: Based on the classification parameters and the relationship between the disease names to be compared, determine the comparison results between the disease names to be compared.

[0105] The encoding parameters, interaction parameters, and classification parameters are determined based on the base positive sample pairs and the augmented negative sample pairs.

[0106] Specifically, the encoding parameters are used to encode each disease name to be compared, resulting in the encoding features of each disease name to be compared. There can be multiple disease names to be compared. After obtaining the encoding features, the interaction parameters are used to interact with each encoding feature to determine the relationship between the disease names to be compared, i.e., the relationship between each pair of disease names. Then, based on the classification parameters and the relationship between each disease name to be identified, the comparison results between each pair of disease names can be determined, i.e., the comparison results between all the disease names to be compared.

[0107] Traditional methods can only compare two disease names at a time. If multiple disease names need to be compared, they must be compared in batches of two, requiring repeated encoding of multiple disease names and reducing comparison efficiency. However, the embodiments of the present invention can encode multiple disease names at once and complete the comparison of multiple disease names in one go, avoiding the problem of low comparison efficiency caused by repeated encoding of disease names.

[0108] Optionally, embodiments of the present invention can train a disease name comparison model based on basic positive sample pairs and enhanced negative sample pairs. Figure 6 This is a schematic diagram of the structure of the disease name comparison model provided by the present invention, as shown below. Figure 6 As shown, the disease name matching model includes an encoding layer, an interaction layer, and a classification layer. Disease names A to F are input into the encoding layer, which encodes them based on the encoding parameters to obtain the encoded features E of disease names A to F. A ~E F Next, the encoded feature E A ~E F The input is fed into the interaction layer, where the interaction parameters are processed by a high-dimensional attention mechanism to encode the feature E. A ~E F By performing pairwise interactions, we can obtain the relationships between any disease name and all disease names, such as R. AA R represents the relationship between disease name A and disease name A'. AB This indicates the relationship between disease name A and disease name B, etc., where R AA ~R AF The corresponding values ​​are two-dimensional matrices. Then, the relationships between the disease names are input into the classification layer. Based on the classification parameters and the relationships between the disease names to be compared, the classification layer determines the comparison results between the disease names, such as S. AF S represents the classification score between disease name A and disease name F, where S AA ~S AF The corresponding value is a two-dimensional matrix, and based on this classification score, the comparison result between disease name A and disease name F can be determined. Specifically, the classification score between disease name j and disease name k...j，k It can be determined based on the following formula:

[0109]

[0110] Among them, score j，k The value range of is (0,1). Flat represents tiling the matrix into a 1-dimensional vector (dimension P*P). MLP stands for Multilayer Perceptron, which maps the dimension from P*P to e. W and b are both trainable parameters, and S... j，k This represents the relationship between disease name j and disease name k.

[0111] Based on any of the above embodiments Figure 7 This is a flowchart illustrating an implementation of step 121 in the disease name comparison method for medical insurance provided by the present invention, as shown below. Figure 7 As shown, step 121 includes:

[0112] Step 1211: Analyze the disease elements of each disease name to be compared to obtain the elements of each disease name to be compared.

[0113] Step 1212: Based on the encoding parameters, encode the elements of each disease name to be compared to obtain the encoding features of each disease name to be compared.

[0114] Specifically, the elements of each disease name to be compared are used to describe the corresponding disease, which may include clinical manifestations, location, etiology, pathology, and key terms, etc., and can be obtained by parsing the disease elements of each disease name. After obtaining the elements of each disease name, the elements are encoded, so that each disease name can be encoded from the aspects of elements that fit the nature of the disease itself, and the encoded features that can accurately represent the nature of the disease corresponding to each disease name are obtained.

[0115] Optionally, when encoding each disease name, the disease name is first parsed using a disease name parsing tool to obtain the elements of each disease name. Each element is then mapped to a vector through a randomly initialized word vector layer, and then input into a BLSTM (bidirectional LSTM) layer for encoding. The last character of the bidirectional sequence is taken and encoded and concatenated (i.e., [F; B], where F is the forward sequence result and B is the reverse sequence result) to obtain the encoding features of each element. Then, the GLU layer is input for processing through a gating mechanism.

[0116] Based on any of the above embodiments, the present invention also provides a method for comparing disease names for medical insurance purposes, the method comprising:

[0117] Obtain multiple disease names to be compared, and input each disease name to be compared into the disease name comparison model to obtain the comparison results output by the disease name comparison model.

[0118] The disease name comparison model comprises an encoding layer, an interaction layer, and a classification layer. The encoding layer encodes the disease names to be compared based on encoding parameters, obtaining the encoded features of each disease name. The interaction layer interacts with the encoded features of each disease name based on interaction parameters, obtaining the relationships between the disease names. The classification layer determines the comparison results between the disease names based on classification parameters and the relationships between them.

[0119] Furthermore, the disease name matching model is trained based on a basic dataset, an augmented dataset, and a discrimination dataset. The model is trained sequentially using a course-based learning approach (from easy to difficult). Specifically, the initial model of the disease name matching model is first trained using the basic dataset to obtain the first model; then the first model is trained using the augmented dataset to obtain the second model; and finally, the second model is trained using the discrimination dataset to obtain the disease name matching model.

[0120] The base dataset is determined based on disease names in medical record texts. Specifically, different disease names from different medical records are used as negative sample pairs in the base dataset, and the disease name from the medical record and its corresponding standard disease name are used as positive sample pairs. The augmented dataset uses positive sample pairs from the base dataset as positive sample pairs, and then replaces the disease names in the positive sample pairs from the base dataset with relevant elements to obtain negative sample pairs. The discrimination dataset is determined based on standard disease names in a disease system (such as ICD-10). Specifically, standard disease names at the same level are used as negative sample pairs in the discrimination dataset, and any standard disease name and its next-level standard disease name are used as positive sample pairs in the discrimination dataset.

[0121] The following describes the disease name comparison device for medical insurance provided by the present invention. The disease name comparison device for medical insurance described below and the disease name comparison method for medical insurance described above can be referred to in correspondence.

[0122] Based on any of the above embodiments Figure 8 This is a schematic diagram of the disease name comparison device for medical insurance provided by the present invention, as shown below. Figure 8 As shown, the device includes:

[0123] The acquisition unit 810 is used to acquire the name of the disease to be compared;

[0124] The comparison unit 820 is used to compare the disease names to be compared based on the basic positive sample pairs and the enhanced negative sample pairs, and obtain the comparison results;

[0125] The basic positive sample pair includes multiple disease names corresponding to the same disease, and the enhanced negative sample pair includes disease names corresponding to different diseases. The enhanced negative sample pair is obtained by replacing elements in the disease names in the basic positive sample pair.

[0126] Based on any of the above embodiments, the device further includes:

[0127] An element replacement unit is used to replace the disease names in the basic positive sample pairs to obtain candidate negative sample pairs.

[0128] An enhanced negative sample determination unit is used to identify the candidate negative sample pair as the enhanced negative sample pair when the disease name of the candidate negative sample pair is the same as the disease name in any basic positive sample pair.

[0129] Based on any of the above embodiments, the comparison unit 820 is used for:

[0130] Based on the base negative sample pairs, the base positive sample pairs, and the enhanced negative sample pairs, the names of the diseases to be compared are compared to obtain the comparison results;

[0131] The base positive sample pairs and the base negative sample pairs are determined based on the medical record disease names in the medical record text and / or the standard disease names in the disease system.

[0132] Based on any of the above embodiments, the comparison unit 820 is used for:

[0133] Based on the disease name comparison model, the disease names to be compared are compared to obtain the comparison results;

[0134] The disease name comparison model is trained sequentially based on a first type of sample, a second type of sample, and a third type of sample. The first type of sample includes basic positive sample pairs and basic negative sample pairs determined based on the disease name in the medical record. The second type of sample includes enhanced negative sample pairs and basic positive sample pairs determined based on the disease name in the medical record. The third type of sample includes basic positive sample pairs and basic negative sample pairs determined based on the standard disease name.

[0135] Based on any of the above embodiments, the device further includes:

[0136] A hierarchy determination unit is used to determine the hierarchy of each standard disease name in the disease system.

[0137] The positive and negative sample partitioning unit is used to partition each standard disease name into positive and negative samples based on the hierarchical relationship between each standard disease name, so as to obtain the basic positive sample pair and the basic negative sample pair.

[0138] Based on any of the above embodiments, the positive and negative example division unit includes:

[0139] The first determining unit is used to classify the two standard disease names into the basic positive sample pair when one of the two standard disease names is a standard disease name at the next higher level than the other standard disease name.

[0140] The second determining unit is used to classify the two standard disease names into the basic negative sample pair when the two standard disease names are at the same level.

[0141] Based on any of the above embodiments, the comparison unit 820 includes:

[0142] The coding unit is used to encode each disease name to be compared based on the coding parameters, so as to obtain the coding features of each disease name to be compared.

[0143] The interaction unit is used to interact with the encoded features of each disease name to be compared based on the interaction parameters, so as to obtain the relationship between each disease name to be compared.

[0144] The classification unit is used to determine the comparison results between the disease names to be compared based on the classification parameters and the relationship between the disease names to be compared.

[0145] The encoding parameters, the interaction parameters, and the classification parameters are determined based on the base positive sample pairs and the enhanced negative sample pairs.

[0146] Based on any of the above embodiments, the encoding unit includes:

[0147] The element parsing unit is used to parse the disease elements of each disease name to be compared, and obtain the elements of each disease name to be compared.

[0148] The element encoding unit is used to encode the elements of each disease name to be compared based on the encoding parameters, so as to obtain the encoding features of each disease name to be compared.

[0149] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 9As shown, the electronic device may include a processor 910, a memory 920, a communications interface 930, and a communications bus 940. The processor 910, memory 920, and communications interface 930 communicate with each other via the communications bus 940. The processor 910 can call logical instructions in the memory 920 to execute a disease name comparison method for medical insurance. This method includes: obtaining the disease names to be compared; comparing the disease names to be compared based on basic positive sample pairs and enhanced negative sample pairs to obtain comparison results; the basic positive sample pairs include multiple disease names corresponding to the same disease, and the enhanced negative sample pairs include disease names corresponding to different diseases, wherein the enhanced negative sample pairs are obtained by replacing elements in the disease names of the basic positive sample pairs.

[0150] Furthermore, the logical instructions in the aforementioned memory 920 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0151] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the disease name comparison method for medical insurance provided by the above methods, the method comprising: obtaining disease names to be compared; comparing the disease names to be compared based on basic positive sample pairs and enhanced negative sample pairs to obtain comparison results; wherein the basic positive sample pairs include multiple disease names corresponding to the same disease, the enhanced negative sample pairs include disease names corresponding to different diseases, and the enhanced negative sample pairs are obtained by replacing elements of the disease names in the basic positive sample pairs.

[0152] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When executed by a processor, the computer program is implemented to perform the above-described disease name comparison methods for medical insurance. The method includes: obtaining disease names to be compared; comparing the disease names to be compared based on basic positive sample pairs and enhanced negative sample pairs to obtain comparison results; wherein the basic positive sample pairs include multiple disease names corresponding to the same disease, and the enhanced negative sample pairs include disease names corresponding to different diseases, and the enhanced negative sample pairs are obtained by replacing elements of the disease names in the basic positive sample pairs.

[0153] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0154] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for comparing disease names for medical insurance purposes, characterized in that, include: Obtain the name of the disease to be compared; The disease name comparison model, trained based on basic positive sample pairs and enhanced negative sample pairs, is used to compare the disease names to be compared and obtain the comparison results. The basic positive sample pair includes multiple disease names corresponding to the same disease, and the enhanced negative sample pair includes disease names corresponding to different diseases. The enhanced negative sample pair is obtained by replacing elements in the disease names in the basic positive sample pair. The enhanced negative sample pairs are determined based on the following steps: Disease parsing is performed on each disease name in the basic positive sample pair to obtain the elements of each disease name; the elements are used to describe the corresponding disease, and the elements include clinical manifestations, location, etiology and pathology; Replace any one or more elements of the disease name in the basic positive sample pair to obtain candidate negative sample pairs; the disease names in the basic positive sample pair correspond to clinical diseases; If the disease name of the candidate negative sample pair is the same as the disease name in any base positive sample pair, the candidate negative sample pair is used as the enhanced negative sample pair. The disease name comparison model, trained based on basic positive sample pairs and enhanced negative sample pairs, compares the disease names to be compared and obtains the comparison results, including: Disease element parsing is performed on each disease name to be compared to obtain the elements of each disease name to be compared; based on the encoding parameters, the elements of each disease name to be compared are encoded to obtain the encoding features of each disease name to be compared; wherein, the encoding includes: mapping each element to a vector through a randomly initialized word vector layer and then inputting it into a BiLSTM layer for encoding, taking the last character of the bidirectional sequence for encoding and concatenating it to obtain the encoding features of each element, and then inputting it into a GLU layer for processing through a gating mechanism; A high-dimensional attention mechanism is used to perform pairwise interactions on the encoded features of each disease name to be compared, so as to obtain the relationship between any disease name and all disease names. Based on the classification parameters and the relationship between each disease name to be compared, the comparison results between each disease name are determined.

2. The disease name comparison method for medical insurance according to claim 1, characterized in that, The disease name comparison model, trained based on basic positive sample pairs and enhanced negative sample pairs, compares the disease names to be compared and obtains the comparison results, including: The disease name comparison model trained based on the basic negative sample pairs, the basic positive sample pairs, and the enhanced negative sample pairs is used to compare the disease names to be compared and obtain the comparison results. The base positive sample pairs and the base negative sample pairs are determined based on the medical record disease names in the medical record text and / or the standard disease names in the disease system.

3. The disease name comparison method for medical insurance according to claim 2, characterized in that, The disease name comparison model is trained sequentially based on a first type of sample, a second type of sample, and a third type of sample. The first type of sample includes basic positive sample pairs and basic negative sample pairs determined based on the disease name in the medical record. The second type of sample includes enhanced negative sample pairs and basic positive sample pairs determined based on the disease name in the medical record. The third type of sample includes basic positive sample pairs and basic negative sample pairs determined based on the standard disease name.

4. The disease name comparison method for medical insurance according to claim 2, characterized in that, The base positive sample pairs and the base negative sample pairs are determined based on the following steps: Determine the hierarchical position of each standard disease name within the disease system; Based on the hierarchical relationship between the standard disease names, positive and negative examples are divided into positive and negative examples for each standard disease name, resulting in the basic positive sample pairs and the basic negative sample pairs.

5. The disease name comparison method for medical insurance according to claim 4, characterized in that, Based on the hierarchical relationship between standard disease names, each standard disease name is divided into positive and negative examples to obtain the basic positive sample pairs and the basic negative sample pairs, including: If one of the two standard disease names is a higher-level standard disease name than the other, the two standard disease names are assigned to the basic positive sample pair. If the two standard disease names are at the same level, the two standard disease names are assigned to the base negative sample pair.

6. A disease name comparison device for medical insurance, characterized in that, include: The acquisition unit is used to acquire the name of the disease to be compared. The comparison unit is used to compare the disease names to be compared by the disease name comparison model trained based on the basic positive sample pairs and the enhanced negative sample pairs, and obtain the comparison results. The basic positive sample pair includes multiple disease names corresponding to the same disease, and the enhanced negative sample pair includes disease names corresponding to different diseases. The enhanced negative sample pair is obtained by replacing elements in the disease names in the basic positive sample pair. The enhanced negative sample pairs are determined based on the following steps: Disease parsing is performed on each disease name in the basic positive sample pair to obtain the elements of each disease name; the elements are used to describe the corresponding disease, and the elements include clinical manifestations, location, etiology and pathology; Replace any one or more elements of the disease name in the basic positive sample pair to obtain candidate negative sample pairs; the disease names in the basic positive sample pair correspond to clinical diseases; If the disease name of the candidate negative sample pair is the same as the disease name in any base positive sample pair, the candidate negative sample pair is used as the enhanced negative sample pair. The disease name comparison model, trained based on basic positive sample pairs and enhanced negative sample pairs, compares the disease names to be compared and obtains the comparison results, including: Disease element parsing is performed on each disease name to be compared to obtain the elements of each disease name to be compared; based on the encoding parameters, the elements of each disease name to be compared are encoded to obtain the encoding features of each disease name to be compared; wherein, the encoding includes: mapping each element to a vector through a randomly initialized word vector layer and then inputting it into a BiLSTM layer for encoding, taking the last character of the bidirectional sequence for encoding and concatenating it to obtain the encoding features of each element, and then inputting it into a GLU layer for processing through a gating mechanism; A high-dimensional attention mechanism is used to perform pairwise interactions on the encoded features of each disease name to be compared, so as to obtain the relationship between any disease name and all disease names. Based on the classification parameters and the relationship between each disease name to be compared, the comparison results between each disease name are determined.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the disease name comparison method for medical insurance as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the disease name comparison method for medical insurance as described in any one of claims 1 to 5.

Images