Key acupuncture point node mining method based on acupuncture point-disease network

Abstract

The invention discloses a key acupuncture point node mining method based on an acupuncture point-disease network, and belongs to the crossing field of information systems and empirical medicine. Comprising the steps of establishing a weighted undirected acupoint-disease network based on a clinical acupuncture prescription, providing a high-specificity key node mining method based on the acupoint-disease network, and balancing contribution of a local area to the network on the basis of meridian distribution characteristics in combination with a traditional Chinese medicine acupoint selection principle. And key acupoints with high influence on global topology are selected according to the evaluation indexes of the key nodes. According to the method, key acupuncture points of meridians and collaterals in each region of a human body are positioned and mined by using a complex network theory in the field of an information system, and the key acupuncture points which have obvious influence on various common diseases of the human body are positioned from a new perspective, so that other key acupuncture points in local meridians and collaterals of the human body are determined; and an empirical medical researcher is assisted to find the optimal prescription acupoint for a specific disease under the resource constraint.

Description

Technical Field

[0001] The present invention belongs to the intersection field of information system and evidence-based medicine, and in particular relates to a key acupoint node mining method based on an acupoint-disease network. Background Art

[0002] As a traditional medical method, acupuncture has been recognized by the World Health Organization and other international organizations for its reliable clinical efficacy. With the rapid development of modern medical technology, researchers have gradually shifted from focusing on the linear relationship between "effect and intervention" to testing the neural correlation or other biological mechanisms of acupuncture. In traditional Chinese medicine culture, the meridian system is the basis of acupuncture treatment. It is a systematic sequence of empirical knowledge. A large number of known or unknown acupuncture points are scattered on the body surface with the meridian system as the carrier, forming a human network closely related to the physiological state.

[0003] Modern medicine believes that the human body is a giant network coupled by multiple functional systems. Physiological systems are closely connected through a common language. From a macroscopic perspective, any abnormal event of physiological function is fed back through multiple network nodes. Studies have shown that acupuncture treatment is actually a comprehensive regulatory process of the network. The regulatory information is cascaded and amplified by the acupoint network activated by physical stimulation, and ultimately acts on the disease based on the distribution characteristics of the meridian system. With the rapid development of complex network science, exploring the relationship between acupoints and diseases with network thinking has become a new trend in TCM research. Network thinking focuses on the interaction between the internal components of the system, thereby clarifying the structure, function and evolution of the system. Different from traditional multi-source statistical analysis methods, network analysis can provide new theoretical support for acupoint-disease research, including visual schematic diagrams, quantitative indicators of the overall or local structure of the network, and mathematical models for the acupoint-disease network.

[0004] There are always some key nodes with great influence in various complex networks. Generally speaking, the influence of nodes can be highlighted from aspects such as the topological structure of the network model or the location attributes of the nodes. For the acupoint-disease network, symptoms involve multi-system dysfunction and multi-targets, and stimulation of a single acupoint node cannot improve all symptoms. At the same time, the areas that can be activated and regulated by each acupoint are also different. Therefore, the synergistic effect of acupoint combination must be one of the focuses of research in the field of evidence-based medicine. Based on this effect, key acupoint prescriptions for treating specific diseases can be excavated. Among them, in the acupoint-disease network, key nodes are not isolated functional nodes, but refer to some core nodes with high influence, wide regulation range, and rich combination patterns. Summary of the invention

[0005] The purpose of the present invention is to propose a method for mining key acupoint nodes based on the acupoint-disease network. First, a weighted undirected acupoint-disease complex network is constructed based on clinical acupuncture prescription literature, and reliable node centrality parameters are obtained by counting the frequency of acupoint selection of acupoint combinations. The acupoint-disease network is a one-dimensional isomorphic network, which is an acupoint association network established based on the synergistic effect of acupoints on diseases. Secondly, a method for mining key acupoint nodes based on the meridian system is proposed. In view of the functional cross-effects of acupoints, the proposed method balances the contribution of acupoints in each community to the network based on the meridian system. After preliminary screening, the present invention designs a multi-constrained key node evaluation index based on the static characteristics of the network for the acupoint-disease network, and mines a group of key acupoint nodes that have a high impact on the global topology according to the node score.

[0006] In order to achieve the above object, the present invention comprises the following steps: a key acupoint node mining method based on acupoint-disease network:

[0007] Step 1: Establish an acupoint-disease network based on the literature on acupoints used in clinical acupuncture prescriptions: Define the nodes and edges of the acupoint-disease network based on complex network theory. In terms of acupoint prescriptions for common diseases, define the reasons for the connection between two acupoints, and complete statistical feature extraction for complex networks on the established network.

[0008] The acupoint-disease network is an acupoint association network established based on the synergistic effect of acupoints on diseases, which is abstracted as a graph G = (V, E), where V is the set of acupoint nodes and E is the set of edges.

[0009] If two acupoint nodes can produce synergistic effects on the same disease, there is an edge between the two nodes; when two acupoints have synergistic effects on multiple types of diseases, the edge weight w is set based on the statistical value of the number of diseases synergistically affected by each pair of acupoints. ij , the number of nodes in the network is recorded as N = |V|, and the number of edges is recorded as M = |E|; the adjacency matrix is ​​used to represent the network. For a network G = (V, E) with N nodes, its adjacency matrix A = {a ij} N×N is an N-order square matrix, where a ij Represents the element in the i-th row and j-th column of the adjacency matrix A. The acupoint-disease network is an undirected weighted network, and the adjacency matrix elements are positioned as follows:

[0010]

[0011] Step 2: Mining key acupuncture point nodes based on meridian distribution characteristics: including community division of meridian distribution characteristics, selecting candidate nodes in each community, and selecting the final key acupuncture point nodes from the candidate nodes based on the proposed key node evaluation indicators.

[0012] 1) Community division based on meridian system

[0013] The nodes in the data source are divided into communities of acupoint disease network according to the meridians they belong to; community division is regarded as a process of classification according to the attributes of acupoints; there are 14 meridians and named extra meridian acupoints in the human body, 12 of which are symmetrically distributed on the head, face, trunk and limbs on the body surface, and 2 meridians are the conception meridian and the governor meridian; the nodes are divided according to the location attributes of the acupoints to form 15 communities;

[0014] 2) Selection of candidate nodes

[0015] The nodes in the 15 communities were preliminarily screened for key node candidate sets, and establishing candidate node sets is a common step in node mining. In the field of evidence-based medicine, the candidate set is composed of the top 10% of nodes in each community according to a certain evaluation index and the nodes connecting different communities. Based on the characteristics of the acupoint-disease network, that is, it is composed of multiple fully connected subnetworks. According to the principle of edge building, a fully connected small-scale network is formed between the acupoints of a single disease type prescription, which is called a fully connected subnetwork. The nodes in each community can belong to multiple subnetworks. In the process of establishing the candidate set, the nodes connecting different communities are ignored to maintain a limited and reliable candidate set dimension; according to the network size and number of communities of the acupoint-disease network, each community will select 4 nodes to form a candidate set with a dimension not exceeding 60;

[0016] Closeness centrality is used as the evaluation index of candidate nodes. For an undirected connected graph, the closeness centrality C of the node C (v i ) is expressed as:

[0017]

[0018] Among them, d ij For node v i To node v j The length of the shortest path, N is the network size; the greater the proximity of the node, the closer the node is to the center of the network, that is, the higher its importance in the network;

[0019] 3) Selection of key acupuncture points

[0020] The importance of a node is judged by the degree of influence of the node on all nodes in the network. For an undirected weighted network, paths of different lengths contribute differently to the node, and the weight on each path determines the influence of the path. Based on the above two principles, a highly specific key node evaluation index is used. The key point node set will consist of the two nodes with the highest scores in the candidate set of each community. The evaluation index of the key point node is shown in the following formula:

[0021]

[0022] in, Represents node v i The weighted influence of all nodes in the network in a path of length k; θ k represents the accumulation of weighted influence of all paths of length k in the network; D is the network diameter, which refers to the maximum value of the shortest distance between any two points in the network, Key_node(v i ) indicates that node v is within the D-hop path range. i The weighted influence on the global network accounts for the proportion of the total network value. In the acupoint-disease network, Key_node(v i ) index contains node v i The disease regulation ability of acupoint combinations;

[0023] The acupoint-disease network is an undirected network. The adjacency matrix of the graph G = (V, W) is a real symmetric matrix. The elements in the k-th power of the weighted adjacency matrix represent the accumulated sum of the weights of all paths of length k between two nodes, that is:

[0024]

[0025] Therefore, the evaluation index of key acupuncture points is expressed as:

[0026]

[0027] The beneficial effects created by the present invention are as follows: the present invention uses complex network theory to mine key acupuncture points for diagnosis and treatment, establishes a weighted and undirected acupuncture point-disease network based on the acupuncture point literature of clinical acupuncture prescriptions, and uses a highly specific key node mining method based on the network to improve the recognition accuracy of key acupuncture point nodes. Combined with the principle of acupuncture points in traditional Chinese medicine, the present invention balances the contribution of local areas to the network based on the distribution characteristics of meridians, and selects 30 key acupuncture points with high influence on the global topology according to the evaluation index of key nodes in a specific application example. Therefore, the key acupuncture points that show high centrality in the acupuncture point-disease network can be used as core acupuncture points to improve and integrate the acupuncture point prescriptions of empirical medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is an example graph of acupoint disease network and its static characteristic parameters established based on clinical acupuncture literature;

[0029] Figure 2 The calculation principle of the weighted influence of node v′1 in the network G′ under 1-hop path and 2-hop path;

[0030] Figure 3 For application cases, human body distribution maps of 30 key acupoints belonging to different meridian communities were mined. DETAILED DESCRIPTION

[0031] A method for mining key acupoint nodes based on acupoint-disease network:

[0032] Step 1: Establish an acupoint-disease network based on the literature on acupoints used in clinical acupuncture prescriptions. Define the nodes and edges of the acupoint-disease network based on complex network theory. Define the concept of edges between acupoints in terms of acupoint prescriptions for common diseases, and complete statistical feature extraction for complex networks on the established network. Establish an acupoint-disease network as follows.

[0033] The acupoint-disease network is an acupoint association network established based on the synergistic effect of acupoints on diseases, represented by the graph G = (V, E), where V is the set of acupoint nodes and E is the set of edges. It is defined that if two acupoint nodes can produce a synergistic effect on the same disease, there is an edge between the two nodes. When two acupoints have a synergistic effect on multiple types of diseases, the edge weight w is set based on the statistical value of the number of diseases synergistically affected by each pair of acupoints. ij The number of nodes in the network is recorded as N = |V|, and the number of edges is recorded as M = |E|. The adjacency matrix is ​​used to represent the network. For a network G = (V, E) with N nodes, its adjacency matrix A = {a ij} N×N is an N-order square matrix, where a ij Represents the element in the i-th row and j-th column of the adjacency matrix A. The acupoint-disease network is an undirected weighted network, and the adjacency matrix elements are positioned as follows:

[0034]

[0035] Step 2: Mining key acupoint nodes based on meridian distribution characteristics:

[0036] Step 2 involves the following processes, including community division based on meridian distribution characteristics, selecting candidate nodes in each community, and selecting the final key acupoint nodes from the candidate nodes based on the proposed key node evaluation indicators.

[0037] 1) Community division based on meridian system

[0038] The community division of the acupuncture point disease network is to divide the nodes in the data source strictly according to the meridians to which they belong. The present invention regards community division as a process of classification according to the attributes of acupuncture points. There are fourteen meridians in the human body, twelve of which are symmetrically distributed on the head, face, trunk and limbs on the body surface. The remaining two meridians (the Conception Meridian and the Governor Vessel) are located on the front midline and the back midline of the human body respectively. In addition to the acupuncture points evenly distributed on these fourteen meridians, there are also named extra meridian acupuncture points. In summary, the nodes are divided according to the location attributes of the acupuncture points (meridians to which they belong) to form fifteen communities.

[0039] Community division is required before ranking the importance of nodes. First, avoid all the key nodes being concentrated on certain specific meridians. The location attributes of each meridian are not the same, and the acupoints on different meridians may have the ability to regulate a variety of common diseases in a local area. At the same time, acupuncture treatment has the principle of "selecting acupoints from the same meridian", which indicates that the acupoints on the same meridian have a wide range of functional intersections. Therefore, the community division method based on the meridian system can highlight some important acupoints with targeted therapeutic effects on special diseases in the network model, and balance the contribution of nodes in each community to the overall network, so as to ensure that Chinese medicine doctors have core acupoints that can be used for reference in each target meridian after dialectics. Secondly, in order to improve the reliability of the key node set, the present invention focuses on the extra meridian acupoints involved in the data source. Extra meridian acupoints are a series of acupoints with names, fixed positions and main therapeutic effects. In the specific application example data set, extra meridian acupoints can work synergistically with fourteen meridian acupoints on common diseases. In summary, the acupoint disease network is divided into 15 communities according to the acupoint attributes, including fourteen meridians and extra meridian acupoints. Each community is shown in different colors Figure 1 , where the acupoints are marked with international numbers.

[0040] The 15 communities are the 14 meridians of the human body plus the extra meridian acupoints. Figure 1 The 15 communities are visualized using a network diagram, which is not necessary because the number and location of the meridians in the human body are known. Figure 3 This is the result map after finding the key acupoints. Each community has key acupoint nodes, with a total of 14 fixed meridian maps. Extra meridian acupoints are added and drawn in the Ren and Du meridian maps.

[0041] Therefore, the explanation of the community is the previous description part, "The acupoint disease network is divided into 15 communities according to the properties of the acupoints, including the fourteen meridians and extra meridian acupoints."

[0042] 2) Selection of candidate nodes

[0043] The nodes in the 15 communities were preliminarily screened for key node candidate sets. Establishing a candidate node set is a common step in node mining. In the field of evidence-based medicine, the candidate set is generally composed of the top 10% of nodes in each community according to a certain evaluation index and the nodes connecting different communities. Based on the characteristics of the acupoint-disease network, that is, it is composed of multiple fully connected subnetworks (according to the edge building principle, a fully connected small-scale network will be formed between the acupoints based on a single disease type prescription, which is called a fully connected subnetwork), the nodes in each community may belong to multiple subnetworks. Therefore, in the process of establishing the candidate set, the nodes connecting different communities are ignored to maintain a limited and reliable candidate set dimension. According to the network size and the number of communities of the acupoint-disease network, 4 nodes will be selected from each community to form a candidate set with a dimension not exceeding 60.

[0044] The evaluation index of candidate nodes generally adopts node centrality parameters. The present invention adopts proximity centrality as the evaluation index of candidate nodes. Proximity is one of the basic concepts in topological space and has a strong dependence on the network topology. For an undirected connected graph, the proximity centrality C of a node is C (v i )The most natural definition can be expressed as:

[0045]

[0046] Among them, d ij For node v i To node v j The length of the shortest path, N is the network size. The greater the proximity of the node, the closer the node is to the center of the network, that is, the higher its importance in the network.

[0047] 3) Selection of key acupuncture points

[0048] The importance of a node can be judged from the degree of influence of the node on all nodes in the network. For an undirected weighted network, paths of different lengths contribute differently to the node. At the same time, the weight on each path determines the influence of the path. Based on the above two principles, the present invention relates to a key node evaluation index with high specificity. The key acupoint node set will consist of the two nodes with the highest scores in the candidate set of each community. The evaluation index of the key acupoint node is shown as follows:

[0049]

[0050] in, Represents node v i The weighted influence of all nodes in the network in a path of length k; θ kIt represents the accumulation of weighted influence of all paths of length k in the network; D is the network diameter, which refers to the maximum value of the shortest distance between any two points in the network. i ) indicates that node v is within the D-hop path range. i The weighted influence on the global network accounts for the proportion of the total network value. In the acupoint-disease network, this indicator focuses on the nodes that contain v i The disease regulation ability of the acupoint combination includes the regulation range. Taking a simple network G′ with N=4 as an example, The calculation process of G′ is explained in detail (the connection of nodes is random). The network model of G′ is as follows: Figure 2 As shown. It is known that the diameter D' of G' is 2. When discussing the importance of node v'1 in the network, it is necessary to calculate the weighted influence of node v'1 on the network in the 1-hop path and the 2-hop path respectively. The weighted influence based on the 1-hop path is actually the weighted degree of node v'1, which reflects the closeness of the connection between the node and its neighboring nodes, as shown in Figure 2 As shown in the figure on the left. In the network construction rules, the neighbor nodes are two prescription acupuncture points that have a direct therapeutic effect on a certain disease. Therefore, the weighted influence of the 1-hop path reflects the ability of node v′1 to directly participate in physiological regulation. The weighted influence of node v′1 on the network in the 2-hop path is as follows: Figure 2 As shown in the figure on the right. In acupuncture theory, remote acupoints are reflected in the number of connection patterns established between two indirectly acting nodes due to different diseases. Unlike the one-hop path, the selection of multi-hop paths will include repeated paths between adjacent nodes. This method helps to strengthen the influence of neighboring nodes on the entire network, and at the same time can reflect the regulatory ability formed between nodes due to different intermediary nodes, that is, different paths. Therefore, although the calculation logic is based on the global topology, for a single node, a node with higher degree centrality will make a greater contribution to the acupoint-disease network, that is, the influence of the local topological structure on the global.

[0051] In summary, the acupoint-disease network is an undirected network, and the adjacency matrix of the graph G = (V, E) is a real symmetric matrix. The elements in the k-th power of the weighted adjacency matrix represent the accumulated sum of the weights of all paths of length k between two nodes, that is:

[0052]

[0053] Therefore, the evaluation index of key acupuncture points can be expressed as:

[0054]

[0055] Embodiment 1:

[0056] The acupoint-disease network was established using the statistical results of acupoints used in high-frequency acupuncture prescriptions for common diseases. This data is a summary of acupoints used for common diseases recorded in 5733 Chinese clinical acupuncture literature. The data source involves a total of 50 common human diseases, including neurological, respiratory, digestive, cardiovascular, dermatological, gynecological, and multiple types of painful surgical diseases. The above common diseases are widely representative in the current treatment goals of Chinese acupuncture. Specifically, taking diarrhea as an example, the data set records 90 prescription acupoints involving this disease, with a cumulative number of acupoints used 737 times. The 10 most frequently used acupoints are Zusanli, Tianshu, Shenque, Zhongshu, Changqiang, Guanyuan, Neiguan, Pishu, Qihai, and Dachangshu. For example, this application example uses the 10 most frequently used acupoints as the acupoint nodes with the strongest ability to regulate the "diarrhea" disease, and establishes a fully connected subnetwork based on a single disease.

[0057] After the community division, a total of 55 nodes are formed into a candidate set based on the sorting results of the centrality parameters. The Key_node (v i ), and obtain the two key acupoint nodes with the highest scores in each community. The screening process maintains the classic mode of node mining to ensure the accuracy of the results while avoiding traversal of the global network. The community attributes and centrality parameters of the key acupoint nodes are shown in Table 1. C (v i ) does not necessarily have a higher Key_node(v i ), for example, acupoint CV4 ranked third in the candidate set of the CV community, C C (CV23) and C C (CV22) are 0.568 and 0.552 respectively, but CV4 has the highest Key_node score in this community. The human distribution diagram of the key acupoints of the 30 different meridian communities mined in this application example is shown in the figure below. Figure 3 shown.

[0058] Table 1 Community attributes and centrality parameters of key acupoint nodes

[0059]

[0060]

Claims

1. A key acupoint node mining method based on acupoint-disease network, characterized by: Step 1: Establish an acupoint-disease network based on the literature on acupoints used in clinical acupuncture prescriptions: Define the nodes and edges of the acupoint-disease network based on complex network theory. In terms of acupoint prescriptions for common diseases, define the reasons for the connection between two acupoints, and complete statistical feature extraction for complex networks on the established network. Step 2: Mining key acupuncture point nodes based on meridian distribution characteristics: including community division of meridian distribution characteristics, selecting candidate nodes in each community, and selecting the final key acupuncture point nodes from the candidate nodes based on the proposed key node evaluation indicators.

2. The key acupoint node mining method based on acupoint-disease network according to claim 1 is characterized by: In the step 1: The acupoint-disease network is an acupoint association network established based on the synergistic effect of acupoints on diseases, which is abstracted as a graph G = (V, E), where V is the set of acupoint nodes and E is the set of edges. If two acupoint nodes can produce synergistic effects on the same disease, there is an edge between the two nodes; when two acupoints have synergistic effects on multiple types of diseases, the edge weight w is set based on the statistical value of the number of diseases synergistically affected by each pair of acupoints. ij , the number of nodes in the network is recorded as N = |V|, and the number of edges is recorded as M = |E|; the adjacency matrix is ​​used to represent the network. For a network G = (V, E) with N nodes, its adjacency matrix A = {a ij } N×N is an N-order square matrix, where a ij Represents the element in the i-th row and j-th column of the adjacency matrix A. The acupoint-disease network is an undirected weighted network, and the adjacency matrix elements are positioned as follows:

3. The key acupoint node mining method based on acupoint-disease network according to claim 1 is characterized by: In the step 2: 1) Community division based on meridian system The nodes in the data source are divided into communities of acupoint disease network according to the meridians they belong to; community division is regarded as a process of classification according to acupoint attributes; There are 14 meridians and named extra meridian acupoints in the human body. Twelve of the 14 meridians are symmetrically distributed on the head, face, trunk and limbs. Two meridians are the Ren Meridian and the Du Meridian. The nodes are divided according to the location attributes of the acupoints to form 15 communities. 2) Selection of candidate nodes The nodes in the 15 communities were preliminarily screened for key node candidate sets, and establishing candidate node sets is a common step in node mining. In the field of evidence-based medicine, the candidate set is composed of the top 10% of nodes in each community according to a certain evaluation index and the nodes connecting different communities. Based on the characteristics of the acupoint-disease network, that is, it is composed of multiple fully connected subnetworks. According to the principle of edge building, a fully connected small-scale network is formed between the acupoints of a single disease type prescription, which is called a fully connected subnetwork. The nodes in each community can belong to multiple subnetworks. In the process of establishing the candidate set, the nodes connecting different communities are ignored to maintain a limited and reliable candidate set dimension; according to the network size and number of communities of the acupoint-disease network, each community will select 4 nodes to form a candidate set with a dimension not exceeding 60; Closeness centrality is used as the evaluation index of candidate nodes. For an undirected connected graph, the closeness centrality C of the node C (v i ) is expressed as: Among them, d ij For node v i To node v j The length of the shortest path, N is the network size; the greater the proximity of the node, the closer the node is to the center of the network, that is, the higher its importance in the network; 3) Selection of key acupuncture points The importance of a node is judged by the degree of influence of the node on all nodes in the network. For an undirected weighted network, paths of different lengths contribute differently to the node, and the weight on each path determines the influence of the path. Based on the above two principles, a highly specific key node evaluation index is used. The key point node set will consist of the two nodes with the highest scores in the candidate set of each community. The evaluation index of the key point node is shown in the following formula: in, Represents node v i The weighted influence of all nodes in the network in a path of length k; θ k represents the accumulation of weighted influence of all paths of length k in the network; D is the network diameter, which refers to the maximum value of the shortest distance between any two points in the network, Key_node(v i ) indicates that node v is within the D-hop path range. i The weighted influence on the global network accounts for the proportion of the total network value. In the acupoint-disease network, Key_node(v i ) index contains node v i The disease regulation ability of acupoint combinations; The acupoint-disease network is an undirected network. The adjacency matrix of the graph G = (V, W) is a real symmetric matrix. The elements in the k-th power of the weighted adjacency matrix represent the accumulated sum of the weights of all paths of length k between two nodes, that is: Therefore, the evaluation index of key acupuncture points is expressed as:

Images