Medical data science popularization method based on artificial intelligence

Abstract

The invention relates to the technical field of data analysis and processing, and discloses a medical data science popularization method based on artificial intelligence, and the method comprises the steps: obtaining medical related literatures through a plurality of data sources; constructing a disease tree database by referring to a disease type division mode of international disease classification (ICD), wherein the disease tree database comprises a plurality of nodes; classifying and sorting disease types of medical related literatures, and storing the classified and sorted disease types into corresponding nodes of the disease tree database; and receiving a question statement input by the user, extracting the disease type from the question statement, and retrieving the medical related literature from the node corresponding to the disease type and returning the medical related literature to the user. According to the method, the multi-order disease tree database is constructed by referring to the ICD disease classification mode, and when a user queries related literatures of a certain disease, the related literatures of the disease type from wide to thin or from thin to wide can be queried, so that more medical knowledge can be popular in science.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis and processing, and in particular to a medical data popularization method based on artificial intelligence. Background Art

[0002] After suffering from a certain disease, users or patients want to search for relevant information about the disease through the Internet. The most authentic documents are published papers, journals, etc., but users cannot search for many documents, and even the public documents are limited. Therefore, users' understanding of disease-related information is limited. How to quickly recommend relevant documents about the disease to users through questions asked by users to achieve the purpose of disease popularization is an urgent problem to be solved. Summary of the invention

[0003] The purpose of the present invention is to improve the deficiencies in the prior art and provide a medical data popularization method based on artificial intelligence.

[0004] In order to achieve the above-mentioned object of the invention, the embodiment of the present invention provides the following technical solutions:

[0005] The medical data popularization method based on artificial intelligence includes the following steps:

[0006] Step 1: Obtain medical-related literature through multiple data sources;

[0007] Step 2, constructing a disease tree database according to the disease type classification method of the International Classification of Diseases (ICD), wherein the disease tree database includes multiple nodes;

[0008] Step 3, classify and sort the medical-related literature by disease type and store them in the corresponding nodes of the disease tree database;

[0009] Step 4, receiving the question sentence input by the user, extracting the disease type from the question sentence, and retrieving medical related literature from the node corresponding to the disease type and returning it to the user.

[0010] Compared with the prior art, the present invention has the following beneficial effects:

[0011] The present invention constructs a multi-level disease tree database with reference to the disease classification method of ICD. When users search for relevant literature on a certain disease, they can search for relevant literature on the disease type from broad to detailed, or from detailed to broad, thereby popularizing more medical knowledge.

[0012] The present invention stores the documents in the closest nodes based on the extracted document segmentation by creating an optimal allocation function, so that when the user queries the disease type, the documents in the corresponding node can be recommended to the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0014] Figure 1 is a flow chart of the method of the present invention;

[0015] Figure 2 This is the classification method of "05 Endocrine, nutritional or metabolic diseases" in the International Classification of Diseases ICD. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0017] It should be noted that similar numbers and letters represent similar items in the following figures, so once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance, or implying any such actual relationship or order between these entities or operations. In addition, the terms "connected", "connected", etc. can be directly connected between elements, or indirectly connected via other elements.

[0018] The present invention is achieved through the following technical solutions: Figure 1 As shown, a medical data popularization method based on artificial intelligence includes the following steps:

[0019] Step 1: Obtain medical-related literature through multiple data sources.

[0020] Access the internal database of the hospital to obtain all medical-related documents. The medical data of this hospital are published by doctors of this hospital, which is more in line with the diagnosis and treatment process of patients in this hospital. Use the Web crawler tool to crawl a number of medical-related documents from public network resources on the Internet, including academic databases, journal websites, institutional libraries, professional associations, research institutions reports and other public network resources. When crawling public network resources, medical documents with large citations and reprints are crawled first. This embodiment does not restrict the source of medical data documents. As long as the data source is reasonable, it can be obtained, thereby forming a large-scale medical-related document, which can be used for subsequent sorting and classification to provide patients with targeted medical science popularization.

[0021] Step 2: construct a disease tree database according to the disease type classification method of the International Classification of Diseases (ICD), wherein the disease tree database includes multiple nodes.

[0022] The disease tree database is constructed with reference to the International Classification of Diseases (ICD) developed and published by the World Health Organization (WHO). ICD is divided into 26 chapters according to etiology, location, etc., and each chapter is further classified until the final code is obtained. For example, Chapter 5 is "05 Endocrine, nutritional or metabolic diseases"; it is further subdivided into "endocrine diseases, nutritional disorders, metabolic disorders, post-operative endocrine or metabolic disorders"; "endocrine diseases" are further subdivided into "thyroid or thyroid hormone system disorders, diabetes, other glucose regulation or pancreatic endocrine disorders, parathyroid or parathyroid hormone system disorders, pituitary hormone system disorders, adrenal or adrenal hormone system disorders, etc."; "diabetes" is further subdivided into "5A10 Type 1 diabetes, 5A11 Type 2 diabetes, 5A12 Malnutrition-related diabetes, 5A13 Diabetes (other specified types), 5A14 Diabetes (unspecified types), acute complications of diabetes". It can be seen that, Figure 2 As shown, "5A11" is the final code. For details, please refer to the official website https: / / icd11.pumch.cn. This plan will not repeat the ICD classification method.

[0023] This scheme refers to the classification method of ICD and selects the classification method of the first 24 chapters to construct a disease tree database. The chapters "01 Certain infectious diseases or parasites, 02 Tumors, 03 Blood or hematopoietic organ diseases, ..., 23 External causes of disease or death, 24 Factors affecting health status or contact with health care institutions" are used as root nodes, and then the descendant nodes are extended downward, but only to the 4-digit code position, such as extending to the level of the leaf node "5A10 Type 1 diabetes", which has exactly 4 digits of code "5A10".

[0024] It should be noted that, due to the different classification results of each chapter, the order, branching degree, number of descendant nodes, height, etc. of each tree are different. This solution completely constructs 24 disease tree databases according to the ICD classification method, that is, there are 24 root nodes, and then further extends its descendant nodes according to the classification of each chapter. It is easy to understand that this tree construction method can be realized based on the existing technology, and this solution does not improve it. The node extension basis can be based on the ICD classification method.

[0025] Step 3: Classify and organize the medical-related literature by disease type and store them in the corresponding nodes of the disease tree database.

[0026] This step extracts the disease type of the document through the name and keywords of the medical-related document. Usually, keywords will be recorded in a document, or the main idea has been mentioned in the document name. For example, the document named "The efficacy and safety of Shuxuening injection in the auxiliary treatment of diabetic patients" can quickly extract the disease type "diabetes", so the document can be stored in the disease tree database corresponding to 05 endocrine, nutritional or metabolic diseases, specifically in the descendant node "diabetes". Since the node "diabetes" is not a leaf node, it also has a child node "5A10 type 1 diabetes, 5A11 type 2 diabetes, etc.", which is the leaf node of this tree, so the content of the document is further queried. If there is a disease type corresponding to a leaf node, the document is stored in the leaf node. For example, if the document "Efficacy and safety of Shuxuening injection in the auxiliary treatment of diabetic patients" contains the content of "type 1 diabetes", the document is stored in leaf node 5A10; if there is also the content of "type 2 diabetes", the document is also stored in leaf node 5A11; if the content represented by any leaf node cannot be found in the document, the document is stored in the node corresponding to "diabetes". It is easy to understand that each parent node contains the addresses of all its descendant nodes, so the documents stored in all its descendant nodes can be queried in the parent node.

[0027] This solution extracts the word segmentation of the document and determines which descendant node a document is stored in by creating an optimal allocation function. The optimal allocation function is improved based on the support vector machine (SVM). Assume that the name and keywords of a document are segmented to obtain the word segmentation results, and then the word segmentation results are encoded by word vectors to obtain the word vector encoding set X={x1,x2,...,x n}, where x i Represents the encoding value of the i-th word segmentation, i=1,2,...,n, n represents the number of word segments of the name and keyword. The optimal allocation function created is:

[0028] Among them, f(x i) represents the optimal allocation function; x i Represents the encoding value of the i-th word; y j represents the encoding value of the jth node, j=1,2,...,K, K represents the number of nodes; sgn represents the step function; Represents kernel function parameters; represents the Gaussian radial basis parameters, ; represents the bias term; h i The weight coefficient of the ith word is related to the number of times the ith word appears in the document. i ,y j Obtained through word2vec word vector encoding technology.

[0029] When this scheme assigns a document to a node in the disease tree database, the following situations exist:

[0030] (1) If a document has been stored in a non-leaf node based on its name and keyword segmentation, then further query whether there is a child node or leaf node with a matching segmentation in the content of the document. If there is only a matching child node, then the document is stored in the matching child node; if there is a matching leaf node, then the document is stored in the matching leaf node.

[0031] (2) If there is no matching node for the name and keyword segmentation of a document, further query is performed to see whether there is a matching node or leaf node in the content of the document. If there is only a matching non-leaf node, the document is stored in the matching non-leaf node. If there is a matching leaf node, the document is stored in the matching leaf node.

[0032] (3) If a document has been stored in a leaf node based on its name and keyword segmentation, matching will no longer be performed based on the content of the document.

[0033] (4) If there is no matching node for the title, keywords, or content segmentation of a document, the document will not be stored.

[0034] It can be seen that the documents need to be stored in higher-order child nodes as much as possible, preferably in leaf nodes. This solution uses the optimal allocation function twice. The first time is to query the matching of the document name and keywords with all nodes, and the second time is to query the matching of the document content with all nodes. There are two differences in using the optimal allocation function:

[0035] 1. The first vector encoding set X={x1,x2,...,x n} represents the encoding value of the document name and keyword segmentation, and the second vector encoding set X={x1,x2,...,x n} represents the encoding value of the word segmentation of the content of the document.

[0036] 2. The first K value represents the number of all nodes in the 24 disease tree databases, while the second K value only represents the number of descendant nodes of the node stored in the first time.

[0037] This scheme uses the optimal allocation function twice. The first allocation is based on the name and keywords of the document. Because the name and keywords of the document are the main idea of ​​the document, it can find out which disease tree database the document belongs to. If the node stored in the first time is a non-leaf node, a second match is performed, and the second match is only performed for the descendants of the node stored in the first time, which has greatly reduced the amount of calculation. If the document has been stored in the leaf node in the first time, there is no need to perform a second match, and there is no need to analyze the content of the document. The length of the content is much larger than the name and keywords, so the amount of calculation is further reduced.

[0038] Step 4, receiving the question sentence input by the user, extracting the disease type from the question sentence, and retrieving medical related literature from the node corresponding to the disease type and returning it to the user.

[0039] For example, if a user inputs the question sentence "How to treat hypothyroidism", it is easy to extract the disease type of "hypothyroidism", and then the medical related literature in the leaf node "5A00 Hypothyroidism" is returned to the user from the disease tree database of 05 endocrine, nutritional or metabolic diseases.

[0040] When feeding back documents, the documents in the leaf node "5A00 Hypothyroidism" are sorted by recommendation value, and documents with higher recommendation values ​​are presented to users first. The recommendation value of a document is calculated by the following formula:

[0041] Among them, SID represents the recommendation value; QSC represents the document's shared citation index; ASC represents the document's interactive comment index; E samp represents the page scrolling index within the unit time T. The faster the page scrolling speed within the unit time T, the less the user likes the document. w1 is the weight of QSC, w2 is the weight of ASC, and w3 is The weight of .

[0042] In addition, the user can be prompted to view the literature stored in the ancestor node of the leaf node "5A00 Hypothyroidism", so that the user can further learn about the medical literature related to "thyroid or thyroid hormone system diseases", "endocrine diseases", and even "endocrine, nutritional or metabolic diseases". It can be seen that this solution divides the disease types from broad to fine with reference to the disease classification method of ICD, and the parent node has the address of its descendant node, so that the user can query upward or downward to find broader or finer literature on related diseases.

[0043] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A medical data popularization method based on artificial intelligence, characterized by: The following steps are involved: Step 1: Obtain medical-related literature through multiple data sources; Step 2, constructing a disease tree database according to the disease type classification method of the International Classification of Diseases (ICD), wherein the disease tree database includes multiple nodes; Step 3, classify and sort the medical-related literature by disease type and store them in the corresponding nodes of the disease tree database; Step 4, receiving the question sentence input by the user, extracting the disease type from the question sentence, and retrieving medical related literature from the node corresponding to the disease type and returning it to the user.

2. The medical data popularization method based on artificial intelligence according to claim 1, characterized in that: In step 2, disease types corresponding to the first 24 chapters of the International Classification of Diseases (ICD) are selected to construct 24 disease tree databases, and the disease types corresponding to the chapters are used as root nodes of the disease tree databases; each disease tree database extends descendant nodes according to the classification method of the International Classification of Diseases (ICD), and uses disease types with 4-bit codes as leaf nodes.

3. The medical data popularization method based on artificial intelligence according to claim 2, characterized in that: The step 3 specifically comprises the following steps: Step 3-1, after word segmentation of the document name and keywords, perform word vector encoding, and use the optimal allocation function to store the document in the node that matches the word segmentation. If the node is a non-leaf node, execute step 3-2. If the node is a leaf node, end; Step 3-2, after the content of the document is segmented, word vector encoding is performed. In the grandchild nodes of the node stored in step 3-1, the optimal allocation function is used to identify the node that matches the segmentation, and the document is stored in the node that matches the segmentation.

4. The medical data popularization method based on artificial intelligence according to claim 3 is characterized by: The optimal allocation function in step 3-1 is: Among them, f(x i ) represents the optimal allocation function; x i Represents the encoding value of the i-th word; y j represents the encoding value of the jth node, j=1,2,...,K, K represents the number of nodes; sgn represents the step function; Represents kernel function parameters; represents the Gaussian radial basis parameters, ; represents the bias term; h i Represents the weight coefficient of the i-th word.

5. The medical data popularization method based on artificial intelligence according to claim 1, characterized in that: In step 4, when retrieving medical related documents from the node corresponding to the disease type and returning them to the user, the recommendation value of the document in the node is calculated: Among them, SID represents the recommendation value; QSC represents the document's shared citation index; ASC represents the document's interactive comment index; E samp represents the page scrolling index within the unit time T. The faster the page scrolling speed within the unit time T, the less the user likes the document. w1 is the weight of QSC, w2 is the weight of ASC, and w3 is The weight of .

6. The medical data popularization method based on artificial intelligence according to claim 1, characterized in that: In step 4, when medical related documents are retrieved from the node corresponding to the disease type and returned to the user, the user is prompted to query the documents in the ancestor node or descendant node of the node.

Images