Medical record text similarity retrieval method, system and computer device

Abstract

The present invention provides a retrieval method, system and computer equipment for medical record text similarity. The retrieval method for medical record text similarity includes: receiving text information; performing word segmentation processing on the text information to generate words; training the words into long text vectors; and obtaining medical record information similar to the text information in a database according to the long text vectors. The retrieval method for medical record text similarity provided by the present invention does not require expert participation, but automatically mines and learns medical knowledge from a database through a medical artificial intelligence method to construct a model for similar medical record comparison. The model can integrate the comparison results of various types of free texts, and can efficiently and accurately obtain similar medical record recommendations, which are highly consistent with the results obtained by manual comparison by doctors, and can provide doctors with clinical pathway reference results with practical value, effectively solving the problem of doctors spending a lot of time on reviewing historical medical records.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a medical record text similarity retrieval method, system and computer equipment. Background Art

[0002] At present, Electronic Medical Record (EMR) is a medical record generated when patients visit medical institutions. It is the carrier of doctors' diagnosis and treatment experience and patterns. Its core value is reflected in auxiliary diagnosis and provides decision support for doctors. The main forms of electronic medical record data are tables, free text, and images. Among them, free text is mainly presented in the form of unstructured data. With the development of hospital informatization, hospitals have accumulated a large amount of unstructured electronic medical record free text, which contains a lot of valuable medical and clinical information. With the improvement of the standardization of medical information, free text covers more standardized and complete patient information. At present, many scholars, institutions and enterprises at home and abroad are committed to the research of auxiliary diagnosis systems based on EMR (electronic medical records). Its field can involve the entire medical process and plays an important role in optimizing work processes, improving work efficiency, reducing medical errors, and improving medical quality. Domestic application research based on Chinese EMR (electronic medical records) is on the one hand aimed at the development of EMR (electronic medical records) systems, and on the other hand, it is based on EMR (electronic medical records) Clinical pathway optimization and similar EMR (electronic medical records) search. Related technologies all use the core technology of similar Chinese medical record text retrieval, which mainly compares through keywords or ontology models and relies on the knowledge of medical experts. However, the information contained in the existing large-scale EMR (electronic medical record) data itself has not been well mined and utilized. Summary of the invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art or related art.

[0004] To this end, a first aspect of the present invention provides a method for retrieving medical record text similarity.

[0005] A second aspect of the present invention provides a medical record text similarity retrieval system.

[0006] A third aspect of the present invention provides a computer device.

[0007] A fourth aspect of the present invention provides a computer-readable storage medium.

[0008] In view of this, the first aspect of the present invention proposes a medical record text similarity retrieval method, including: receiving text information; performing word segmentation processing on the text information to generate words; training the words into long text vectors; and obtaining medical record information similar to the text information in a database based on the long text vectors.

[0009] The medical record text similarity retrieval method provided by the present invention performs word segmentation on the received text information, and the word segmentation includes word ambiguity segmentation and recognition of unregistered words. Diseases, symptoms and time can be segmented, and the segmented words are used for the next step of training. Accurate word segmentation determines the accuracy of the next step. The generated words are trained into long text vectors to obtain corresponding long text digital identifiers, and then medical record information similar to the text information is obtained in the database according to the long text vector. Retrieving medical record information by this method does not require expert participation, but automatically mines and learns medical knowledge from the database through the method of medical artificial intelligence, and constructs a model for similar medical record comparison. The model can integrate the comparison results of various types of free texts, can efficiently and accurately obtain similar medical record recommendations, and is highly consistent with the results obtained by manual comparison by doctors. It can provide doctors with clinical pathway reference results with practical value, effectively solving the problem that doctors spend a lot of time on consulting historical medical records. At the same time, this method can also be used to assist doctors who lack medical experience, so that patients can be diagnosed and treated better and in a timely manner, thereby improving clinical consultation efficiency and clinical diagnosis accuracy.

[0010] Specifically, the main processing objects of this method are the chief complaint, current medical history, past medical history, personal history, family history and general examination results in the free text, so as to obtain a complete auxiliary diagnosis for the patient.

[0011] The above-mentioned medical record text similarity retrieval method provided by the present invention may also have the following additional technical features:

[0012] In the above technical solution, preferably, the medical record text similarity retrieval method also includes: segmenting the text information, and after the step of generating words, it also includes: marking the parts of speech of the words; and classifying the words according to the marking of the parts of speech of the words.

[0013] In this technical solution, the text information is preprocessed by the named entity recognition application, the word parts of speech are annotated, and then the words are classified according to the annotations, and the correct lexical tag is assigned to each word in the sentence, and a category is assigned to each word. Furthermore, the named entity recognition application can accurately segment unregistered words, and the part-of-speech tagging is mainly divided into rule-based and statistical-based methods. Specifically, the CRF (conditional random field) algorithm is first used to tag the words in the long text segmentation with parts of speech, and the words with the annotated parts of speech are used as RNN (recurrent neural network) inputs. According to the category of the part of speech, the vocabulary classification of diseases and symptoms appearing in the long text is fed back.

[0014] In any of the above technical solutions, preferably, the step of segmenting the text information to generate words specifically includes: segmenting the text information according to a disease dictionary, regular expressions, and removing stop words to generate words.

[0015] In this technical solution, the text information is segmented according to the disease dictionary, regular expressions, and stop words, thereby achieving the effect of removing interference words. At the same time, the maximum matching method is used to improve the accuracy of word segmentation.

[0016] In any of the above technical solutions, preferably, the step of training words into long text vectors specifically includes: training words into word vectors; and composing word vectors into long text vectors.

[0017] In this technical solution, the divided words are first trained into word vectors, and then the word vectors in each sentence are combined to form a long text vector, and then the digital symbols of the long text of the medical record are obtained.

[0018] In any of the above technical solutions, preferably, the step of obtaining medical record information similar to the text information in the database according to the long text vector specifically includes: obtaining multiple long texts similar to the text information in the database, and dividing the multiple long texts into word sets as screening sets; obtaining long texts that match the word set after word segmentation processing of the text information in the screening set, and taking it as a priority result; calculating the correlation between the word set that does not match the text information in the screening set and the word set after word segmentation processing of the text information according to the long text vector; judging whether the correlation is greater than a preset threshold; if the correlation is greater than the preset threshold, arranging the long texts that do not match the text information in ascending order according to the size of the correlation.

[0019] In this technical solution, the edit distance is first used to solve the most similar positive order of EMR (electronic medical record) words, and then divided into corresponding word sets. The jaccard distance is used to calculate the long text that completely matches the text information in the word set, and its priority is set to the highest. The cosine distance is used to obtain the correlation between words for the incompletely matched long texts, and a preset threshold is set. If the correlation is less than the preset threshold, the correlation is 0, which can be considered as unrelated. The associated word distances are added and sorted in positive order to solve the secondary priority long text match. Specifically, for example, the current long text segmentation set {A, B} and a set {C, A} in the library, the weighted similarity distance obtained after the cosine distance calculation is: (B·C) / (||B||·||C||).

[0020] In a second aspect of the present invention, a medical record text similarity retrieval system is proposed, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following are achieved: receiving text information; performing word segmentation on the text information to generate words; training the words into long text vectors; and obtaining medical record information similar to the text information in a database based on the long text vectors.

[0021] The medical record text similarity retrieval system provided by the present invention performs word segmentation processing on the received text information, and the word segmentation includes word ambiguity segmentation and recognition of unregistered words. Diseases, symptoms and time can be segmented, and the segmented words are used for the next step of training. Accurate word segmentation determines the accuracy of the next step. The generated words are trained into long text vectors to obtain corresponding long text digital identifiers, and then medical record information similar to the text information is obtained in the database according to the long text vector. Through this system, medical record information is retrieved without the participation of experts, but medical knowledge is automatically mined and learned from the database through the method of medical artificial intelligence, and a model for similar medical record comparison is constructed. The model can integrate the comparison results of various types of free texts, and can efficiently and accurately obtain similar medical record recommendations, which are highly consistent with the results obtained by manual comparison by doctors, and can provide doctors with clinical pathway reference results with practical value, effectively solving the problem that doctors spend a lot of time on consulting historical medical records. At the same time, this system can also be used to assist doctors who lack medical experience, so that patients can be diagnosed and treated better and in a timely manner, thereby improving clinical consultation efficiency and clinical diagnosis accuracy.

[0022] Specifically, the main processing objects of this system are the chief complaint, current medical history, past medical history, personal history, family history and general examination results in free text, so as to obtain a complete auxiliary diagnosis for the patient.

[0023] The above-mentioned medical record text similarity retrieval system provided by the present invention may also have the following additional technical features:

[0024] In the above technical solution, preferably, when the processor executes the computer program, it also implements: segmenting the text information, and after the step of generating words, it also includes: marking the parts of speech of the words; and classifying the words according to the marking of the parts of speech of the words.

[0025] In this technical solution, the text information is preprocessed by the named entity recognition application, the word parts of speech are annotated, and then the words are classified according to the annotations, and the correct lexical tag is assigned to each word in the sentence, and a category is assigned to each word. Furthermore, the named entity recognition application can accurately segment unregistered words, and the part-of-speech tagging is mainly divided into rule-based and statistical-based methods. Specifically, the CRF (conditional random field) algorithm is first used to tag the words in the long text segmentation with parts of speech, and the words with the annotated parts of speech are used as RNN (recurrent neural network) inputs. According to the category of the part of speech, the vocabulary classification of diseases and symptoms appearing in the long text is fed back.

[0026] In any of the above technical solutions, preferably, when the processor executes the computer program, it implements the step of segmenting the text information and generating words, which specifically includes: segmenting the text information according to a disease dictionary, regular expressions, and removing stop words to generate words.

[0027] In this technical solution, the text information is segmented according to the disease dictionary, regular expressions, and stop words, thereby achieving the effect of removing interference words. At the same time, the maximum matching method is used to improve the accuracy of word segmentation.

[0028] In any of the above technical solutions, preferably, when the processor executes the computer program, the step of training words into long text vectors is implemented, specifically including: training words into word vectors; and composing word vectors into long text vectors.

[0029] In this technical solution, the divided words are first trained into word vectors, and then the word vectors in each sentence are combined to form a long text vector, and then the digital symbols of the long text of the medical record are obtained.

[0030] In any of the above technical solutions, preferably, when the processor executes the computer program, it implements the step of obtaining medical record information similar to the text information in the database according to the long text vector, specifically including: obtaining multiple long texts similar to the text information in the database, and dividing the multiple long texts into word sets as screening sets; obtaining long texts that match the word set after word segmentation processing of the text information in the screening set, and taking it as a priority result; calculating the correlation between the word set that does not match the text information in the screening set and the word set after word segmentation processing of the text information according to the long text vector; judging whether the correlation is greater than a preset threshold; if the correlation is greater than the preset threshold, arranging the long texts that do not match the text information in positive order according to the size of the correlation.

[0031] In this technical solution, the edit distance is first used to solve the most similar positive order of EMR (electronic medical record) words, and then divided into corresponding word sets. The jaccard distance is used to calculate the long text that completely matches the text information in the word set, and its priority is set to the highest. The cosine distance is used to obtain the correlation between words for the incompletely matched long texts, and a preset threshold is set. If the correlation is less than the preset threshold, the correlation is 0, which can be considered as unrelated. The associated word distances are added and sorted in positive order to solve the secondary priority long text match. Specifically, for example, the current long text segmentation set {A, B} and a set {C, A} in the library, the weighted similarity distance obtained after the cosine distance calculation is: (B·C) / (||B||·||C||).

[0032] In a third aspect of the present invention, a computer device is proposed, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for retrieving the medical record text similarity as described in any of the above technical solutions is implemented.

[0033] The technical solution proposed in the present invention includes the medical record text similarity retrieval method described in any technical solution of the first aspect, and therefore has all the beneficial effects of the medical record text similarity retrieval method.

[0034] The fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method described in any of the above technical solutions are implemented, thereby having all the technical effects of the medical record text similarity retrieval method, which will not be repeated here.

[0035] Additional aspects and advantages of the present invention will become apparent from the following description or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0037] Figure 1 A schematic diagram showing a flow chart of a method for retrieving medical record text similarity according to an embodiment of the present application;

[0038] Figure 2 Another schematic diagram showing a process of retrieving medical record text similarity according to an embodiment of the present application;

[0039] Figure 3 Another schematic diagram showing a process of retrieving medical record text similarity according to an embodiment of the present application;

[0040] Figure 4 Another schematic diagram showing a process of retrieving medical record text similarity according to an embodiment of the present application;

[0041] Figure 5 A block diagram of a medical record text similarity retrieval system according to an embodiment of the present application is shown;

[0042] Figure 6 Another block diagram of a medical record text similarity retrieval system according to an embodiment of the present application is shown;

[0043] Figure 7 Another block diagram of a medical record text similarity retrieval system according to an embodiment of the present application is shown;

[0044] Figure 8 A schematic block diagram of a computer device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0045] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0046] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0047] Refer to the following Figures 1 to 8 The present invention describes a method, system and computer device for retrieving medical record text similarity according to some embodiments of the present invention.

[0048] Figure 1 FIG. 1 is a flow chart showing a method for retrieving medical record text similarity according to an embodiment of the present application. Figure 1 As shown, the method includes:

[0049] Step 102, receiving text information;

[0050] Step 104, performing word segmentation processing on the text information to generate words;

[0051] Step 106, training the words into long text vectors;

[0052] Step 108: Obtain medical record information similar to the text information in the database according to the long text vector.

[0053] The medical record text similarity retrieval method provided by the present invention performs word segmentation on the received text information, and the word segmentation includes word ambiguity segmentation and recognition of unregistered words. Diseases, symptoms and time can be segmented, and the segmented words are used for the next step of training. Accurate word segmentation determines the accuracy of the next step. The generated words are trained into long text vectors to obtain corresponding long text digital identifiers, and then medical record information similar to the text information is obtained in the database according to the long text vector. Retrieving medical record information by this method does not require expert participation, but automatically mines and learns medical knowledge from the database through the method of medical artificial intelligence, and constructs a model for similar medical record comparison. The model can integrate the comparison results of various types of free texts, can efficiently and accurately obtain similar medical record recommendations, and is highly consistent with the results obtained by manual comparison by doctors. It can provide doctors with clinical pathway reference results with practical value, effectively solving the problem that doctors spend a lot of time on consulting historical medical records. At the same time, this method can also be used to assist doctors who lack medical experience, so that patients can be diagnosed and treated better and in a timely manner, thereby improving clinical consultation efficiency and clinical diagnosis accuracy.

[0054] Specifically, the main processing objects of this method are the chief complaint, current medical history, past medical history, personal history, family history and general examination results in the free text, so as to obtain a complete auxiliary diagnosis for the patient.

[0055] In the above embodiment, preferably, after the step of segmenting the text information to generate words, the process further includes: marking the parts of speech of the words; and classifying the words according to the marking of the parts of speech of the words.

[0056] In this embodiment, the text information is preprocessed by a named entity recognition application, the word parts of speech are annotated, and then the words are classified according to the annotations, and the correct lexical tag is assigned to each word in the sentence, and a category is assigned to each word. Further, the named entity recognition application can accurately segment unregistered words, and the part-of-speech tagging is mainly divided into rule-based and statistical-based methods. Specifically, the CRF (conditional random field) algorithm is first used to tag the words of the long text segmentation with parts of speech, and the words with the annotated parts of speech are used as RNN (recurrent neural network) inputs, and the vocabulary classification of diseases and symptoms appearing in the long text is fed back according to the category of the part of speech.

[0057] In any of the above embodiments, preferably, the step of segmenting the text information to generate words specifically includes: segmenting the text information according to a disease dictionary, regular expressions, and removing stop words to generate words.

[0058] In this embodiment, the text information is segmented according to the disease dictionary, regular expressions, and stop words are removed, thereby achieving the effect of removing interference words. At the same time, the maximum matching method is used to improve the accuracy of the segmentation.

[0059] In any of the above embodiments, preferably, the step of training words into long text vectors specifically includes: training words into word vectors; and composing word vectors into long text vectors.

[0060] In this embodiment, the divided words are first trained into word vectors, and then the word vectors in each sentence are combined to form a long text vector, thereby obtaining the digital symbols of the long text of the medical record.

[0061] In any of the above embodiments, preferably, the step of obtaining medical record information similar to the text information in the database according to the long text vector specifically includes: obtaining multiple long texts similar to the text information in the database, and dividing the multiple long texts into word sets as screening sets; obtaining long texts that match the word set after word segmentation processing of the text information in the screening set, and taking it as a priority result; calculating the correlation between the word set that does not match the text information in the screening set and the word set after word segmentation processing of the text information according to the long text vector; judging whether the correlation is greater than a preset threshold; if the correlation is greater than the preset threshold, arranging the long texts that do not match the text information in ascending order according to the size of the correlation.

[0062] In this embodiment, the edit distance is first used to solve the most similar positive order of EMR (electronic medical record) words, and then divided into corresponding word sets. The long text that completely matches the text information is calculated in the word set using the jaccard distance, and its priority is set to the highest. The incompletely matched long text is used to obtain the correlation between words using the cosine distance, and a preset threshold is set. If the correlation is less than the preset threshold, the correlation is 0, which can be considered as unrelated. The associated word distances are added and sorted in positive order to solve the secondary priority long text match. Specifically, for example, the current long text segmentation set {A, B} and a set {C, A} in the library, the weighted similarity distance obtained after the cosine distance calculation is: (B·C) / (||B||·||C||).

[0063] Figure 2 Another flow chart of a method for retrieving medical record text similarity according to an embodiment of the present application is shown. Figure 2 As shown, the method includes:

[0064] Step 202, receiving the patient's medical record information;

[0065] Step 204, segmenting the patient's medical record chief complaint information to generate words;

[0066] Step 206, training the words into long text vectors;

[0067] Step 208, filtering the search scope according to whether the disease name or specific symptom is included;

[0068] Step 210, calculating the similarity of the main complaints according to the combined distance algorithm.

[0069] In this embodiment, the received data objects are the patient's chief complaint data (text type) and disease history (numeric type). First, the similarity of the chief complaint data is calculated, such as Figure 2 As shown in the figure, based on the patient's chief complaint input by the doctor, the CRF (conditional random field) algorithm, RNN (recurrent neural network) and Doc2Vec (sentiment analysis) are used to train the chief complaint into a long text vector, and the retrieval scope is filtered according to whether the chief complaint contains the disease name or specific symptoms. The edit distance is used to narrow the retrieval scope and reduce the time complexity to achieve the effect of fast retrieval. Then, the jaccard distance and cosine distance are used to combine and calculate the similarity of the chief complaint.

[0070] Figure 3 Another flow chart of a method for retrieving medical record text similarity according to an embodiment of the present application is shown. Figure 3 As shown, the method includes:

[0071] Step 302, obtaining medical history information in the patient's medical record according to medical history statistics;

[0072] Step 304, automatically coding the medical history;

[0073] Step 306, performing word segmentation processing on the medical history to generate words;

[0074] Step 308, training the words into long text vectors;

[0075] Step 310, calculating the medical history similarity based on the long text vector.

[0076] In this embodiment, medical history records in the medical record are obtained through medical history statistics, the medical history is encoded using one-hot encoding, and then the similarity between the medical histories is calculated to obtain the medical history similarity.

[0077] Figure 4 Another flow chart of a method for retrieving medical record text similarity according to an embodiment of the present application is shown. Figure 4 As shown, the method includes:

[0078] Step 402, receiving text information;

[0079] Step 404, performing word segmentation processing on the text information to generate words;

[0080] Step 406, training the words into long text vectors;

[0081] Step 408, calculating the similarity of the chief complaint and the similarity of the medical history;

[0082] Step 410, normalizing the similarity of the chief complaint and the similarity of the medical history;

[0083] Step 412, feature selection;

[0084] Step 414, calculating the weight ratio of each feature through feature selection;

[0085] Step 416, weighted sum of the chief complaint similarity and the medical history similarity is performed according to the obtained weight ratio to obtain the comprehensive similarity.

[0086] In this embodiment, after obtaining the similarity between the chief complaint and the medical history, the comprehensive similarity between the two is calculated. Figure 4 As shown, the similarity of the chief complaint and the similarity of the medical history are normalized, and the input data format is standardized; through feature selection, the weight ratio of each feature is calculated; according to the obtained weight ratio, the weighted sum of the similarity of the chief complaint and the similarity of the medical history is used to obtain the comprehensive similarity.

[0087] In the second aspect of the present invention, a medical record text similarity retrieval system 50 is proposed, comprising: a memory 502, a processor 504 and a computer program stored in the memory 502 and executable on the processor 504, wherein the processor 504, when executing the computer program, achieves: receiving text information; performing word segmentation processing on the text information to generate words; training the words into long text vectors; and obtaining medical record information similar to the text information in a database based on the long text vectors.

[0088] like Figure 5 As shown, the medical record text similarity retrieval system 50 provided by the present invention performs word segmentation processing on the received text information, and the word segmentation includes word ambiguity segmentation and recognition of unregistered words. Diseases, symptoms and time can be segmented, and the segmented words are used for the next step of training. Accurate word segmentation determines the accuracy of the next step. The generated words are trained into long text vectors to obtain corresponding long text digital identifiers, and then medical record information similar to the text information is obtained in the database according to the long text vector. Retrieving medical record information through this system does not require expert participation, but automatically mines and learns medical knowledge from the database through the method of medical artificial intelligence, and constructs a model for similar medical record comparison. The model can integrate the comparison results of various types of free texts, and can efficiently and accurately obtain similar medical record recommendations, which are highly consistent with the results obtained by manual comparison by doctors, and can provide doctors with clinical pathway reference results with practical value, effectively solving the problem that doctors spend a lot of time on reviewing historical medical records. At the same time, this system can also be used to assist doctors who lack medical experience, so that patients can get better and more timely diagnosis and treatment, thereby improving clinical treatment efficiency and clinical diagnosis accuracy.

[0089] Specifically, the main processing objects of this system are the chief complaint, current medical history, past medical history, personal history, family history, and general examination results in free text, so as to obtain a complete auxiliary diagnosis for the patient.

[0090] In the above embodiment, preferably, when the processor 504 executes the computer program, it also implements: segmenting the text information, and after the step of generating words, it also includes: marking the parts of speech of the words; and classifying the words according to the marking of the parts of speech of the words.

[0091] In this embodiment, the text information is preprocessed by a named entity recognition application, the word parts of speech are annotated, and then the words are classified according to the annotations, and the correct lexical tag is assigned to each word in the sentence, and a category is assigned to each word. Further, the named entity recognition application can accurately segment unregistered words, and the part-of-speech tagging is mainly divided into rule-based and statistical-based methods. Specifically, the CRF (conditional random field) algorithm is first used to tag the words of the long text segmentation with parts of speech, and the words with the annotated parts of speech are used as RNN (recurrent neural network) inputs, and the vocabulary classification of diseases and symptoms appearing in the long text is fed back according to the category of the part of speech.

[0092] In any of the above embodiments, preferably, when the processor 504 executes the computer program, it implements the step of segmenting the text information and generating words, which specifically includes: segmenting the text information according to a disease dictionary, regular expressions, and removing stop words to generate words.

[0093] In this embodiment, the text information is segmented according to the disease dictionary, regular expressions, and stop words are removed, thereby achieving the effect of removing interference words. At the same time, the maximum matching method is used to improve the accuracy of the segmentation.

[0094] In any of the above embodiments, preferably, when the processor 504 executes the computer program, the step of training words into long text vectors is implemented, specifically including: training words into word vectors; and composing word vectors into long text vectors.

[0095] In this embodiment, the divided words are first trained into word vectors, and then the word vectors in each sentence are combined to form a long text vector, thereby obtaining the digital symbols of the long text of the medical record.

[0096] In any of the above embodiments, preferably, when the processor executes the computer program, it implements the step of obtaining medical record information similar to the text information in the database according to the long text vector, specifically including: obtaining multiple long texts similar to the text information in the database, and dividing the multiple long texts into word sets as screening sets; obtaining long texts that match the word set after word segmentation processing of the text information in the screening set, and taking it as a priority result; calculating the correlation between the word set that does not match the text information in the screening set and the word set after word segmentation processing of the text information according to the long text vector; judging whether the correlation is greater than a preset threshold; if the correlation is greater than the preset threshold, arranging the long texts that do not match the text information in ascending order according to the size of the correlation.

[0097] In this embodiment, the edit distance is first used to solve the most similar positive order of EMR (electronic medical record) words, and then divided into corresponding word sets. The long text that completely matches the text information is calculated in the word set using the jaccard distance, and its priority is set to the highest. The incompletely matched long text is used to obtain the correlation between words using the cosine distance, and a preset threshold is set. If the correlation is less than the preset threshold, the correlation is 0, which can be considered as unrelated. The associated word distances are added and sorted in positive order to solve the secondary priority long text match. Specifically, for example, the current long text segmentation set {A, B} and a set {C, A} in the library, the weighted similarity distance obtained after the cosine distance calculation is: (B·C) / (||B||·||C||).

[0098] Specifically, Figure 6 As shown, the patient's medical record 6 is input, similar medical records 62 are obtained from the medical record database 60, and the results are returned to the doctor. Specifically, after the patient describes the symptoms, the doctor can retrieve similar medical records in long texts based on experience, make corresponding clinical diagnoses, and provide a suitable treatment plan for the patient.

[0099] Specifically, Figure 7 As shown, the doctor separates the chief complaint input data 70, the patient's medical history data 72, and the general examination data 74 in the medical record data based on the new patient's input medical record data 7, performs chief complaint similarity calculation 702, medical history similarity calculation 722, and comprehensive similarity calculation 742 based on the separated data, obtains similar medical records from the Chinese electronic medical record database 78, and returns the examination results 76 to assist the doctor in making a clinical diagnosis.

[0100] like Figure 8 As shown, in the third aspect of the present invention, a computer device 8 is proposed, including a memory 80, a processor 82, and a computer program stored in the memory 80 and executable on the processor 82, and when the processor 82 executes the computer program, a method for retrieving medical record text similarity as described in any of the above embodiments is implemented.

[0101] The embodiment proposed by the present invention includes the medical record text similarity retrieval method described in any of the above embodiments, and therefore has all the beneficial effects of the medical record text similarity retrieval method.

[0102] The fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method described in any of the above embodiments are implemented, thereby having all the technical effects of the medical record text similarity retrieval method, which will not be repeated here.

[0103] In the present invention, the term "plurality" refers to two or more than two, unless otherwise clearly defined. The terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense. For example, "connected" can be a fixed connection, a detachable connection, or an integral connection; "connected" can be a direct connection or an indirect connection through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0104] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "specific embodiments", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0105] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A medical record text similarity retrieval method, It is characterized in that include: Receive text messages; Performing word segmentation processing on the text information to generate words; Training the words into long text vectors; Acquire medical record information similar to the text information in a database according to the long text vector; After the step of segmenting the text information to generate words, the method further includes: Marking the part of speech of the words; Classify the words according to the part of speech marking of the words; The step of acquiring medical record information similar to the text information in a database according to the long text vector specifically includes: Acquire multiple long texts similar to the text information in the database, and divide the multiple long texts into word sets respectively as screening sets; Obtaining, from the screening set, a long text that matches the word set after the word segmentation processing of the text information, and taking it as a priority result; Calculating the relevance between the word set in the screening set that does not match the text information and the word set after the text information is segmented according to the long text vector; Determining whether the correlation is greater than a preset threshold; If the relevance is greater than the preset threshold, the long texts that do not match the text information are arranged in positive order according to the relevance; The step of acquiring a plurality of long texts similar to the text information in the database and dividing the plurality of long texts into word sets respectively comprises: Using the edit distance, the text information is sorted in the most similar order, and the text information is divided into corresponding word sets; The step of obtaining a long text in the screening set that matches the word set after the text information is segmented specifically includes: Use the jaccard distance to calculate the long text that completely matches the text information in the word set. The calculating, according to the long text vector, the relevance between the word set in the screening set that does not match the text information and the word set after the text information is segmented specifically includes: The correlation between the word set in the screening set that does not match the text information and the word set after the text information is segmented is calculated according to the cosine distance.

2. The medical record text similarity retrieval method according to claim 1, It is characterized in that The step of performing word segmentation processing on the text information to generate words specifically includes: The text information is segmented according to a disease dictionary, regular expressions, and stop words are removed to generate words.

3. The medical record text similarity retrieval method according to claim 1, It is characterized in that The step of training the words into long text vectors specifically includes: Training the words into word vectors; The word vectors are combined into the long text vector.

4. A medical record text similarity retrieval system, It is characterized in that include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following is achieved: Receive text messages; Performing word segmentation processing on the text information to generate words; Training the words into long text vectors; Acquire medical record information similar to the text information in a database according to the long text vector; When the processor executes the computer program, after the step of performing word segmentation processing on the text information and generating words, the processor further includes: Marking the part of speech of the words; Classify the words according to the part of speech marking of the words; When the processor executes the computer program, the step of acquiring medical record information similar to the text information in the database according to the long text vector specifically includes: Acquire multiple long texts similar to the text information in the database, and divide the multiple long texts into word sets respectively as screening sets; Obtaining, from the screening set, a long text that matches the word set after the word segmentation processing of the text information, and taking it as a priority result; Calculating the relevance between the word set in the screening set that does not match the text information and the word set after the text information is segmented according to the long text vector; Determining whether the correlation is greater than a preset threshold; If the relevance is greater than the preset threshold, the long texts that do not match the text information are arranged in positive order according to the relevance; The step of acquiring a plurality of long texts similar to the text information in the database and dividing the plurality of long texts into word sets respectively comprises: Using the edit distance, the text information is sorted in the most similar order, and the text information is divided into corresponding word sets; The step of obtaining a long text in the screening set that matches the word set after the text information is segmented specifically includes: Using the jaccard distance, calculate the long text that completely matches the text information in the word set; The calculating, according to the long text vector, the relevance between the word set in the screening set that does not match the text information and the word set after the text information is segmented specifically includes: The correlation between the word set in the screening set that does not match the text information and the word set after the text information is segmented is calculated according to the cosine distance.

5. The medical record text similarity retrieval system according to claim 4, It is characterized in that When the processor executes the computer program, the step of performing word segmentation processing on the text information and generating words specifically includes: The text information is segmented according to a disease dictionary, regular expressions, and stop words are removed to generate words.

6. The medical record text similarity retrieval system according to claim 5, It is characterized in that When the processor executes the computer program, the step of training the words into long text vectors is implemented, specifically including: Training the words into word vectors; The word vectors are combined into the long text vector.

7. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the method for retrieving the medical record text similarity according to any one of claims 1 to 3 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the medical record text similarity retrieval method according to any one of claims 1 to 3 is implemented.

Images