Similar case matching system and method based on historical medical record information

Abstract

Disclosed is a similar case matching system and method based on historical medical record information. First, the sequence of the first medical record attribute item text description obtained by dividing the first medical record information and the second medical record information is semantically encoded to obtain a sequence of the first medical record attribute item text description semantic understanding feature vector and a sequence of the second medical record attribute item text description semantic understanding feature vector. Then, the semantic consistency factor between each group of corresponding two feature vectors is calculated to obtain a medical record attribute granularity comparison vector composed of multiple semantic consistency factors. Then, the medical record attribute granularity comparison vector is fully connected encoded to obtain a medical record attribute granularity comparison weight feature. Finally, based on the medical record attribute granularity comparison weight feature, the matching level label between the first medical record information and the second medical record information is determined. In this way, doctors can improve the efficiency of diagnosis and treatment.

Description

Technical Field

[0001] The present application relates to the field of case matching, and more specifically, to a similar case matching system and method based on historical medical record information. Background Art

[0002] The explosive growth and complexity of information in the medical field make it difficult for doctors to quickly and accurately find relevant cases for reference and study in their daily work. Traditional retrieval methods based on keyword matching often cannot meet doctors' needs for accurate matching, because keyword matching can only consider the surface information of the case and cannot capture the deep similarities between cases.

[0003] Therefore, a similar case matching system based on historical medical record information is desired. Summary of the invention

[0004] In view of this, the present application proposes a similar case matching system and method based on historical medical record information, which can help doctors improve the efficiency of diagnosis and treatment, and also promote the dissemination and updating of medical knowledge.

[0005] According to one aspect of the present application, a similar case matching system based on historical medical record information is provided, which includes:

[0006] A medical record information collection module, used to obtain first medical record information and second medical record information;

[0007] An attribute division module, used for dividing the first medical record information and the second medical record information according to the attributes to obtain a sequence of text descriptions of the first medical record attribute items and a sequence of text descriptions of the second medical record attribute items;

[0008] A medical record attribute item semantic understanding module, used to semantically encode the sequence of the first medical record attribute item text description and the sequence of the second medical record attribute item text description to obtain a sequence of semantic understanding feature vectors of the first medical record attribute item text description and a sequence of semantic understanding feature vectors of the second medical record attribute item text description;

[0009] The medical record attribute granularity comparison module is used to calculate the semantic consistency factor between each group of the first medical record attribute item text description semantic understanding feature vector and the second medical record attribute item text description semantic understanding feature vector in the sequence of the first medical record attribute item text description semantic understanding feature vector and the sequence of the second medical record attribute item text description semantic understanding feature vector to obtain the medical record attribute granularity comparison vector composed of multiple semantic consistency factors;

[0010] A fully connected encoding module for attribute granularity between medical records, used for performing fully connected encoding on the attribute granularity comparison vector between medical records to obtain a weight feature for attribute granularity comparison between medical records; and

[0011] The medical record information matching module is used to determine the matching level label between the first medical record information and the second medical record information based on the attribute granularity comparison weight feature between the medical records.

[0012] According to another aspect of the present application, a similar case matching method based on historical medical record information is provided, which includes:

[0013] Obtaining first medical record information and second medical record information;

[0014] Dividing the first medical record information and the second medical record information according to attributes to obtain a sequence of text descriptions of first medical record attribute items and a sequence of text descriptions of second medical record attribute items;

[0015] Semantically encoding the sequence of the first medical record attribute item text description and the sequence of the second medical record attribute item text description to obtain a sequence of semantic understanding feature vectors of the first medical record attribute item text description and a sequence of semantic understanding feature vectors of the second medical record attribute item text description;

[0016] Calculate the semantic consistency factor between each group of the first medical record attribute item text description semantic understanding feature vector and the second medical record attribute item text description semantic understanding feature vector in the sequence of the first medical record attribute item text description semantic understanding feature vector and the sequence of the second medical record attribute item text description semantic understanding feature vector to obtain an inter-medical record attribute granularity comparison vector composed of multiple semantic consistency factors;

[0017] Performing full connection encoding on the attribute granularity comparison vector between medical records to obtain attribute granularity comparison weight features between medical records; and

[0018] Based on the attribute granularity comparison weight feature between the medical records, a matching level label between the first medical record information and the second medical record information is determined.

[0019] According to an embodiment of the present application, firstly, the sequence of the text description of the first medical record attribute item and the sequence of the text description of the second medical record attribute item obtained by dividing the first medical record information and the second medical record information are semantically encoded to obtain a sequence of the semantic understanding feature vector of the text description of the first medical record attribute item and a sequence of the semantic understanding feature vector of the text description of the second medical record attribute item, then, the semantic consistency factor between each group of corresponding two feature vectors is calculated to obtain a medical record attribute granularity comparison vector composed of multiple semantic consistency factors, then, the medical record attribute granularity comparison vector is fully connected encoded to obtain a medical record attribute granularity comparison weight feature, and finally, based on the medical record attribute granularity comparison weight feature, the matching level label between the first medical record information and the second medical record information is determined. In this way, doctors can improve the efficiency of diagnosis and treatment.

[0020] Other features and aspects of the present application will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the present application and, together with the description, serve to explain the principles of the present application.

[0022] Figure 1 A block diagram of a similar case matching system based on historical medical record information according to an embodiment of the present application is shown.

[0023] Figure 2 A flowchart of a similar case matching method based on historical medical record information according to an embodiment of the present application is shown.

[0024] Figure 3 A schematic diagram of the architecture of a similar case matching method based on historical medical record information according to an embodiment of the present application is shown.

[0025] Figure 4 An application scenario diagram of a similar case matching system based on historical medical record information according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0027] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0028] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0029] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0030] In view of the above technical problems, the technical solution of this application proposes a similar case matching system based on historical medical record information, which can retrieve similar cases from a massive case database based on the patient's condition and historical medical records for doctors to refer to and learn from. The system can help doctors improve the efficiency of diagnosis and treatment, and can also promote the dissemination and updating of medical knowledge.

[0031] Specifically, the technical concept of the present application is to collect the medical record information of the detected patient and the medical record information in the database in the process of similarity detection and matching, and introduce data processing and semantic understanding algorithms in the back end to perform semantic analysis of the medical record information of the two and text semantic matching of each attribute item. In this way, multiple attributes in the case, such as age, gender, symptoms, diagnosis, treatment, etc., as well as the weights between the attributes, can be comprehensively considered to obtain a comprehensive similarity, so as to facilitate more accurate matching and retrieval of medical record information. In this way, similar cases can be automatically retrieved from the database based on the patient's condition and historical medical records for doctors to refer to and learn from. In this way, similar cases can be matched more accurately, providing more valuable reference and learning resources, and helping doctors improve diagnostic efficiency.

[0032] Figure 1 FIG. 1 is a block diagram of a similar case matching system based on historical medical record information according to an embodiment of the present application. Figure 1As shown, according to the embodiment of the present application, the similar case matching system 100 based on historical medical record information includes: a medical record information acquisition module 110, used to obtain the first medical record information and the second medical record information; an attribute division module 120, used to divide the first medical record information and the second medical record information according to the attributes to obtain a sequence of text descriptions of the first medical record attribute items and a sequence of text descriptions of the second medical record attribute items; a medical record attribute item semantic understanding module 130, used to semantically encode the sequence of text descriptions of the first medical record attribute items and the sequence of text descriptions of the second medical record attribute items, respectively, to obtain a sequence of semantic understanding feature vectors of the first medical record attribute item text description and a sequence of semantic understanding feature vectors of the second medical record attribute item text description; an attribute granularity comparison module between medical records 140, used to calculate the semantic consistency factor between each group of the first medical record attribute item text description semantic understanding feature vector and the second medical record attribute item text description semantic understanding feature vector in the sequence of the first medical record attribute item text description semantic understanding feature vector and the sequence of the second medical record attribute item text description semantic understanding feature vector to obtain an attribute granularity comparison vector between medical records composed of multiple semantic consistency factors; an attribute granularity comparison vector between medical records fully connected encoding module 150, used to perform fully connected encoding on the attribute granularity comparison vector between medical records to obtain an attribute granularity comparison weight feature between medical records; and a medical record information matching module 160, used to determine the matching level label between the first medical record information and the second medical record information based on the attribute granularity comparison weight feature between medical records.

[0033] More specifically, in the technical solution of the present application, first, the first medical record information and the second medical record information are obtained. Next, considering that there are multiple attribute items in the first medical record information and the second medical record information, such as age, gender, symptoms, diagnosis, treatment, etc., these data items not only have their own semantic information, but also each data item in each medical record information has a mutual correlation relationship. Therefore, in order to more accurately perform semantic matching of the medical record information of the two, it is necessary to divide the first medical record information and the second medical record information according to the attributes to obtain a sequence of text descriptions of the first medical record attribute items and a sequence of text descriptions of the second medical record attribute items. In this way, semantic analysis of the medical record information of the two based on attribute granularity can be performed, which is conducive to subsequent semantic comparison analysis and matching based on attribute granularity.

[0034] Then, the sequence of the text descriptions of the first medical record attribute items is semantically encoded through a semantic encoder including a word embedding layer, so as to extract the semantic understanding feature information of each of the text descriptions of the first medical record attribute items, thereby obtaining a sequence of semantic understanding feature vectors of the text descriptions of the first medical record attribute items. Furthermore, the sequence of the text descriptions of the second medical record attribute items is semantically encoded through the semantic encoder including a word embedding layer, so as to extract the semantic understanding feature information of each of the text descriptions of the second medical record attribute items, thereby obtaining a sequence of semantic understanding feature vectors of the text descriptions of the second medical record attribute items.

[0035] Correspondingly, the medical record attribute item semantic understanding module 130 includes: a first semantic encoding unit, used to pass the sequence of the first medical record attribute item text description through a semantic encoder including a word embedding layer to obtain a sequence of semantic understanding feature vectors of the first medical record attribute item text description; and a second semantic encoding unit, used to pass the sequence of the second medical record attribute item text description through the semantic encoder including a word embedding layer to obtain a sequence of semantic understanding feature vectors of the second medical record attribute item text description.

[0036] It is worth mentioning that the word embedding layer is a technology in natural language processing (NLP) that is used to map words in a text into a continuous vector space. Its purpose is to convert discrete word representations into continuous vector representations so that computers can better understand and process text data. The role of the word embedding layer is to map each word to a fixed-length vector in a high-dimensional vector space by learning the semantic and grammatical relationships between words. The advantage of this is that similar words are also close in distance in the vector space, so that the computer can obtain semantic information between words by calculating the distance or similarity between vectors. The word embedding layer is usually part of a neural network model, after the input layer of the model, and is used to convert text data into a vector representation. It can be trained based on a given corpus, or a pre-trained word embedding model such as Word2Vec, GloVe or BERT can be used. In the medical record attribute item semantic understanding module, the word embedding layer is used to convert the text description in the medical record attribute item into a semantic understanding feature vector. By mapping each word in the text sequence to the vector representation of the word embedding layer, the semantic relationship and contextual information between words can be captured. In this way, the module can better understand the meaning of medical record attribute items and further perform subsequent semantic processing and analysis.

[0037] Furthermore, in order to measure the semantic similarity between two medical record information at the attribute granularity, in the technical solution of the present application, the semantic consistency factor between each group of corresponding first medical record attribute item text description semantic understanding feature vectors and second medical record attribute item text description semantic understanding feature vectors in the sequence of the first medical record attribute item text description semantic understanding feature vectors and the sequence of the second medical record attribute item text description semantic understanding feature vectors is further calculated to obtain an attribute granularity comparison vector between medical records composed of multiple semantic consistency factors. By calculating the consistency factor between the semantics of each group of corresponding medical record attribute item text descriptions, the semantic similarities and differences of the medical record information of the two at the attribute granularity can be captured for subsequent matching and sorting of similar cases. In this way, the semantic similarities between multiple attributes in the medical record can be considered more comprehensively, thereby improving the accuracy and effect of similar case matching.

[0038] Correspondingly, the medical record attribute granularity comparison module 140 is used to calculate the semantic consistency factor between each group of the first medical record attribute item text description semantic understanding feature vector and the second medical record attribute item text description semantic understanding feature vector in the sequence of the first medical record attribute item text description semantic understanding feature vector and the sequence of the second medical record attribute item text description semantic understanding feature vector using the following correlation calculation formula to obtain the medical record attribute granularity comparison vector composed of multiple semantic consistency factors; wherein the correlation calculation formula is: in, and are the first medical record attribute item text description semantic understanding feature vector sequence The first medical record attribute item text description semantic understanding feature vector and the second medical record attribute item text description semantic understanding feature vector in the sequence corresponding to the The feature values ​​of each position of the second medical record attribute item text description semantic understanding feature vector, It is the The semantic understanding feature vector of the text description of the first medical record attribute item and the corresponding The scale of the semantic understanding feature vector of the second medical record attribute item text description, represents the logarithmic function with base 2, It is the eigenvalue of each position in the attribute granularity comparison vector between the medical records.

[0039] In the similar case matching system, since the attribute granularity comparison vector between medical records is obtained by calculating the semantic consistency factor of the medical record information of both based on the attribute granularity, it reflects the similarity of the medical records in attribute granularity. However, using only the attribute granularity comparison vector cannot fully express the similarity between medical records, because there may be complex associations and interactions between different attributes. Therefore, in order to further improve the matching accuracy of medical record information, it is necessary to perform association analysis on the semantic consistency between the text descriptions of these medical record attribute items. Based on this, in the technical solution of the present application, the attribute granularity comparison vector between medical records is further passed through a weight generator based on a fully connected layer to obtain an attribute granularity comparison weight feature vector between medical records. It should be understood that by processing with the weight generator based on a fully connected layer, the complex relationships between the attribute items of the medical record information can be learned, and these relationships can be encoded into the generated attribute granularity comparison feature vector between medical records. In particular, considering that each attribute item has different importance in the process of semantic comparison analysis and medical record information matching, by introducing the weight generator, it is possible to perform weighted association analysis of each attribute item according to the input attribute text description semantics to generate a more representative feature vector. This feature vector can capture richer medical record similarity information, including the association, interaction and importance between attributes.

[0040] Correspondingly, the medical record attribute granularity fully connected encoding module 150 is used to: pass the medical record attribute granularity comparison vector through a weight generator based on a fully connected layer to obtain a medical record attribute granularity comparison weight feature vector as the medical record attribute granularity comparison weight feature.

[0041] Then, the attribute granularity comparison weight feature vector between the medical records is passed through a classifier to obtain a classification result, and the classification result is used to represent the matching level label between the first medical record information and the second medical record information. In other words, the weighted semantic full connection association feature information based on the attribute items between the first medical record information and the second medical record information is used for classification processing to determine the matching level between the first medical record information and the second medical record information. In this way, multiple attributes in the case, such as age, gender, symptoms, diagnosis, treatment, etc., as well as the weights between the attributes, can be comprehensively considered, which is conducive to more accurate matching and retrieval of medical record information.

[0042] Correspondingly, the medical record information matching module 160 is used to: pass the attribute granularity comparison weight feature vector between the medical records through a classifier to obtain a classification result, and the classification result is used to represent the matching level label between the first medical record information and the second medical record information.

[0043] Specifically, the medical record information matching module 160 is further used to: use the fully connected layer of the classifier to fully connect the attribute granularity comparison weight feature vector between the medical records to obtain a coded classification feature vector; and input the coded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0044] It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression, SVM, etc. are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-classification, but this is prone to errors and inefficient. Commonly used multi-classification methods include the Softmax classification function.

[0045] Furthermore, in the technical solution of the present application, the similar case matching system based on historical medical record information also includes a training module for training the semantic encoder including the word embedding layer, the weight generator based on the fully connected layer, and the classifier.

[0046] In one example, the training module includes: a training data acquisition unit, used to obtain training data, wherein the training data includes the first training medical record information, the second training medical record information, and the true value of the matching level label between the first training medical record information and the second training medical record information; a training attribute division unit, used to divide the first training medical record information and the second training medical record information according to attributes to obtain a sequence of text descriptions of the first training medical record attribute items and a sequence of text descriptions of the second training medical record attribute items; a training medical record attribute item semantic understanding unit, used to respectively pass the sequence of text descriptions of the first training medical record attribute items and the sequence of text descriptions of the second training medical record attribute items through the semantic encoder including the word embedding layer to obtain a sequence of semantic understanding feature vectors of the text descriptions of the first training medical record attribute items and a sequence of semantic understanding feature vectors of the text descriptions of the second training medical record attribute items; an attribute granularity comparison unit between training medical records, used to calculate each group of pairs in the sequence of semantic understanding feature vectors of the text descriptions of the first training medical record attribute items and the sequence of semantic understanding feature vectors of the text descriptions of the second training medical record attribute items A semantic consistency factor between the semantic understanding feature vector of the text description of a first training medical record attribute item and the semantic understanding feature vector of the text description of a second training medical record attribute item is used to obtain a training medical record attribute granularity comparison vector composed of multiple semantic consistency factors; a training medical record attribute granularity fully connected encoding unit is used to pass the training medical record attribute granularity comparison vector through the weight generator based on the fully connected layer to obtain a training medical record attribute granularity comparison weight feature vector; a classification loss function value calculation unit is used to pass the training medical record attribute granularity comparison weight feature vector through the classifier to obtain a classification loss function value; a specific loss function value calculation unit is used to calculate the specific loss function value of the sequence of the first training medical record attribute item text description semantic understanding feature vector and the sequence of the second training medical record attribute item text description semantic understanding feature vector; and a loss training unit is used to train the semantic encoder including the word embedding layer, the weight generator based on the fully connected layer and the classifier with the weighted sum of the classification loss function value and the specific loss function value as the loss function value.

[0047] The classification loss function value calculation unit is used to process the attribute granularity comparison weight feature vector between the training medical records through the classifier using the following classification training formula to obtain a training classification result. The classification training formula is: ,in, arrive is the weight matrix, arrive is the bias vector, The weight feature vector of the attribute granularity comparison between the training medical records; and the cross entropy value between the training classification result and the true value is calculated as the classification loss function value.

[0048] In particular, in the above technical scheme, the sequence of semantic understanding feature vectors of the text description of the first training medical record attribute item and the sequence of semantic understanding feature vectors of the text description of the second training medical record attribute item respectively express the encoded text semantic features of the text description of the first training medical record attribute item and the second training medical record attribute item. Here, considering the text attribute differences and source text semantic differences of the first training medical record information and the second training medical record information themselves, the obtained sequence of semantic understanding feature vectors of the text description of the first training medical record attribute item and the sequence of semantic understanding feature vectors of the text description of the second training medical record attribute item will also have differences in the feature group density representation on the distribution dimension of the overall feature vector, thereby affecting the calculation accuracy of the semantic consistency factor between each group of the first training medical record attribute item text description semantic understanding feature vector and the second training medical record attribute item text description semantic understanding feature vector in the sequence of the first training medical record attribute item text description semantic understanding feature vector and the second training medical record attribute item text description semantic understanding feature vector.

[0049] Therefore, in order to improve the consistency of the feature group density representation of the first training medical record attribute item text description semantic understanding feature vector sequence and the second training medical record attribute item text description semantic understanding feature vector sequence, the applicant of the present application further introduced a specific loss function for the first training medical record attribute item text description semantic understanding feature vector sequence and the second training medical record attribute item text description semantic understanding feature vector sequence.

[0050] Accordingly, in one example, the specific loss function value calculation unit is used to calculate the specific loss function value of the sequence of the first training medical record attribute item text description semantic understanding feature vector and the sequence of the second training medical record attribute item text description semantic understanding feature vector using the following specific loss function value calculation formula; wherein the specific loss function value calculation formula is: in, and are respectively a first cascade feature vector and a second cascade feature vector obtained by cascading a sequence of the first training medical record attribute item text description semantic understanding feature vectors and a sequence of the second training medical record attribute item text description semantic understanding feature vectors, is the length of the eigenvector, and represents the square of the vector's two-norm, represents the first cascade feature vector eigenvalues, represents the second concatenated feature vector eigenvalues, represents vector subtraction, represents an exponential operation of a value, wherein the exponential operation of the value represents calculating the value of a natural exponential function with the value as a power, is the value of a specific loss function.

[0051] Here, the loss function performs group count attention based on feature group density, and performs adaptive attention to different density representation patterns between the sequence of the first training medical record attribute item text description semantic understanding feature vector and the sequence of the second training medical record attribute item text description semantic understanding feature vector by using group count as a recursive mapping of the output feature group density. By using it as a loss function to train the model, the model can avoid overestimation and underestimation for the different density patterns under the feature distribution of the sequence of the first training medical record attribute item text description semantic understanding feature vector and the sequence of the second training medical record attribute item text description semantic understanding feature vector, and learn the corresponding relationship between the feature value distribution and the group density distribution, thereby achieving the semantic distribution consistency optimization between the sequence of the first training medical record attribute item text description semantic understanding feature vector and the sequence of the second training medical record attribute item text description semantic understanding feature vector with different feature densities. In this way, the calculation accuracy of the semantic consistency factor is improved, thereby improving the expression effect of the medical record attribute granularity comparison feature vector obtained from the medical record attribute granularity comparison vector composed of multiple semantic consistency factors, and improving the accuracy of the classification result obtained by the classifier. In this way, similar cases can be automatically retrieved from the database based on the patient's condition and historical medical records for doctors to refer to and learn from. In this way, similar cases can be matched more accurately, thereby providing more valuable reference and learning resources, helping doctors improve diagnostic efficiency.

[0052] In summary, the similar case matching system 100 based on historical medical record information according to the embodiment of the present application is explained, which can help doctors improve the efficiency of diagnosis and treatment, and also promote the dissemination and updating of medical knowledge.

[0053] As described above, the similar case matching system 100 based on historical medical record information according to the embodiment of the present application can be implemented in various terminal devices, such as a server with a similar case matching algorithm based on historical medical record information. In one example, the similar case matching system 100 based on historical medical record information can be integrated into the terminal device as a software module and / or a hardware module. For example, the similar case matching system 100 based on historical medical record information can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the similar case matching system 100 based on historical medical record information can also be one of the many hardware modules of the terminal device.

[0054] Alternatively, in another example, the similar case matching system 100 based on historical medical record information and the terminal device may also be separate devices, and the similar case matching system 100 based on historical medical record information may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0055] Figure 2 A flowchart of a similar case matching method based on historical medical record information according to an embodiment of the present application is shown. Figure 3 A schematic diagram showing the system architecture of a similar case matching method based on historical medical record information according to an embodiment of the present application. Figure 2 and Figure 3 As shown, according to the embodiment of the present application, the similar case matching method based on historical medical record information includes: S110, obtaining first medical record information and second medical record information; S120, dividing the first medical record information and the second medical record information according to attributes to obtain a sequence of text descriptions of the first medical record attribute items and a sequence of text descriptions of the second medical record attribute items; S130, semantically encoding the sequence of text descriptions of the first medical record attribute items and the sequence of text descriptions of the second medical record attribute items, respectively, to obtain a sequence of semantic understanding feature vectors of the text descriptions of the first medical record attribute items and a sequence of semantic understanding feature vectors of the text descriptions of the second medical record attribute items; S140, calculating the The sequence of semantic understanding feature vectors of the text description of the first medical record attribute item and the sequence of semantic understanding feature vectors of the text description of the second medical record attribute item correspond to the semantic consistency factor between the semantic understanding feature vectors of the text description of the first medical record attribute item and the semantic understanding feature vectors of the text description of the second medical record attribute item to obtain a medical record attribute granularity comparison vector composed of multiple semantic consistency factors; S150, fully connected encoding is performed on the medical record attribute granularity comparison vector to obtain a medical record attribute granularity comparison weight feature; and, S160, based on the medical record attribute granularity comparison weight feature, a matching level label between the first medical record information and the second medical record information is determined.

[0056] In a possible implementation, the sequence of text descriptions of the first medical record attribute items and the sequence of text descriptions of the second medical record attribute items are semantically encoded respectively to obtain a sequence of semantic understanding feature vectors of the first medical record attribute item text descriptions and a sequence of semantic understanding feature vectors of the second medical record attribute item text descriptions, including: passing the sequence of text descriptions of the first medical record attribute item through a semantic encoder including a word embedding layer to obtain a sequence of semantic understanding feature vectors of the first medical record attribute item text descriptions; and passing the sequence of text descriptions of the second medical record attribute item through the semantic encoder including a word embedding layer to obtain a sequence of semantic understanding feature vectors of the second medical record attribute item text descriptions.

[0057] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned similar case matching method based on historical medical record information have been referred to above. Figure 1 The similar case matching system based on historical medical record information has been introduced in detail, and therefore, its repeated description will be omitted.

[0058] Figure 4 FIG. 1 shows an application scenario diagram of a similar case matching system based on historical medical record information according to an embodiment of the present application. Figure 4 As shown, in this application scenario, first, obtain the first medical record information (for example, Figure 4 D1 as shown in the figure) and the second medical record information (e.g. Figure 4 Then, the first medical record information and the second medical record information are input into a server (for example, Figure 4 S) as shown in , wherein the server is able to use the similar case matching algorithm based on historical medical record information to process the first medical record information and the second medical record information to obtain a classification result for representing a matching level label between the first medical record information and the second medical record information.

[0059] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can also be executed in the opposite order sometimes, depending on the function involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be realized by a dedicated hardware-based system that performs the function or action of the specification, or can be realized by a combination of special-purpose hardware and computer instructions.

[0060] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A similar case matching system based on historical medical record information, characterized in that: include: A medical record information collection module, used to obtain first medical record information and second medical record information; An attribute division module, used for dividing the first medical record information and the second medical record information according to the attributes to obtain a sequence of text descriptions of the first medical record attribute items and a sequence of text descriptions of the second medical record attribute items; A medical record attribute item semantic understanding module, used to semantically encode the sequence of the first medical record attribute item text description and the sequence of the second medical record attribute item text description to obtain a sequence of semantic understanding feature vectors of the first medical record attribute item text description and a sequence of semantic understanding feature vectors of the second medical record attribute item text description; The medical record attribute granularity comparison module is used to calculate the semantic consistency factor between each group of the first medical record attribute item text description semantic understanding feature vector and the second medical record attribute item text description semantic understanding feature vector in the sequence of the first medical record attribute item text description semantic understanding feature vector and the sequence of the second medical record attribute item text description semantic understanding feature vector to obtain the medical record attribute granularity comparison vector composed of multiple semantic consistency factors; A fully connected encoding module for attribute granularity between medical records, used for performing fully connected encoding on the attribute granularity comparison vector between medical records to obtain a weight feature for attribute granularity comparison between medical records; as well as A medical record information matching module, used to determine a matching level label between the first medical record information and the second medical record information based on the attribute granularity comparison weight feature between the medical records; The medical record attribute granularity comparison module is used to calculate the semantic consistency factor between each group of the first medical record attribute item text description semantic understanding feature vector and the second medical record attribute item text description semantic understanding feature vector in the sequence of the first medical record attribute item text description semantic understanding feature vector and the sequence of the second medical record attribute item text description semantic understanding feature vector using the following correlation calculation formula to obtain the medical record attribute granularity comparison vector composed of multiple semantic consistency factors; wherein the correlation calculation formula is: ;in, and are the first medical record attribute item text description semantic understanding feature vector sequence The first medical record attribute item text description semantic understanding feature vector and the second medical record attribute item text description semantic understanding feature vector in the sequence corresponding to the The feature values ​​of each position of the second medical record attribute item text description semantic understanding feature vector, It is the The semantic understanding feature vector of the text description of the first medical record attribute item and the corresponding The scale of the semantic understanding feature vector of the second medical record attribute item text description, represents the logarithmic function with base 2, It is the eigenvalue of each position in the attribute granularity comparison vector between the medical records.

2. The similar case matching system based on historical medical record information according to claim 1 is characterized in that: The medical record attribute item semantic understanding module includes: A first semantic encoding unit, configured to pass the sequence of the first medical record attribute item text description through a semantic encoder including a word embedding layer to obtain a sequence of semantic understanding feature vectors of the first medical record attribute item text description; and The second semantic encoding unit is used to pass the sequence of the text description of the second medical record attribute item through the semantic encoder including the word embedding layer to obtain a sequence of semantic understanding feature vectors of the text description of the second medical record attribute item.

3. The similar case matching system based on historical medical record information according to claim 2 is characterized in that: The medical record attribute granularity fully connected encoding module is used to: pass the medical record attribute granularity comparison vector through a weight generator based on a fully connected layer to obtain a medical record attribute granularity comparison weight feature vector as the medical record attribute granularity comparison weight feature.

4. The similar case matching system based on historical medical record information according to claim 3 is characterized in that: The medical record information matching module is used to: pass the attribute granularity comparison weight feature vector between the medical records through a classifier to obtain a classification result, and the classification result is used to represent a matching level label between the first medical record information and the second medical record information.

5. The similar case matching system based on historical medical record information according to claim 4 is characterized in that: It also includes a training module for training the semantic encoder including the word embedding layer, the weight generator based on the fully connected layer, and the classifier.

6. The similar case matching system based on historical medical record information according to claim 5 is characterized in that: The training module comprises: A training data acquisition unit, used to acquire training data, wherein the training data includes first training medical record information, second training medical record information, and a true value of a matching level label between the first training medical record information and the second training medical record information; A training attribute division unit, configured to divide the first training medical record information and the second training medical record information according to attributes to obtain a sequence of text descriptions of first training medical record attribute items and a sequence of text descriptions of second training medical record attribute items; A training medical record attribute item semantic understanding unit, used for respectively passing the sequence of the first training medical record attribute item text description and the sequence of the second training medical record attribute item text description through the semantic encoder including the word embedding layer to obtain a sequence of the first training medical record attribute item text description semantic understanding feature vectors and a sequence of the second training medical record attribute item text description semantic understanding feature vectors; The attribute granularity comparison unit between training medical records is used to calculate the semantic consistency factor between each group of the first training medical record attribute item text description semantic understanding feature vector and the second training medical record attribute item text description semantic understanding feature vector in the sequence of the first training medical record attribute item text description semantic understanding feature vector and the sequence of the second training medical record attribute item text description semantic understanding feature vector to obtain an attribute granularity comparison vector between training medical records composed of multiple semantic consistency factors; A fully connected encoding unit for attribute granularity between training medical records, used for passing the attribute granularity comparison vector between training medical records through the weight generator based on the fully connected layer to obtain an attribute granularity comparison weight feature vector between training medical records; A classification loss function value calculation unit, used for passing the attribute granularity comparison weight feature vector between the training medical records through the classifier to obtain a classification loss function value; a specific loss function value calculation unit, used to calculate a specific loss function value of a sequence of semantic understanding feature vectors of text descriptions of the first training medical record attribute items and a sequence of semantic understanding feature vectors of text descriptions of the second training medical record attribute items; and A loss training unit is used to train the semantic encoder including the word embedding layer, the weight generator based on the fully connected layer, and the classifier by taking the weighted sum of the classification loss function value and the specific loss function value as the loss function value.

7. The similar case matching system based on historical medical record information according to claim 6 is characterized in that: The classification loss function value calculation unit is used to: The classifier processes the attribute granularity comparison weight feature vector between the training medical records using the following classification training formula to obtain a training classification result, and the classification training formula is: ;in, arrive is the weight matrix, arrive is the bias vector, is a weight feature vector for comparing the attribute granularity between the training medical records; and A cross entropy value between the training classification result and the true value is calculated as the classification loss function value.

8. A similar case matching method based on historical medical record information, using the similar case matching system based on historical medical record information according to claim 1, characterized in that: include: Obtaining first medical record information and second medical record information; Dividing the first medical record information and the second medical record information according to attributes to obtain a sequence of text descriptions of first medical record attribute items and a sequence of text descriptions of second medical record attribute items; Semantically encoding the sequence of the first medical record attribute item text description and the sequence of the second medical record attribute item text description to obtain a sequence of semantic understanding feature vectors of the first medical record attribute item text description and a sequence of semantic understanding feature vectors of the second medical record attribute item text description; Calculate the semantic consistency factor between each group of the first medical record attribute item text description semantic understanding feature vector and the second medical record attribute item text description semantic understanding feature vector in the sequence of the first medical record attribute item text description semantic understanding feature vector and the sequence of the second medical record attribute item text description semantic understanding feature vector to obtain an inter-medical record attribute granularity comparison vector composed of multiple semantic consistency factors; Performing full connection encoding on the attribute granularity comparison vector between medical records to obtain the attribute granularity comparison weight feature between medical records; as well as Based on the attribute granularity comparison weight feature between the medical records, a matching level label between the first medical record information and the second medical record information is determined.

9. The similar case matching method based on historical medical record information according to claim 8, characterized in that: Semantically encoding the sequence of the first medical record attribute item text description and the sequence of the second medical record attribute item text description to obtain a sequence of semantic understanding feature vectors of the first medical record attribute item text description and a sequence of semantic understanding feature vectors of the second medical record attribute item text description, including: Passing the sequence of the text descriptions of the first medical record attribute items through a semantic encoder including a word embedding layer to obtain a sequence of semantic understanding feature vectors of the text descriptions of the first medical record attribute items; and The sequence of the text descriptions of the second medical record attribute items is passed through the semantic encoder including the word embedding layer to obtain a sequence of semantic understanding feature vectors of the text descriptions of the second medical record attribute items.

Images