An abnormal medical behavior identification method, device and storage medium

Abstract

Embodiments of the present application provide an abnormal medical behavior identification method and device and a storage medium. The diagnosis and medical items can be extracted from real medical records. The medical items in the medical records can be traversed. If the current traversed medical item does not appear in the preset medical knowledge graph or the associated diagnosis in the medical knowledge graph does not appear in the target medical record, the first affinity and the second affinity between the current traversed medical item and the extracted diagnosis can be further obtained. If the two affinities do not satisfy the preset condition, it is determined that the current traversed medical item belongs to an abnormal medical item. In this way, the abnormal medical behavior identification can be performed in combination with dynamic real data and static experience data. In the case that the experience data is not comprehensive, the affinity between the medical items and the diagnosis can be used as a further identification basis, so that the identification accuracy and reliability of the abnormal medical behavior can be effectively improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, device and storage medium for identifying abnormal medical behavior. Background Technology

[0002] To strengthen the supervision and management of the use of medical insurance funds, ensure fund security, promote the effective use of funds, and safeguard citizens' legitimate rights and interests in medical insurance, the government has formulated the "Regulations on the Supervision and Management of the Use of Medical Insurance Funds," which has been promulgated and implemented. Medical insurance funds involve numerous users, a long chain of transactions, many risk points, and significant challenges in supervision, leading to frequent illegal activities and widespread fraud.

[0003] Currently, the main medical insurance payment method is fee-for-service (hospitals are gradually shifting to DRG, DIP and other grouped bundled payment methods). However, in many medical treatment processes, there are abnormal medical behaviors, such as the use of unnecessary drugs and unnecessary procedures, which leads to the waste of medical insurance funds.

[0004] Therefore, it is necessary to identify these abnormal medical behaviors. Summary of the Invention

[0005] This application provides a method, device, and storage medium for identifying abnormal medical behavior, in order to more accurately identify abnormal medical behavior.

[0006] This application provides a method for identifying abnormal medical behavior, including:

[0007] Extract diagnoses and medical items from the target medical records;

[0008] Iterate through the medical items contained in the target medical records;

[0009] If the currently traversed medical item does not appear in the preset medical knowledge graph, or the diagnosis associated with the currently traversed medical item in the medical knowledge graph does not appear in the target medical record, then obtain the first affinity of the currently traversed medical item to the extracted diagnosis and the second affinity of the extracted diagnosis to the currently traversed medical item.

[0010] If the first affinity and the second affinity do not meet the preset conditions, the medical item being traversed is determined to be an abnormal medical item.

[0011] This application also provides a computing device, including a memory and a processor;

[0012] The memory is used to store one or more computer instructions;

[0013] The processor is coupled to the memory and is used to execute the one or more computer instructions for:

[0014] Extract diagnoses and medical items from the target medical records;

[0015] Iterate through the medical items contained in the target medical records;

[0016] If the currently traversed medical item does not appear in the preset medical knowledge graph, or the diagnosis associated with the currently traversed medical item in the medical knowledge graph does not appear in the target medical record, then obtain the first affinity of the currently traversed medical item to the extracted diagnosis and the second affinity of the extracted diagnosis to the currently traversed medical item.

[0017] If the first affinity and the second affinity do not meet the preset conditions, the medical item being traversed is determined to be an abnormal medical item.

[0018] This application also provides a computer-readable storage medium for storing computer instructions, which, when executed by one or more processors, cause the one or more processors to perform the aforementioned abnormal medical behavior identification method.

[0019] In this embodiment, diagnoses and medical items can be extracted from real medical records. The medical items in the medical records can be traversed. If the currently traversed medical item does not appear in a preset medical knowledge graph, or if the associated diagnosis in the medical knowledge graph does not appear in the target medical record, then a first affinity and a second affinity between the currently traversed medical item and the extracted diagnosis can be further obtained. If the first affinity and the second affinity do not meet preset conditions, the currently traversed medical item is determined to be an abnormal medical item. In this way, dynamic real data and static empirical data can be combined for abnormal medical behavior identification. Furthermore, when empirical data is incomplete, the affinity between medical items and diagnoses can be used as a further identification criterion, thereby effectively improving the accuracy and reliability of abnormal medical behavior identification. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0021] Figure 1 A flowchart illustrating an exemplary embodiment of this application for identifying abnormal medical behavior;

[0022] Figure 2A logical schematic diagram of an abnormal medical behavior identification method provided for an exemplary embodiment of this application;

[0023] Figure 3 A logical schematic diagram of an application scheme provided for an exemplary embodiment of this application;

[0024] Figure 4 This is a schematic diagram of the structure of a computing device provided for another exemplary embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] Currently, many abnormal medical behaviors occur during medical treatment, leading to a waste of medical insurance funds. Therefore, it is necessary to identify these abnormal medical behaviors. To this end, in some embodiments of this application: diagnoses and medical items can be extracted from real medical records; and the medical items in the medical records can be traversed. If the currently traversed medical item does not appear in a preset medical knowledge graph, or if the diagnosis associated with the medical knowledge graph does not appear in the target medical record, then the first affinity and second affinity between the currently traversed medical item and the extracted diagnosis can be further obtained; and if the first affinity and second affinity do not meet preset conditions, the currently traversed medical item is determined to be an abnormal medical item. In this way, abnormal medical behavior can be identified by combining dynamic real data and static empirical data, and when empirical data is incomplete, the affinity between medical items and diagnoses can be used as a further identification basis, thereby effectively improving the accuracy and reliability of abnormal medical behavior identification.

[0027] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0028] Figure 1 This is a flowchart illustrating an exemplary embodiment of an abnormal medical behavior identification method provided in this application. Figure 2 This is a logical schematic diagram of an abnormal medical behavior identification method provided as an exemplary embodiment of this application. The method can be executed by a data processing device, which can be implemented as a combination of software and / or hardware, and can be integrated into a computing device. (Reference) Figure 1 The method includes:

[0029] Step 100: Extract diagnoses and medical items from the target medical records;

[0030] Step 101: Traverse the medical items contained in the target medical record;

[0031] Step 102: If the medical item being traversed does not appear in the preset medical knowledge graph or the diagnosis associated with the medical item being traversed in the medical knowledge graph does not appear in the target medical record, then obtain the first affinity of the medical item being traversed to the extracted diagnosis and the second affinity of the extracted diagnosis to the medical item being traversed.

[0032] Step 103: If neither the first affinity nor the second affinity meets the preset conditions, determine that the medical item being traversed is an abnormal medical item.

[0033] The abnormal medical behavior identification method provided in this embodiment can be applied to scenarios such as medical insurance fund supervision and medical institution behavior management. This embodiment does not limit the application scenario. The abnormal medical behavior mentioned in this embodiment can refer to medical behavior that exceeds actual needs.

[0034] refer to Figure 1 and Figure 2 In step 100, diagnoses and medical items can be extracted from the target medical record. The target medical record can be any record requiring abnormal medical behavior identification. Each patient visit generates a medical record, which typically records the patient's symptoms and treatment details according to regulations; that is, the medical record usually contains rich and comprehensive information related to the visit. Based on this, various text recognition technologies can be used to extract the diagnoses and medical items contained in the target medical record. Diagnoses can be used to record the diagnostic results that occurred during the visit; medical items can be used to record the medical procedures used during the visit, which are usually reflected in the billing details. A medical record may contain multiple diagnoses, such as a primary diagnosis and secondary diagnoses; it may also contain multiple medical items, such as medications used, surgical procedures, and nursing care. In this embodiment, the reasonableness of multiple medical items in the target medical record will be determined separately.

[0035] Therefore, step 101 proposes that the medical items contained in the target medical record can be traversed. That is, in this embodiment, steps 102 and 103 will be executed for each medical item in the target medical record to determine whether each medical item is an abnormal medical item. The judgment logic will be described below from the perspective of "currently traversed medical items". It should be understood that "currently traversed medical items" can be any medical item in the target medical record.

[0036] refer to Figure 1 and Figure 2 In step 102, if the currently traversed medical item does not appear in the preset medical knowledge graph, or if the diagnosis associated with the currently traversed medical item in the medical knowledge graph does not appear in the target medical record, then the first affinity of the currently traversed medical item to the extracted diagnosis and the second affinity of the extracted diagnosis to the currently traversed medical item are obtained. The medical knowledge graph can serve as a basis for identifying abnormal medical behavior. The medical knowledge graph can be derived from experience: it can be expert experience or experience learned from historical data; this embodiment does not limit this. Thus, in step 102, the rationality of the currently traversed medical item can first be judged based on the medical knowledge graph. If the medical knowledge graph supports the rationality of the currently traversed medical item, then it can be directly determined that the currently traversed medical item is not an abnormal medical item. If the medical knowledge graph cannot support the rationality of the currently traversed medical item, then the affinity calculation in step 102 and subsequent steps can be further executed to further analyze the rationality of the currently traversed medical item, thus avoiding the limitations of the medical knowledge graph.

[0037] In this embodiment, various implementation methods can be used to determine whether the currently traversed medical item appears in the preset medical knowledge graph. Various implementation methods can also be used to determine whether the diagnosis associated with the currently traversed medical item in the medical knowledge graph appears in the target medical record. Specific implementation methods will be detailed later.

[0038] refer to Figure 1 and Figure 2 In step 102, if the currently traversed medical item does not appear in the preset medical knowledge graph, or if the diagnosis associated with the currently traversed medical item in the medical knowledge graph does not appear in the target medical record, the first affinity of the currently traversed medical item to the extracted diagnosis and the second affinity of the extracted diagnosis to the currently traversed medical item can be obtained. It is important to emphasize that for a diagnosis and a medical item, the affinity of the diagnosis to the medical item and the affinity of the medical item to the diagnosis are different; affinity can be understood as support. In this embodiment, based on historical data, the affinity of each medical item to each diagnosis and the affinity of each diagnosis to each medical item can be pre-calculated. Based on this, in step 102, the required first and second affinity can be directly obtained from the pre-calculated affinity set. The affinity set can be indexed using identifiers such as the names of diagnoses and medical items.

[0039] It is worth noting that if the target medical record contains multiple diagnoses, then in step 102, the first affinity of the currently traversed medical item to each extracted diagnosis can be obtained, and the second affinity of each extracted diagnosis to the currently traversed medical item can be obtained. In this embodiment, the first affinity and the second affinity are merely approximations of affinity in different directions.

[0040] Based on this, in step 103, if neither the first affinity nor the second affinity meets the preset conditions, the currently traversed medical item can be determined to be an abnormal medical item. However, if the currently traversed medical item appears in a preset medical knowledge graph and the diagnosis associated with it in the medical knowledge graph also appears in the target medical record, or although the currently traversed medical item does not appear in the preset medical knowledge graph or the diagnosis associated with it in the medical knowledge graph does not appear in the target medical record, but either the first affinity or the second affinity meets the preset conditions, then the currently traversed medical item is determined not to be an abnormal medical item. The preset conditions can be that the first affinity is not lower than a first threshold or the second affinity is not lower than a second threshold. The first threshold and the second threshold can be the same or different; in practical applications, they can be set according to the desired recognition effect. Here, when the target medical record contains multiple diagnoses, if the first and second affinity of each diagnosis in the target medical record does not meet the preset conditions with respect to the currently traversed medical item, then the currently traversed medical item is determined to be an abnormal medical item; otherwise, the currently traversed medical item is determined not to be an abnormal medical item. In other words, if there is at least one diagnosis in the target medical record that supports the rationality of the currently traversed medical item, then the currently traversed medical item can be considered not to be an abnormal medical item.

[0041] For example, in medical record 1, which includes medical item x1 and diagnoses d1 and d2, if the affinity P(x1|d1) of medical item x1 to diagnosis d1 and the affinity P(d1|x1) of diagnosis d1 to medical item x1 are both less than their respective specified thresholds, and the affinity P(d2|x1) of diagnosis d2 to medical item x1 is less than its corresponding specified threshold, but since the affinity P(x1|d2) of medical item x1 to diagnosis d2 is greater than its corresponding specified threshold, then medical item x1 will be determined not to be an abnormal medical item, because medical item x1 is reasonable for diagnosis d1.

[0042] In this way, by iterating through the records, it can be determined whether each medical item in the target medical record is an abnormal medical item.

[0043] Based on this, the percentage of abnormal medical items in the target medical record can be calculated. If the percentage exceeds a specified standard, the target medical record is determined to be an abnormal visit. An exemplary statistical scheme could be: configure a counter for the target medical record; increment the counter by 1 for each abnormal medical item found in the record to count the total number of abnormal medical items, and then calculate the percentage of abnormal medical items in the record (i.e., the ratio between the number of abnormal medical items and the total number of medical items). A specified standard can be defined for this percentage, for example, 0.2. If the calculated percentage exceeds the specified standard, the target medical record is determined to be an abnormal visit. Using the "percentage" method ensures a certain degree of robustness and prevents uncommon cases from being identified.

[0044] Accordingly, in this embodiment, diagnoses and medical items can be extracted from real medical records. The medical items in the medical records can be traversed. If the currently traversed medical item does not appear in a preset medical knowledge graph, or if the associated diagnosis in the medical knowledge graph does not appear in the target medical record, then a first affinity and a second affinity between the currently traversed medical item and the extracted diagnosis can be further obtained. If neither the first affinity nor the second affinity meets preset conditions, the currently traversed medical item is determined to be an abnormal medical item. In this way, abnormal medical behavior can be identified by combining dynamic real data and static empirical data. Furthermore, when empirical data is incomplete, the affinity between medical items and diagnoses can be used as a further identification criterion, thereby effectively improving the accuracy and reliability of abnormal medical behavior identification.

[0045] In the above or following embodiments, the method for calculating affinity may be:

[0046] From a specified range of historical medical records, diagnoses, medical procedures, and the relationships between them are extracted. Diagnoses and medical procedures appearing in the same historical medical record are considered to have a relationship. Based on these relationships, the affinity of each diagnosis to each medical procedure and the affinity of each medical procedure to each diagnosis are calculated to generate a first affinity and a second affinity. Here, the specified range of historical medical records can be medical records occurring within a specified historical time frame, such as those occurring within the last six months.

[0047] In this embodiment, an exemplary scheme is provided for calculating the first affinity and the second affinity between the currently traversed medical item and the target diagnosis in the target medical record: Based on the association relationship, the proportion of the number of visits using the currently traversed medical item and diagnosed with the target diagnosis to the total number of visits corresponding to the target diagnosis is calculated as the first affinity of the currently traversed medical item to the target diagnosis; based on the association relationship, the proportion of the number of visits using the currently traversed medical item and diagnosed with the target diagnosis to the total number of visits using the currently traversed medical item is calculated as the second affinity of the target diagnosis to the currently traversed medical item; wherein, the target diagnosis is any one of the diagnoses included in the target medical record.

[0048] In this exemplary scheme, conditional probability is used to calculate affinity. In this exemplary calculation scheme, the first affinity of the currently traversed medical item x to the target diagnosis d can be represented as P(x|d) = the number of visits using medical item x and being diagnosed as d / the number of visits being diagnosed as d; and the second affinity of the target diagnosis d to the currently traversed medical item x can be represented as P(d|x) = the number of visits using medical item x and being diagnosed as d / the number of visits using medical item x. Of course, other implementation schemes can also be used to calculate the first affinity and the second affinity in this embodiment, and this embodiment is not limited to this.

[0049] Typically, with a sufficient amount of historical medical records, a relatively comprehensive range of medical items and diagnoses can be obtained. In this exemplary scheme, all relevant affinities can be pre-calculated, thereby generating an affinity set. In this case, the process of calculating the first and second affinities for the currently traversed medical items is essentially occurring during the pre-calculation process, thus calculating a relatively comprehensive affinity. Based on this, during the process of identifying abnormal medical behavior in a target medical record, the required affinity can be conveniently and quickly found from the pre-calculated affinity set without real-time calculation. Of course, this embodiment is not limited to this. In this embodiment, the above calculation logic can also support the real-time calculation of the required first and second affinities during the process of identifying abnormal medical behavior in a target medical record. During the real-time calculation process, historical medical records can be called up as needed.

[0050] In the above or below embodiments, during the process of traversing the medical items in the target medical record, it can be first determined whether the currently traversed medical item appears in a preset medical knowledge graph.

[0051] The medical items extracted from the target medical records may have a name diversity problem. That is, there may be a name inconsistency between the extracted medical items and the same medical items in the pre-built medical knowledge graph.

[0052] Therefore, in this embodiment, various implementation methods can be used to solve the problem of inconsistent names, thereby more accurately determining whether the currently traversed medical item is included in the preset medical knowledge graph.

[0053] In one optional implementation, the standard name corresponding to the currently traversed medical item can be determined; if no pre-defined medical item matching the standard name exists in the medical knowledge graph, it is determined that the currently traversed medical item does not appear in the medical knowledge graph. That is, in this embodiment, the names of medical items extracted from the target medical records can be aligned to ensure the accuracy of the search within the medical knowledge graph.

[0054] In this implementation, based on historical medical records, different medical item names corresponding to the same medical item can be pre-normalized to a unified standard name. Here, two scenarios may exist:

[0055] In one scenario, if a standard name directory exists, an exemplary approach to pre-constructing standard names in this implementation could be to match the currently traversed medical item name to the corresponding standard name. Specifically: in the standard name directory, search for a standard name that is completely identical to the medical item name. If such a name exists, the medical item name can be normalized to that standard name; if not, the similarity between the medical item name and each standard name in the standard name directory can be calculated using methods such as longest common subsequence, Jaccard similarity, and edit distance similarity. Then, based on the calculated similarity, the medical item name can be normalized to the standard name with the highest similarity or the one that matches the average similarity. The specific implementation scheme for selecting a standard name based on similarity is not limited here. The longest common subsequence method can refer to using the length of the longest common character sequence between two strings to measure the similarity between them. Other similarity calculation methods will not be detailed here. The same method can be used to normalize the extracted diagnosis names. The extracted diagnosis names can be normalized using the same scheme as the medical project names, which will not be repeated here.

[0056] In another scenario, if a standard name directory does not exist in the scenario, an exemplary approach to pre-constructing standard names could be: a self-aggregating implementation could be used to determine the standard name corresponding to the currently traversed medical item. In this exemplary approach: medical item names can be extracted from a specified range of historical medical records; the similarity between each pair of medical item names can be calculated; the medical item names can be clustered based on the similarity to obtain multiple independent sets of medical item names; and the medical item names in the same set can be normalized to a unified standard name to generate the standard name for the medical item. Similarly, the standard name for the diagnosis can be normalized. The similarity measurement performed here can cover all medical item names and diagnosis names extracted from historical medical records. The similarity measurement method can use methods such as the longest common subsequence, Jaccard similarity, and edit distance similarity mentioned earlier. For example, the length of the longest common subsequence between two medical item names can be calculated, and the ratio between the calculated length and the average string length of the two medical item names can be used as the similarity, or the ratio between the calculated length and the length of the longest string in the two medical item names can be used as the similarity. An exemplary clustering scheme could be as follows: if the similarity between two medical item names exceeds a given threshold, an edge can be created between the two medical item names to link similar medical item names together, thus constructing a medical item association graph. In the medical item association graph, similar medical item names are linked by edges; the medical item association graph will contain multiple disconnected subgraphs, each corresponding to a set of medical item names. The same scheme can be used to cluster diagnostic names. In this implementation, the clustering results based on similarity can be further optimized during the clustering process. An exemplary optimization dimension could be: determining whether two diagnostic names / medical item names whose similarity meets the clustering requirements correspond to the same part / organ / tissue / structure, etc. If not, the clustering relationship between the two names is broken. Here, "part" can refer to a body part. For example, although the similarity between the two diagnostic names "malignant renal tumor" and "malignant lung tumor" meets the clustering requirements, the two diagnostic names correspond to different parts—one is the kidney, and the other is the lung—therefore, these two diagnostic names are not clustered together. Another exemplary optimization dimension could be: removing non-distinguishing characters from each diagnosis / medical project name to obtain a simplified name; calculating the similarity based on the simplified names corresponding to each diagnosis / medical project name. This can be done by using word segmentation tools to segment each diagnosis / medical project name, counting the frequency of each segment, and selecting the k most frequent segments as non-distinguishing characters. For accuracy, manual verification can be introduced to ensure the rationality of the selected non-distinguishing characters.For example, the characters "chronic" in "chronic kidney disease" and "disease" in "hypertension" are characters without distinctiveness. By removing these non-distinctive characters before measuring similarity, the problem of increased similarity due to too many non-distinctive characters in the names can be avoided, effectively improving the accuracy and rationality of similarity measurement. Another exemplary optimization dimension could be: determining whether two diagnosis names / medical item names that meet the clustering requirements have antonyms. If so, the clustering relationship between the two names can be broken. For example, the diagnosis names "benign tumor" and "malignant tumor" meet the clustering requirements in terms of similarity, but if there are antonyms, then these two diagnosis names do not correspond to the same diagnosis. By optimizing the clustering results based on similarity, the accuracy of the diagnosis name set / medical item name set can be effectively improved, thereby improving the accuracy of the normalization results. In addition, in this exemplary scheme, the number of visits corresponding to each item name within the same item name set can be counted; the item name with the most visits is used as the standard name for normalization of the item name set. Of course, the diagnosis name / medical procedure name with the shortest string length can also be used as the normalization result; the diagnosis name / medical procedure name corresponding to the median number of visits can also be used as the normalization result, and so on.

[0057] Based on the pre-constructed standard names, this implementation method can use methods such as longest common subsequence, Jaccard similarity, and edit distance similarity to calculate the similarity between the name of the currently traversed medical item and the pre-determined standard names of each medical item, thereby finding the standard name corresponding to the currently traversed medical item. The process of finding the standard name can be referred to the previous text and will not be repeated here.

[0058] After determining the standard name corresponding to the medical item being traversed, we can continue to check whether there is a pre-defined medical item in the medical knowledge graph that matches the standard name. If it does not exist, we can determine that the current medical item does not appear in the medical knowledge graph.

[0059] An exemplary scheme for determining whether a pre-defined medical item matching a standard name exists in a medical knowledge graph can be as follows: Calculate the similarity between the standard name corresponding to the currently traversed medical item and the names of pre-defined medical items contained in the medical knowledge graph; if there are candidate pre-defined medical items with a similarity not lower than a first specified threshold, it can be determined that a pre-defined medical item matching the standard name exists in the medical knowledge graph. Further, in this exemplary scheme, pre-defined medical items that do not meet the name consistency requirements can be further removed from the candidate pre-defined medical items; if all candidate pre-defined medical items are removed, it is then determined that no pre-defined medical item matching the standard name exists in the medical knowledge graph. This can more accurately determine whether a pre-defined medical item matching the standard name exists in the medical knowledge graph. The name consistency requirements may include one or more of the following: part / organ / tissue consistency, positive / negative consistency, negative consistency, object consistency, action consistency, or morphological consistency. Among them, the consistency requirement of part / organ / tissue can be that there is no different part / organ / tissue between the standard name of the currently traversed medical item and the name of the preset medical item; the positive-negative consistency requirement can be that there is no antonym between the standard name of the currently traversed medical item and the name of the preset medical item. Other types of consistency requirements will not be detailed here.

[0060] Here, if all the candidate pre-set medical items are removed, it can be determined that the medical item being traversed does not appear in the pre-set medical knowledge graph, and the calculation of affinity and subsequent operations in step 102 above can be continued to further analyze the rationality of the medical item being traversed.

[0061] If not all of the candidate pre-set medical items are eliminated, it can be determined that the currently traversed medical item appears in the pre-set medical knowledge graph, and it can be further determined whether the diagnosis associated with the currently traversed medical item in the medical knowledge graph appears in the target medical record.

[0062] An exemplary embodiment for determining whether the diagnosis associated with the currently traversed medical item in the medical knowledge graph appears in the target medical record may be as follows: First, determine the pre-set medical item matching the standard name from the remaining candidate pre-set medical items, and use it as the matching medical item; second, search the medical knowledge graph for the diagnoses associated with each matching medical item, and use them as related diagnoses; third, if there is no identical diagnosis between the diagnoses extracted from the target medical record and the related diagnoses, then determine that the diagnosis associated with the currently traversed medical item in the medical knowledge graph does not appear in the target medical record. Various implementation schemes can be used to determine the matching medical item, including but not limited to: selecting the one with the highest similarity from the remaining candidate pre-set medical items as the matching medical item; or, using the remaining candidate pre-set medical items as the matching medical item; or, selecting the K most similar items from the remaining candidate pre-set medical items as the matching medical item, where K is an integer greater than 1. In this exemplary scheme, the diagnoses associated with each matched medical item can form a set of related diagnoses. Based on this, it can be determined whether there is a related diagnosis in the set of related diagnoses that matches any diagnosis name extracted from the target medical record. If there is, it proves that there is a common diagnosis between the two, and it can be determined that the diagnosis associated with the currently traversed medical item in the medical knowledge graph appears in the target medical record. If they do not match, preferably, the similarity between the diagnosis extracted from the target medical record and the related diagnoses can be further calculated separately. If there are candidate diagnoses with a similarity of not less than a second preset value among the related diagnoses, then diagnoses that do not meet the name consistency requirement are removed from the candidate diagnoses. If all candidate diagnoses are removed, it is determined that there is no common diagnosis between the diagnosis extracted from the target medical record and the related diagnoses. If not all candidate diagnoses are removed, it can still be determined that the diagnosis associated with the currently traversed medical item in the medical knowledge graph appears in the target medical record. Here, the requirements for name consistency may include, but are not limited to, one or more of the following: consistency of part / organ / tissue, positive / negative consistency, negative consistency, object consistency, action consistency, or morphological consistency. Please refer to the previous text for details, which will not be repeated here.

[0063] If the diagnosis associated with the currently traversed medical item appears in the target medical record, it indicates a correlation between the currently traversed medical item and the diagnosis contained in the target medical record, meaning the currently traversed medical item is reasonable and not an abnormal medical item. However, if the diagnosis associated with the currently traversed medical item does not appear in the target medical record, then it is necessary to continue calculating the affinity and performing subsequent operations as described in step 102 to further analyze the reasonableness of the currently traversed medical item.

[0064] Of course, in this embodiment, other implementation methods can also be used to determine whether the currently traversed medical item appears in the preset medical knowledge graph, and other implementation methods can also be used to determine whether the diagnosis associated with the currently traversed medical item in the medical knowledge graph appears in the target medical record. This embodiment is not limited to these. For example, first search for the preset medical items associated with each diagnosis in the target medical record in the medical knowledge graph, and then determine whether the currently traversed medical item is included in the preset medical items, etc., which will not be described in detail here.

[0065] This embodiment provides a solution to the problem of name diversity in diagnoses and medical items in real data. Using standardized names ensures consistency, thereby improving data quality and recognition accuracy. Furthermore, starting with medical items in the target medical record, it searches for related pre-defined medical items and diagnoses in the medical knowledge graph, gradually determining whether the medical knowledge graph can support the rationality of the currently traversed medical items.

[0066] Figure 3 A logical schematic diagram of an application scheme provided for an exemplary embodiment of this application. (Reference) Figure 3 In this application scheme, the medical knowledge graph contains the relationships between diagnoses and pre-defined medical items. The target medical record contains diagnoses d = {d1, d2, d3} and medical items x = {x1, x2, x3, x4}. In this exemplary scheme, each medical item can be traversed.

[0067] For each medical item, the search is first performed in the medical knowledge graph. If an equivalent match is found, it is found directly. If no equivalent match is found, an approximate match is used. Specifically, a set of candidate medical items with a similarity of no less than a given threshold t1 is obtained. Then, some special candidate medical items (i.e., those that do not meet the consistency test, such as inconsistent parts or conflicting antonyms) are removed from this set. If the set is not empty after removing these candidate medical items, it can be determined that the medical item being traversed has been found in the medical knowledge graph.

[0068] Then, based on the remaining candidate medical items, the associated diagnoses in the medical knowledge graph are searched, resulting in a set of related diagnoses. Next, it is determined whether this set intersects with the target consultation's diagnosis set d (i.e., they share the same diagnosis). If they do, it means the diagnosis associated with the currently traversed medical item in the medical knowledge graph appears in the target consultation. If not, for each related diagnosis, its similarity to each diagnosis in the target consultation is calculated, resulting in a set of candidate diagnoses with a similarity at least equal to a given threshold t2. Then, some special diagnoses (i.e., those that do not meet consistency checks, such as inconsistent locations or conflicting antonyms) are removed from this set. If the set is not empty after the removal operation, it also means the diagnosis associated with the currently traversed medical item in the medical knowledge graph appears in the target consultation. Both of these cases indicate that the currently traversed medical item is reasonable.

[0069] If, based on the medical knowledge graph, it is initially determined that the medical item being traversed is unreasonable, then two affinities between the medical item and each diagnosis of the target patient can be obtained. As long as there exists any affinity between the medical item and one of the diagnoses that is not lower than the given threshold t3, it indicates that the medical item is also reasonable.

[0070] If both of the above dimensions determine that the currently traversed medical item is unreasonable, this indicates that the medical item is unreasonable. The count of unreasonable items is then incremented by 1.

[0071] Finally, after iterating through every medical item in the target medical record, the number of unreasonable items is divided by the total number of medical items in the target medical record. If the percentage is lower than the given threshold t4, it means that the medical visit was normal; otherwise, it is abnormal.

[0072] Combination Figure 3 During the process of traversing the medical projects x = {x1, x2, x3, x4}, the following situations may occur.

[0073] For x1:

[0074] First, we search the medical knowledge graph. Let's assume we find an equivalent match, which is kx2.

[0075] The set of diagnoses associated with kx2 is kd = {kd1}. The question is whether the set kd = {kd1} shares the same diagnosis as the set of diagnoses d = {d1, d2, d3} for this specific visit. Specifically, iterate through each diagnosis in kd, performing an equivalence check. If an equivalence diagnosis is found, then the diagnoses are considered the same, and the check is complete. This indicates that x1 can be found in the knowledge graph, and the associated diagnosis also appears in this visit, suggesting that x1 is a normal item.

[0076] If no equivalent value in kd appears in d, then iterate through each diagnosis kd_i in kd (here only kd1), perform similarity judgment, and obtain the set of diagnoses in d whose similarity is not lower than the specified threshold t2, let's say it's {d1, d2}. Then remove diagnoses that fail consistency checks such as inconsistent parts or non-contradictory antonyms, let's say the filtered set is {d1}. Since it's not empty, it means that x1 can be found in the medical knowledge graph, and the associated diagnosis also appears in this visit, indicating that x1 is a normal item.

[0077] For x2:

[0078] First, a search is performed in the medical knowledge graph. Assuming no equivalent match is found, a similarity check is performed. Let's say the set with a similarity score no lower than a specified threshold t1 is {kx1, kx3, kx4}. We then check if there are any discrepancies in location / tissue / organ, or if there are conflicts in the antonyms or other consistency checks. If so, these are removed. Let's say the filtered set is {kx3, kx4}. The associated diagnosis set for {kx3, kx4} is kd = {kd2, kd3}. We then check if the set kd = {kd2, kd3} has the same diagnosis as the diagnosis set d = {d1, d2, d3} from this particular visit. Specifically, we iterate through each diagnosis in kd, performing an equivalence check. If an equivalent diagnosis is found, then the same diagnosis exists, and the check is complete. This indicates that x2 can be found in the medical knowledge graph, and the associated diagnosis also appears in this visit, meaning x1 is a normal item.

[0079] If no equivalent value in kd appears in d, then iterate through each diagnosis kd_i (where i = 2, 3) in kd, perform similarity judgment, and obtain the set of diagnoses in d whose similarity to kd_i is not lower than the specified threshold t2, let's say it's {d1, d2}. And remove diagnoses that fail consistency checks such as inconsistent parts or contradictory antonyms. Let's say the unfiltered diagnoses are {d1, d2}. Since they are not empty, it means that x2 can be found in the medical knowledge graph, and the associated diagnosis also appears in this visit, indicating that x1 is a normal item.

[0080] For x3:

[0081] First, a search is performed in the medical knowledge graph. Assuming no equivalent match is found, a similarity check is performed. Let {kx7} be the set of items with a similarity score at or above a specified threshold t1, and let kd = {kd4} be the set of diagnoses associated with {kx7}. The question is whether set kd = {kd4} shares the same diagnosis as the diagnosis set d = {d1,d2,d3} for this particular visit. If kd = {kd4} and d = {d1,d2,d3} are not equivalent, and there is no similar set at or above the given threshold t2, it means that x3 cannot find an item-diagnosis relationship for this visit in the knowledge graph. Next, affinity is used for further evaluation. Based on six affinity values: p(x3|d = {d1,d2,d3}) and p(d = {d1,d2,d3}x3), if any one of these affinity values ​​is at or above the specified threshold t3, x3 is considered normal based on affinity; otherwise, it is considered abnormal, and the count is incremented by 1. Here, we assume it is normal.

[0082] For x4:

[0083] First, a search is performed in the medical knowledge graph. Assuming no equivalent match is found, a similarity check is performed. If no item set has a similarity score greater than or equal to a specified threshold t1, it indicates the item is not in the knowledge graph. Further, an affinity-based approach is used. Based on the six affinities p(x3|d={d1,d2,d3}) and p(d={d1,d2,d3}x3), assuming all six are below the given threshold t3, x4 is considered an anomaly. The count is incremented by 1.

[0084] Based on the previous calculations, only x4 is abnormal, and the count of abnormal items is 1. The total number of items in this visit is 4, and the abnormality rate is 1 / 4, which is lower than the given threshold t4 (assumed to be 0.3). This indicates that although there are suspected abnormal items in this visit, the overall situation is normal.

[0085] It should be noted that some processes described in the above embodiments and accompanying drawings include multiple operations appearing in a specific order. However, it should be clearly understood that these operations may not be executed in the order they appear in this document, or they may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should also be noted that the descriptions such as "first" and "second" in this document are used to distinguish different degrees of affinity, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0086] Figure 4 This is a schematic diagram of the structure of a computing device provided for another exemplary embodiment of this application. For example... Figure 4As shown, the computing device includes a memory 40 and a processor 41.

[0087] Processor 41, coupled to memory 40, is used to execute computer programs in memory 40 for:

[0088] Extract diagnoses and medical items from the target medical records;

[0089] Iterate through the medical items included in the target medical records;

[0090] If the medical item being traversed does not appear in the preset medical knowledge graph, or if the diagnosis associated with the medical item being traversed does not appear in the target medical record, then obtain the first affinity of the medical item being traversed to the extracted diagnosis and the second affinity of the extracted diagnosis to the medical item being traversed.

[0091] If the first affinity and the second affinity do not meet the preset conditions, the medical item being traversed is determined to be an abnormal medical item.

[0092] In one optional embodiment, the preset conditions include a first affinity not lower than a first threshold or a second affinity not lower than a second threshold.

[0093] In an optional embodiment, the processor 41, in the process of obtaining the first affinity of the currently traversed medical item to the extracted diagnosis and the second affinity of the extracted diagnosis to the currently traversed medical item, may:

[0094] From a specified range of historical medical records, extract diagnoses, medical procedures, and the relationships between diagnoses and medical procedures. Among these, diagnoses and medical procedures appearing in the same historical medical record have a relationship.

[0095] Based on the association relationship, the affinity of each diagnosis to each medical item and the affinity of each medical item to each diagnosis are calculated separately to generate the first affinity and the second affinity.

[0096] In an alternative embodiment, the processor 41, in the process of calculating the first affinity and the second affinity based on the adjoint relationship, can be used to:

[0097] Based on the association relationship, the percentage of visits that use the currently traversed medical items and are diagnosed with the target diagnosis is calculated out of the total number of visits corresponding to the target diagnosis, which is used as the first affinity of the currently traversed medical items to the target diagnosis.

[0098] Based on the association relationship, the percentage of visits that use the currently traversed medical items and are diagnosed as the target diagnosis is calculated as the second affinity of the target diagnosis to the currently traversed medical items.

[0099] The target diagnosis is any one of the diagnoses contained in the target medical record.

[0100] In an optional embodiment, when the processor 41 determines that the currently traversed medical item belongs to an abnormal medical item if the first affinity and the second affinity do not meet the preset conditions, it can be used to:

[0101] If the target medical record contains multiple diagnoses, and the first affinity and second affinity corresponding to each diagnosis in the target medical record do not meet the preset conditions, then the medical item being traversed is determined to be an abnormal medical item.

[0102] Otherwise, determine that the medical item being traversed is not an abnormal medical item.

[0103] In an optional embodiment, the processor 41 is further configured to:

[0104] The target is to determine the percentage of abnormal medical items in the patient's medical records.

[0105] If the proportion exceeds the specified standard, the target medical record will be identified as an abnormal medical record.

[0106] In an optional embodiment, the processor 41 is further configured to:

[0107] Determine the standard name corresponding to the medical item being traversed;

[0108] If there is no pre-defined medical item in the medical knowledge graph that matches the standard name, then it is determined that the medical item being traversed does not appear in the medical knowledge graph.

[0109] In an optional embodiment, the processor 41 is further configured to:

[0110] Calculate the similarity between the standard name of the currently traversed medical item and the name of the pre-defined medical items contained in the medical knowledge graph;

[0111] If there are candidate pre-built medical items with a similarity of not less than the first specified threshold, then the pre-built medical items that do not meet the name consistency requirements will be removed from the candidate pre-built medical items.

[0112] If all candidate pre-defined medical items are removed, it is determined that there are no pre-defined medical items in the medical knowledge graph that match the standard name.

[0113] In one alternative embodiment, the name consistency requirement includes one or more of the following: part / organ / tissue consistency, positive / negative consistency, negative consistency, object consistency, action consistency, or morphological consistency.

[0114] In an optional embodiment, the processor 41 is further configured to:

[0115] If not all of the candidate pre-set medical items are eliminated, then the pre-set medical items that match the standard name are determined from the remaining candidate pre-set medical items and used as the matching medical items.

[0116] Search the medical knowledge graph for the diagnoses associated with each matching medical item, and use them as relevant diagnoses;

[0117] If there is no identical diagnosis between the diagnosis extracted from the target medical record and related diagnoses, it is determined that the diagnosis associated with the currently traversed medical item in the medical knowledge graph does not appear in the target medical record.

[0118] In an alternative embodiment, the processor 41, in the process of determining a preset medical item that matches the standard name from the remaining candidate preset medical items as a matching medical item, can be used to:

[0119] Select the one with the highest similarity from the remaining pre-selected medical services as the matching medical service; or...

[0120] Use the remaining pre-selected medical items as the matching medical items; or...

[0121] Select the K most similar medical items from the remaining pre-selected medical items as matching medical items, where K is an integer greater than 1.

[0122] In an optional embodiment, the processor 41 is further configured to:

[0123] Calculate the similarity between the diagnoses extracted from the target medical records and related diagnoses;

[0124] If there are candidate diagnoses with a similarity of not less than the second preset value among the relevant diagnoses, then the diagnoses that do not meet the name consistency requirements will be removed from the candidate diagnoses.

[0125] If all candidate diagnoses are eliminated, it is determined that there is no identical diagnosis between the diagnoses extracted from the target medical record and the relevant diagnoses.

[0126] In one optional embodiment, the process of generating the standard name includes:

[0127] Extract medical service names from a specified range of historical medical records;

[0128] Calculate the similarity between medical item names pairwise;

[0129] Clustering of medical project names based on similarity yields multiple independent sets of medical project names.

[0130] The names of various medical projects in the same set of medical project names are normalized into a unified standard name.

[0131] In an optional embodiment, similarity is calculated using the longest common subsequence, Jaccard similarity, and edit distance similarity.

[0132] Furthermore, such as Figure 3 As shown, the computing device also includes other components such as a communication component 42 and a power supply component 43. Figure 4 The diagram only shows some components and does not mean that the computing device includes only these components. Figure 4 The components shown.

[0133] It is worth noting that the technical details of the above-mentioned embodiments of the computing device can be referred to the relevant descriptions in the foregoing method embodiments. To save space, they will not be repeated here, but this should not cause any loss to the scope of protection of this application.

[0134] Accordingly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed, can perform the steps that can be executed by a computing device in the above method embodiments.

[0135] The above Figure 4 The memory in a computer is used to store computer programs and can be configured to store various other data to support operation on a computing platform. Examples of this data include instructions for any application or method operating on the computing platform, contact data, phone book data, messages, pictures, videos, etc. The memory can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disks, or optical disks.

[0136] The above Figure 4The communication component is configured to facilitate wired or wireless communication between the device containing the communication component and other devices. The device containing the communication component can access wireless networks based on communication standards, such as WiFi, 2G, 3G, 4G / LTE, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication component further includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID), Infrared Data Association (IrDA) technology, Ultra-Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0137] The above Figure 4 The power supply component provides power to the various components of the device in which it resides. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which it resides.

[0138] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0139] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0140] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0141] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0142] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0143] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0144] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0145] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0146] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for identifying abnormal medical behavior, comprising: Extract diagnoses and medical items from the target medical records; Iterate through the medical items contained in the target medical records; If the currently traversed medical item does not appear in the preset medical knowledge graph, or if the diagnosis associated with the currently traversed medical item in the medical knowledge graph does not appear in the target medical record, then the accompanying relationship between the diagnosis and the medical item is extracted from the historical medical records within the specified range, wherein the diagnosis and medical item appearing in the same historical medical record have an accompanying relationship. Based on the aforementioned association relationship, the percentage of visits to medical items currently being traversed and diagnosed as the target diagnosis is calculated as the percentage of visits to medical items currently being traversed that correspond to the target diagnosis in the total number of visits to the target diagnosis. This percentage is used as the first affinity of the medical items currently being traversed to the extracted diagnosis. Based on the association relationship, the percentage of visits that use the currently traversed medical items and are diagnosed with the target diagnosis is calculated as the percentage of the total number of visits using the currently traversed medical items, which is used as the second affinity of the extracted diagnosis to the currently traversed medical items. If the first affinity is not lower than the first preset threshold and the second affinity is not lower than the second threshold, the currently traversed medical item is determined to be an abnormal medical item.

2. The method according to claim 1, wherein determining that the currently traversed medical item belongs to an abnormal medical item when the first affinity and the second affinity do not meet a preset condition includes: If the target medical record contains multiple diagnoses, and if neither the first affinity nor the second affinity corresponding to each diagnosis in the target medical record meets the preset condition, then the medical item being traversed is determined to be an abnormal medical item. Otherwise, it is determined that the currently traversed medical item is not an abnormal medical item.

3. The method according to claim 1, further comprising: The percentage of abnormal medical items in the target medical records is calculated. If the percentage exceeds the specified standard, the target medical record is determined to be an abnormal medical record.

4. The method according to claim 1, further comprising: Determine the standard name corresponding to the currently traversed medical item; If there is no pre-defined medical item in the medical knowledge graph that matches the standard name, then it is determined that the currently traversed medical item does not appear in the medical knowledge graph.

5. The method according to claim 4, further comprising: Calculate the similarity between the standard name corresponding to the currently traversed medical item and the name of the pre-set medical item contained in the medical knowledge graph; If there are candidate pre-built medical items with a similarity of not less than the first specified threshold, then the pre-built medical items that do not meet the name consistency requirements will be removed from the candidate pre-built medical items. If all candidate pre-set medical items are eliminated, it is determined that there are no pre-set medical items in the medical knowledge graph that match the standard name.

6. The method according to claim 5, wherein the name consistency requirement includes one or more of the following: consistency of location / organ / tissue, positive / negative consistency, negative consistency, object consistency, action consistency, or morphological consistency.

7. The method according to claim 5, further comprising: If not all of the candidate pre-set medical items are eliminated, then the pre-set medical items that match the standard name are determined from the remaining candidate pre-set medical items and used as the matching medical items. The medical knowledge graph is used to find the diagnosis associated with each matching medical item, which is then used as the relevant diagnosis. If there is no identical diagnosis between the diagnosis extracted from the target medical record and the related diagnosis, then it is determined that the diagnosis associated with the currently traversed medical item in the medical knowledge graph does not appear in the target medical record. Among them, the pre-set medical items that match the standard name are determined from the remaining candidate pre-set medical items, and these are used as the matching medical items, including: Select the one with the highest similarity from the remaining pre-selected medical services as the matching medical service; or... Use the remaining pre-selected medical items as the matching medical items; or... Select the K most similar medical items from the remaining pre-selected medical items as matching medical items, where K is an integer greater than 1.

8. The method according to claim 7, further comprising: Calculate the similarity between the diagnoses extracted from the target medical records and the related diagnoses; If there are candidate diagnoses among the relevant diagnoses with a similarity not lower than the second preset value, then diagnoses that do not meet the name consistency requirement are removed from the candidate diagnoses. If all candidate diagnoses are eliminated, it is determined that there is no identical diagnosis between the diagnoses extracted from the target medical record and the relevant diagnoses.

9. The method according to claim 4, wherein the process of generating the standard name includes: Extract medical service names from a specified range of historical medical records; Calculate the similarity between medical item names pairwise; Clustering of medical project names based on similarity yields multiple independent sets of medical project names. The names of various medical projects in the same set of medical project names are normalized into a unified standard name.

10. The method according to any one of claims 5, 8 or 9, wherein the similarity is calculated using one or more of the following methods: longest common subsequence, Jaccard similarity, and edit distance similarity.

11. A computing device, comprising a memory and a processor; The memory is used to store one or more computer instructions; The processor is coupled to the memory and is used to execute one or more computer instructions to perform the abnormal medical behavior recognition method according to any one of claims 1-10.

12. A computer-readable storage medium storing computer instructions that, when executed by one or more processors, cause the one or more processors to perform the abnormal medical behavior recognition method according to any one of claims 1-10.

Images