Co-disease trajectory clustering method and system

Abstract

The invention relates to a co-disease trajectory clustering method and system, belongs to the technical field of medical health, and solves the problem of inaccurate co-disease trajectory clustering in the prior art. The method comprises the following steps: acquiring common disease trajectory data, and calculating the correlation between common disease trajectories; constructing an initial trajectory set, and performing bipartite clustering on the initial trajectory set based on the correlation between the common disease trajectories to obtain two subclass trajectory sets; for each subclass trajectory set, obtaining the attribute of each common disease trajectory in the subclass trajectory set based on the connectivity of the disease chains in the subclass trajectory set, and judging whether the subclass trajectory set is a to-be-clustered trajectory subset based on the attribute of the common disease trajectory; the attributes comprise a hub and a non-hub; if the to-be-clustered track subsets exist, carrying out two-way nested clustering on each to-be-clustered track subset; and obtaining a clustering result of the co-disease trajectory. And accurate co-disease trajectory clustering is realized.

Description

Technical Field

[0001] The present invention relates to the field of medical health technology, and in particular to a comorbidity trajectory clustering method and system. Background Art

[0002] In the vast field of medical health research, comorbidity trajectory clustering technology has shown great potential and application value. This technology is deeply involved in the complex data of disease evolution, and uses advanced clustering algorithms to classify them according to the similarities between the occurrence and development of diseases, aiming to reveal the inherent laws and evolution patterns of comorbidity phenomena. This not only provides clinicians with a powerful tool to identify and manage patient groups with similar health challenges, but also promotes the formulation of more precise treatment strategies and preventive intervention measures. In addition, this technology is also of far-reaching significance for the exploration of disease pathogenesis and in-depth analysis of epidemiology, laying a solid foundation for the scientific formulation of public health strategies.

[0003] However, existing trajectory clustering methods have significant limitations for clustering disease trajectories. Due to the extensive and close association between some common diseases (such as hypertension) and many other diseases, existing clustering methods often mistakenly merge these disease trajectories with low correlation into the same group. Take hypertension as an example. It appears in the development paths of many diseases, but current clustering techniques often indiscriminately cluster disease trajectories containing hypertension, even if these trajectories are not highly correlated in clinical practice. Therefore, it is impossible to capture the characteristics of comorbidities, and the clustering accuracy for comorbidities is low, which in turn affects the accuracy of treatment plans and preventive measures. Summary of the invention

[0004] In view of the above analysis, the embodiments of the present invention aim to provide a comorbidity trajectory clustering method and system to solve the problem of low accuracy of existing comorbidity trajectory clustering.

[0005] On the one hand, an embodiment of the present invention provides a method for clustering comorbidity trajectories, comprising the following steps:

[0006] Obtain comorbidity trajectory data and calculate the correlation between comorbidity trajectories;

[0007] constructing an initial trajectory set, and performing binary clustering on the initial trajectory set based on the correlation between the comorbidity trajectories to obtain two subclass trajectory sets;

[0008] For each subclass trajectory set, the attribute of each comorbidity trajectory in the subclass trajectory set is obtained based on the connectivity of the disease chain in the subclass trajectory set, and whether the subclass trajectory set is a trajectory subset to be clustered is determined based on the attribute of the comorbidity trajectory; the attribute includes hub and non-hub;

[0009] If there are trajectory subsets to be clustered, binary nested clustering is performed on each trajectory subset to be clustered; the clustering results of the comorbidity trajectory are obtained.

[0010] Based on the further improvement of the above method, the attributes of each comorbidity trajectory in the subclass trajectory set are obtained based on the connectivity of the disease chain in the subclass trajectory set, including:

[0011] Calculate the connectivity of each disease chain in the subclass trajectory set, and obtain the attributes of each disease chain based on the connectivity of each disease chain;

[0012] For each comorbidity trajectory in the subclass trajectory set, if the comorbidity trajectory covers the hub disease chain, then the comorbidity trajectory is a hub comorbidity trajectory, otherwise the comorbidity trajectory is a non-hub comorbidity trajectory.

[0013] Based on the further improvement of the above method, the connectivity of each disease chain in the subclass trajectory set is calculated, and the attributes of each disease chain are obtained based on the connectivity of each disease chain, including:

[0014] For each disease chain in the subclass trajectory set, the connectivity of the disease chain is the number of comorbidity trajectories in the subclass trajectory set that contain at least one disease in the disease chain;

[0015] If the connectivity of the disease chain is d e ≥q×(|A ij |), then the disease chain is a hub disease chain, otherwise the disease chain is a non-hub disease chain, q represents the proportion parameter, |A ij | represents the subclass trajectory set A ij The number of comorbidity trajectories.

[0016] Based on the further improvement of the above method, a binary nested clustering is performed on each subset of trajectories to be clustered, including:

[0017] S41, for each trajectory subset to be clustered, remove the hub comorbidity trajectory in the trajectory subset to be clustered; if there are unconnected comorbidity trajectories in the trajectory subset to be clustered, remove the unconnected comorbidity trajectories;

[0018] S42, performing binary clustering on each subset of trajectories to be clustered after removing unconnected comorbid trajectories to obtain two subclass trajectory sets corresponding to each subset of trajectories to be clustered;

[0019] S43, obtaining the attribute of each comorbidity trajectory in the subclass trajectory set based on the connectivity of the disease chain in the subclass trajectory set, and judging whether the subclass trajectory set is a trajectory subset to be clustered based on the attribute of the comorbidity trajectory; and taking the trajectory subset not to be clustered as a cluster set in the clustering result;

[0020] S44, if there is a subset of trajectories to be clustered, return to step S41, otherwise, end clustering and execute step S45;

[0021] S45. If there are unconnected comorbidity trajectories, for each unconnected comorbidity trajectory, the cluster to which the unconnected comorbidity trajectory belongs is obtained based on the connectivity between the unconnected comorbidity trajectory and other comorbidity trajectories.

[0022] Based on the further improvement of the above method, the cluster to which the unconnected comorbidity trajectory belongs is obtained based on the connectivity between the unconnected comorbidity trajectory and other comorbidity trajectories, including:

[0023] The correlation between the unconnected comorbidity trajectory and other comorbidity trajectories is calculated by calculating the correlation between the unconnected comorbidity trajectory and each cluster, and the unconnected comorbidity trajectory is assigned to the cluster with the largest correlation.

[0024] Based on the further improvement of the above method, the correlation between comorbidity trajectories was calculated in the following way:

[0025]

[0026] in, represents the ith comorbidity trajectory V i and the jth comorbidity trajectory V j The correlation between represents the ith comorbidity trajectory V i and the jth comorbidity trajectory V j The number of diseases included in common, represents the ith comorbidity trajectory V i and the jth comorbidity trajectory V j The total number of diseases included.

[0027] On the other hand, an embodiment of the present invention provides a comorbidity trajectory clustering system, including the following modules:

[0028] A correlation calculation module is used to obtain comorbidity trajectory data and calculate the correlation between comorbidity trajectories;

[0029] An initial clustering module is used to construct an initial trajectory set, and perform binary clustering on the initial trajectory set based on the correlation between the comorbidity trajectories to obtain two sub-class trajectory sets; for each sub-class trajectory set, the attribute of each comorbidity trajectory in the sub-class trajectory set is obtained based on the connectivity of the disease chain in the sub-class trajectory set, and whether the sub-class trajectory set is a trajectory subset to be clustered is determined based on the attributes of the comorbidity trajectory; the attributes include hubs and non-hubs;

[0030] The binary nested clustering module is used to perform binary nested clustering on each subset of trajectories to be clustered if there are any subsets of trajectories to be clustered, and obtain the clustering results of the comorbidity trajectories.

[0031] Based on the further improvement of the above system, the properties of each comorbidity trajectory in the subclass trajectory set are obtained based on the connectivity of the disease chain in the subclass trajectory set, including:

[0032] Calculate the connectivity of each disease chain in the subclass trajectory set, and obtain the attributes of each disease chain based on the connectivity of each disease chain;

[0033] For each comorbidity trajectory in the subclass trajectory set, if the comorbidity trajectory covers the hub disease chain, then the comorbidity trajectory is a hub comorbidity trajectory, otherwise the comorbidity trajectory is a non-hub comorbidity trajectory.

[0034] Based on the further improvement of the above system, the connectivity of each disease chain in the subclass trajectory set is calculated, and the attributes of each disease chain are obtained based on the connectivity of each disease chain, including:

[0035] For each disease chain in the subclass trajectory set, the connectivity of the disease chain is the number of comorbidity trajectories in the subclass trajectory set that contain at least one disease in the disease chain;

[0036] If the connectivity of the disease chain is d e ≥q×(|A ij |), then the disease chain is a hub disease chain, otherwise the disease chain is a non-hub disease chain, q represents the proportion parameter, |A ij | represents the subclass trajectory set A ij The number of comorbidity trajectories.

[0037] Based on the further improvement of the above system, binary nested clustering is performed on each subset of trajectories to be clustered, including:

[0038] S41, for each trajectory subset to be clustered, remove the hub comorbidity trajectory in the trajectory subset to be clustered; if there are unconnected comorbidity trajectories in the trajectory subset to be clustered, remove the unconnected comorbidity trajectories;

[0039] S42, performing binary clustering on each subset of trajectories to be clustered after removing unconnected comorbid trajectories to obtain two subclass trajectory sets corresponding to each subset of trajectories to be clustered;

[0040] S43, obtaining the attribute of each comorbidity trajectory in the subclass trajectory set based on the connectivity of the disease chain in the subclass trajectory set, and judging whether the subclass trajectory set is a trajectory subset to be clustered based on the attribute of the comorbidity trajectory; and taking the trajectory subset not to be clustered as a cluster set in the clustering result;

[0041] S44, if there is a subset of trajectories to be clustered, return to step S41, otherwise, end clustering and execute step S45;

[0042] S45. If there are unconnected comorbidity trajectories, for each unconnected comorbidity trajectory, the cluster to which the unconnected comorbidity trajectory belongs is obtained based on the connectivity between the unconnected comorbidity trajectory and other comorbidity trajectories.

[0043] Compared with the prior art, the present invention constructs an initial trajectory set by calculating the correlation between comorbidity trajectories, and performs binary classification on the initial trajectory set based on the correlation between trajectories to obtain two sub-class trajectory sets. Based on the attributes of the comorbidity trajectories in the sub-class trajectory set, it is determined whether the set is a trajectory subset to be clustered. The trajectory subset not to be clustered is directly used as a class set in the clustering result, and binary nested clustering is continued for the trajectory subset to be clustered until there is no trajectory subset to be clustered, and the clustering result of the comorbidity trajectory is obtained. Therefore, according to the characteristics of the trajectory structure, the trajectories are directly clustered based on the trajectory, and the hub trajectories across different groups are accurately identified, which overcomes the defects of the traditional method that does not allow overlapping structures between groups and unclear affiliations, and can directly capture the continuity and correlation of disease evolution at the trajectory level, thereby achieving more accurate and stable clustering of comorbidity trajectories. In addition, since the trajectories are directly clustered, when the disease nodes in the trajectory are scattered in different clusters, the final classification of the trajectory does not need to rely on unstable external rules, such as the majority principle, which may cause fuzzy or unreasonable classification of some trajectories. Therefore, the clustering result is more stable and accurate.

[0044] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the entire drawings, the same reference symbols represent the same components;

[0046] Figure 1 This is a flow chart of a method for clustering comorbidity trajectories according to an embodiment of the present invention;

[0047] Figure 2 This is a schematic diagram of clustering results according to an embodiment of the present invention;

[0048] Figure 3 This is a block diagram of a comorbidity trajectory clustering system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0050] Existing technologies often cluster diseases using diseases as nodes. Although they can identify key diseases within a single cluster group, they cannot accurately identify hub diseases connecting different cluster groups, let alone hub comorbidity trajectories across different cluster groups. Traditional non-fuzzy clustering methods do not allow overlapping structures between different groups when dealing with clustering, which to some extent limits the practicality and interpretability of clustering. Although fuzzy clustering methods allow overlap, they lack clear boundaries when defining the affiliation of members and subclasses, resulting in the inability to form clustering scenarios where hub nodes belong to multiple subclasses rather than hub nodes belonging to only a single subclass. Finally, although existing technologies can present a certain degree of hierarchy when displaying cluster hierarchical structures, there is a lack of actual node connections between the levels, making this hierarchy lack practical clinical guidance significance. How to accurately identify and analyze hub disease chains, and how these nodes interweave to affect the clustering process, and ultimately shape a complex clustering structure with overlapping characteristics, remains an unresolved problem.

[0051] A specific embodiment of the present invention discloses a method for clustering comorbidity trajectories, such as Figure 1 As shown, the following steps are included:

[0052] S1. Obtain comorbidity trajectory data and calculate the correlation between comorbidity trajectories;

[0053] S2, constructing an initial trajectory set, and performing binary clustering on the initial trajectory set based on the correlation between the comorbidity trajectories to obtain two subclass trajectory sets;

[0054] S3. For each subclass trajectory set, the attribute of each comorbidity trajectory in the subclass trajectory set is obtained based on the connectivity of the disease chain in the subclass trajectory set, and whether the subclass trajectory set is a trajectory subset to be clustered is determined based on the attribute of the comorbidity trajectory; the attribute includes hub and non-hub;

[0055] S4. If there are trajectory subsets to be clustered, perform binary nested clustering on each trajectory subset to be clustered to obtain the clustering results of the comorbidity trajectory.

[0056] It should be noted that the co-morbidity trajectory is the sequence of disease occurrence, for example, D1→D2→D3, which means that disease D1 develops into disease D2, and then develops into disease D3. The disease chain is a chain consisting of two adjacent diseases in the co-morbidity trajectory and the directed edges between them. For example, the co-morbidity trajectory D1→D2→D3 includes the disease chain D1→D2 and the disease chain D2→D3. The hub disease chain refers to the directed edge that plays a key connecting role in the co-morbidity network. The co-morbidity trajectory containing these hub disease chains is the hub co-morbidity trajectory. The hub co-morbidity trajectory connects multiple co-morbidity trajectory clusters and belongs to multiple clusters at the same time, while the non-hub co-morbidity trajectory belongs to only one cluster.

[0057] Compared with the prior art, the comorbidity trajectory clustering method provided in this embodiment calculates the correlation between comorbidity trajectories to construct an initial trajectory set, performs binary classification on the initial trajectory set based on the correlation between trajectories to obtain two sub-class trajectory sets, and judges whether the set is a trajectory subset to be clustered based on the attributes of the comorbidity trajectories in the sub-class trajectory set. The trajectory subset not to be clustered is directly used as a class set in the clustering result, and binary nested clustering is continued on the trajectory subset to be clustered until there is no trajectory subset to be clustered, and the clustering result of the comorbidity trajectory is obtained, so as to classify the trajectory based on the trajectory structure characteristics. The trajectories are directly clustered to accurately identify hub trajectories across different groups, overcoming the traditional defects of not allowing overlapping structures between groups and unclear affiliations. It can directly capture the continuity and correlation of disease evolution at the trajectory level, thereby achieving more accurate and stable clustering of comorbidity trajectories. Moreover, since the trajectories are clustered directly, there is no problem that when the disease nodes in the trajectory are scattered in different clusters, the final classification of the trajectory needs to rely on unstable external rules, such as the majority principle, which may lead to fuzzy or unreasonable classification of some trajectories. Therefore, the clustering results are more stable and accurate.

[0058] During implementation, comorbidity trajectory data is obtained and the correlation between comorbidity trajectories is calculated; illustratively, the comorbidity trajectories are shown in Table 1.

[0059] Table 1: Disease trajectory table

[0060]

[0061]

[0062] Specifically, the correlation between comorbidity trajectories was calculated in the following way:

[0063]

[0064] in, represents the ith comorbidity trajectory V i and the jth comorbidity trajectory V j The correlation between represents the ith comorbidity trajectory V i and the jth comorbidity trajectory V j The number of diseases included in common, represents the ith comorbidity trajectory V i and the jth comorbidity trajectory V j The total number of diseases included.

[0065] For example, the number of diseases included in the first comorbidity trajectory D1→D2→D3 and the second comorbidity trajectory D1→D4→D5 is 1, and the total number of diseases included is 5, so the correlation is 1 / 5.

[0066] The correlation between any two trajectories is characterized by calculating the correlation. During implementation, a correlation matrix between trajectories can be constructed for subsequent use.

[0067] Afterwards, an initial trajectory set is constructed based on all comorbidity trajectories. It should be noted that if the correlation between a comorbidity trajectory and all other comorbidity trajectories is zero, the comorbidity trajectory is removed from the initial trajectory set.

[0068] During implementation, for each comorbidity trajectory in the initial trajectory set, the correlation between the comorbidity trajectory and each comorbidity trajectory in the initial trajectory set is used as the characteristic data of the comorbidity trajectory, that is, as the input data for clustering, and binary clustering is performed to cluster the initial trajectory set into two categories to obtain two sub-category trajectory sets. During implementation, the k-means clustering algorithm can be used for binary clustering. For example, the initial trajectory set in Table 1 is subjected to binary clustering to obtain two sub-category trajectory sets A 11 and A 12 , A 11 Including comorbidity trajectory V1~V 11 , A 12 Including comorbidity trajectory V 12 ~V 16 .

[0069] For each subclass trajectory set, the attributes of each comorbidity trajectory in the subclass trajectory set are obtained based on the connectivity of the disease chain in the subclass trajectory set, including:

[0070] S31, calculating the connectivity of each disease chain in the subclass trajectory set, and obtaining the attributes of each disease chain based on the connectivity of each disease chain;

[0071] S32. For each comorbidity trajectory in the subclass trajectory set, if the comorbidity trajectory covers a hub disease chain, the comorbidity trajectory is a hub comorbidity trajectory; otherwise, the comorbidity trajectory is a non-hub comorbidity trajectory.

[0072] Specifically, the connectivity of each disease chain in the subclass trajectory set is calculated, and the attributes of each disease chain are obtained based on the connectivity of each disease chain, including:

[0073] For each disease chain in the subclass trajectory set, the connectivity of the disease chain is the number of comorbidity trajectories in the subclass trajectory set that contain at least one disease in the disease chain;

[0074] If the connectivity of the disease chain is d e ≥q×(|A ij |), then the disease chain is a hub disease chain, otherwise the disease chain is a non-hub disease chain, q represents the proportion parameter, |A ij | represents the subclass trajectory set A ij The number of comorbidity trajectories.

[0075] For example, for the disease chain D1→D2, the subclass trajectory set A 11 There are 11 comorbidity trajectories containing disease D1 or disease D2, so the connectivity of the disease chain is 11. For the disease chain D2→D3, the subclass trajectory set A 11 There are 6 comorbidity trajectories that include disease D2 or disease D3, so the connectivity of the disease chain is 6.

[0076] According to the formula d e ≥q×(|A ij |), if the connectivity is greater than q×|A ij |, then the disease chain is a hub disease chain, otherwise it is a non-hub disease chain. The value of q satisfies 0<q≤1. The larger the value of q, the fewer hub disease chains are obtained, and the fewer trajectories are identified as hub trajectories. In implementation, q=0.9 can be taken.

[0077] After determining all hub disease chains in the subclass trajectory set, the comorbidity trajectory in the set that covers the hub disease chain is the hub comorbidity trajectory, thereby obtaining the attributes of each comorbidity trajectory in the set, that is, whether the comorbidity trajectory is a hub comorbidity trajectory or a non-hub comorbidity trajectory. For example, the disease chain D1→D2 is a hub disease chain, and the comorbidity trajectory D1→D2→D3 is a hub comorbidity trajectory.

[0078] If there is a hub co-morbidity trajectory in the subclass trajectory set, the proportion of hub co-morbidity trajectories is less than the first threshold p, and the number of co-morbidity trajectories in the subclass trajectory set is greater than the second threshold τ, then the subclass trajectory set is a trajectory subset to be clustered. If it is not a trajectory subset to be clustered, then the subclass trajectory set is directly used as a cluster in the clustering result.

[0079] The value of the first threshold p satisfies 0<p≤1. The larger the value of p, the larger the range of the hub trajectory in the clustering will be when the subclass terminates the recursion. In implementation, p=1 can be taken. The value of the second threshold τ is an integer greater than 0. It can control the number of trajectories in each category in the trajectory clustering result not to be too small. The larger the value of τ, the larger the scale of the subclass will be when the recursion is terminated. In implementation, τ=10 can be taken. For example, the subclass trajectory set A 12 There is no hub comorbidity trajectory, so it is directly used as a cluster set in the clustering result. 11 There is a hub comorbidity trajectory D1→D2→D3, and the proportion of hub comorbidity trajectories is less than the first threshold, and the number of comorbidity trajectories in the set is greater than the second threshold, so it is a trajectory subset to be clustered.

[0080] If there are trajectory subsets to be clustered, binary nested clustering is performed on each trajectory subset to be clustered, that is, nested recursive clustering is performed until there are no trajectory subsets to be clustered, thereby obtaining the clustering results of the comorbidity trajectory.

[0081] Specifically, binary nested clustering is performed on each subset of trajectories to be clustered, including:

[0082] S41, for each trajectory subset to be clustered, remove the hub comorbidity trajectory in the trajectory subset to be clustered; if there are unconnected comorbidity trajectories in the trajectory subset to be clustered, remove the unconnected comorbidity trajectories;

[0083] S42, performing binary clustering on each subset of trajectories to be clustered after removing unconnected comorbid trajectories to obtain two subclass trajectory sets corresponding to each subset of trajectories to be clustered;

[0084] S43, obtaining the attribute of each comorbidity trajectory in the subclass trajectory set based on the connectivity of the disease chain in the subclass trajectory set, and judging whether the subclass trajectory set is a trajectory subset to be clustered based on the attribute of the comorbidity trajectory; and taking the trajectory subset not to be clustered as a cluster set in the clustering result;

[0085] S44, if there is a subset of trajectories to be clustered, return to step S41, otherwise, end clustering and execute step S45;

[0086] S45. If there are unconnected comorbidity trajectories, for each unconnected comorbidity trajectory, the cluster to which the unconnected comorbidity trajectory belongs is obtained based on the connectivity between the unconnected comorbidity trajectory and other comorbidity trajectories.

[0087] During implementation, the hub comorbidity trajectory in the subset of trajectories to be clustered is first proposed, and then, if there is an unconnected comorbidity trajectory, the unconnected comorbidity trajectory is proposed. It should be noted that if the correlation between a comorbidity trajectory in the subset of trajectories to be clustered and other comorbidity trajectories in the subset of trajectories to be clustered is zero, then the comorbidity trajectory is an unconnected comorbidity trajectory.

[0088] For example, for the trajectory subset A to be clustered 11 , where there is no unconnected comorbidity trajectory, the hub comorbidity trajectory is removed, and the removed trajectory subsets to be clustered are subjected to binary clustering to obtain two subclass trajectory sets A 111 and A 112 , A 111 and A 112 , A 111 Including comorbidity trajectories V2 to V6, A 112 Including comorbidity trajectory V7~V 11 . For each subclass trajectory set, the attributes of each comorbidity trajectory in the subclass trajectory set are obtained based on the connectivity of the disease chain in the subclass trajectory set. The specific process can refer to steps S31 to S32. Then, based on the attributes of the comorbidity trajectory, it is determined whether the subclass trajectory set is a trajectory subset to be clustered. That is, if there is a hub comorbidity trajectory in the subclass trajectory set, the proportion of hub comorbidity trajectories is less than the first threshold p, and the number of comorbidity trajectories in the subclass trajectory set is greater than the second threshold τ, then the subclass trajectory set is a trajectory subset to be clustered. Similarly, the non-clustered trajectory subset is used as a class set in the clustering results; if there is a trajectory subset to be clustered, return to step S41 and continue with the next layer of clustering. If there is no trajectory subset to be clustered, the clustering is terminated. Two subclass trajectory sets A 111 and A 112 are not the subsets of trajectories to be clustered, so the clustering is terminated and the clustering results are obtained, such as Figure 2 As shown in the figure, the trajectories are clustered into three categories, C1, C2 and C3, among which the first category C1 and the second category C2 are connected by the hub comorbidity trajectory V1. That is, the hub trajectory V1 belongs to both the first category C1 and the second category C2.

[0089] For the unconnected comorbidity trajectory removed in step S41, the cluster to which the unconnected comorbidity trajectory belongs is obtained based on the correlation between the unconnected comorbidity trajectory and other comorbidity trajectories, specifically including:

[0090] The correlation between the unconnected comorbidity trajectory and each cluster is calculated based on the correlation between the unconnected comorbidity trajectory and other comorbidity trajectories, and the unconnected comorbidity trajectory is assigned to the cluster with the largest correlation.

[0091] During implementation, the average value of the correlation between the unconnected comorbidity trajectory and the comorbidity trajectory in each cluster is calculated to obtain the correlation between the unconnected comorbidity trajectory and the cluster, so that the unconnected comorbidity trajectory is assigned to the cluster with the greatest correlation.

[0092] Based on the clustering results of the comorbidity trajectories, in-depth research can be conducted on the type groups of each comorbidity trajectory. For example, all trajectories belonging to type C1 can be studied, and the interaction relationships such as the development time between diseases within the type can be analyzed. The comorbidity patterns, key disease development paths and potential common risks unique to type C1 can be identified, so as to develop more personalized and effective intervention plans, such as recommendations on drug treatment, lifestyle intervention, health management, etc., to block or delay the development of comorbidities.

[0093] In addition, the individual disease trajectory of the patient can also be compared with the trajectory clustering results to find out the trajectory type to which the individual disease trajectory of the patient belongs. Referring to the common characteristics of the diseases within the trajectory type group, an integrated diagnosis and treatment plan for the trajectory type group can be formulated to replace the traditional diagnosis and treatment plan for a single disease. For example, if the individual disease trajectory of the patient is D1→D8, and the individual disease trajectory belongs to trajectory type C1, an integrated diagnosis and treatment plan can be formulated based on the common characteristics of the diseases within C1, thereby replacing the traditional diagnosis and treatment plan for a single disease, so as to block or delay the development of comorbidities.

[0094] By performing clustering directly at the trajectory level, the present invention can more comprehensively consider the continuity and dynamics of disease evolution and reduce the errors and uncertainties caused by redistributing trajectories after clustering at the node level. This direct clustering method can more accurately reflect the essential characteristics of co-morbidity trajectories and improve the accuracy and stability of clustering. Secondly, it reveals more complex co-morbidity relationships: Traditional methods have limitations in dealing with complex trajectories containing nodes across clustering boundaries, while the present invention can effectively identify and handle these complex situations. By directly clustering trajectories, the present invention can reveal more detailed and complex co-morbidity relationships, providing a new perspective for a deeper understanding of the interactions between diseases.

[0095] The present invention has broad application prospects and important practical value in the field of medical health. Its new method of clustering directly based on comorbidity trajectories will bring significant beneficial effects to clinical medicine, public health policy formulation, medical research and other aspects. Based on the clustering results of the present invention, clinicians can more accurately identify patient groups with similar health problems, so as to formulate more personalized and accurate treatment plans. In addition, the clustering results based on trajectories can also be used to predict patients' future health risks, provide a basis for early intervention and preventive treatment, and improve the practicality and effectiveness of clinical applications. Through in-depth analysis and clustering of comorbidity trajectories, the present invention can provide more scientific and comprehensive data support for the formulation of public health policies. Based on the clustering results, policymakers can understand the distribution characteristics, influencing factors and development trends of different comorbidity patterns, so as to formulate more targeted and effective public health strategies.

[0096] A specific embodiment of the present invention discloses a comorbidity trajectory clustering system, such as Figure 3 As shown, it includes the following modules:

[0097] A correlation calculation module is used to obtain comorbidity trajectory data and calculate the correlation between comorbidity trajectories;

[0098] An initial clustering module is used to construct an initial trajectory set, and perform binary clustering on the initial trajectory set based on the correlation between the comorbidity trajectories to obtain two sub-class trajectory sets; for each sub-class trajectory set, the attribute of each comorbidity trajectory in the sub-class trajectory set is obtained based on the connectivity of the disease chain in the sub-class trajectory set, and whether the sub-class trajectory set is a trajectory subset to be clustered is determined based on the attributes of the comorbidity trajectory; the attributes include hubs and non-hubs;

[0099] The binary nested clustering module is used to perform binary nested clustering on each subset of trajectories to be clustered if there are any subsets of trajectories to be clustered, and obtain the clustering results of the comorbidity trajectories.

[0100] Preferably, the attributes of each comorbidity trajectory in the subclass trajectory set are obtained based on the connectivity of the disease chain in the subclass trajectory set, including:

[0101] Calculate the connectivity of each disease chain in the subclass trajectory set, and obtain the attributes of each disease chain based on the connectivity of each disease chain;

[0102] For each comorbidity trajectory in the subclass trajectory set, if the comorbidity trajectory covers the hub disease chain, then the comorbidity trajectory is a hub comorbidity trajectory, otherwise the comorbidity trajectory is a non-hub comorbidity trajectory.

[0103] Preferably, the connectivity of each disease chain in the subclass trajectory set is calculated, and the attributes of each disease chain are obtained based on the connectivity of each disease chain, including:

[0104] For each disease chain in the subclass trajectory set, the connectivity of the disease chain is the number of comorbidity trajectories in the subclass trajectory set that contain at least one disease in the disease chain;

[0105] If the connectivity of the disease chain is d e ≥q×(|A ij |), then the disease chain is a hub disease chain, otherwise the disease chain is a non-hub disease chain, q represents the proportion parameter, |A ij | represents the subclass trajectory set A ij The number of comorbidity trajectories.

[0106] Preferably, binary nested clustering is performed on each subset of trajectories to be clustered, including:

[0107] S41, for each trajectory subset to be clustered, remove the hub comorbidity trajectory in the trajectory subset to be clustered; if there are unconnected comorbidity trajectories in the trajectory subset to be clustered, remove the unconnected comorbidity trajectories;

[0108] S42, performing binary clustering on each subset of trajectories to be clustered after removing unconnected comorbid trajectories to obtain two subclass trajectory sets corresponding to each subset of trajectories to be clustered;

[0109] S43, obtaining the attribute of each comorbidity trajectory in the subclass trajectory set based on the connectivity of the disease chain in the subclass trajectory set, and judging whether the subclass trajectory set is a trajectory subset to be clustered based on the attribute of the comorbidity trajectory; and taking the trajectory subset not to be clustered as a cluster set in the clustering result;

[0110] S44, if there is a subset of trajectories to be clustered, return to step S41, otherwise, end clustering and execute step S45;

[0111] S45. If there are unconnected comorbidity trajectories, for each unconnected comorbidity trajectory, the cluster to which the unconnected comorbidity trajectory belongs is obtained based on the connectivity between the unconnected comorbidity trajectory and other comorbidity trajectories.

[0112] The above method embodiments and system embodiments are based on the same principle, and their related parts can be used for reference, and can achieve the same technical effect. The specific implementation process refers to the above embodiment, which will not be repeated here.

[0113] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0114] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A comorbidity trajectory clustering method, characterized in that: The following steps are involved: Obtain comorbidity trajectory data and calculate the correlation between comorbidity trajectories; constructing an initial trajectory set, and performing binary clustering on the initial trajectory set based on the correlation between the comorbidity trajectories to obtain two subclass trajectory sets; For each subclass trajectory set, the attribute of each comorbidity trajectory in the subclass trajectory set is obtained based on the connectivity of the disease chain in the subclass trajectory set, and whether the subclass trajectory set is a trajectory subset to be clustered is determined based on the attribute of the comorbidity trajectory; the attribute includes hub and non-hub; If there are trajectory subsets to be clustered, perform binary nested clustering on each trajectory subset to be clustered; The clustering results of comorbidity trajectories were obtained.

2. The comorbidity trajectory clustering method according to claim 1, characterized in that: Based on the connectivity of the disease chain in the subclass trajectory set, the attributes of each comorbidity trajectory in the subclass trajectory set are obtained, including: Calculate the connectivity of each disease chain in the subclass trajectory set, and obtain the attributes of each disease chain based on the connectivity of each disease chain; For each comorbidity trajectory in the subclass trajectory set, if the comorbidity trajectory covers the hub disease chain, then the comorbidity trajectory is a hub comorbidity trajectory, otherwise the comorbidity trajectory is a non-hub comorbidity trajectory.

3. The comorbidity trajectory clustering method according to claim 2, characterized in that: Calculate the connectivity of each disease chain in the subclass trajectory set, and obtain the properties of each disease chain based on the connectivity of each disease chain, including: For each disease chain in the subclass trajectory set, the connectivity of the disease chain is the number of comorbidity trajectories in the subclass trajectory set that contain at least one disease in the disease chain; If the connectivity of the disease chain is d e ≥q×(|A ij |), then the disease chain is a hub disease chain, otherwise the disease chain is a non-hub disease chain, q represents the proportion parameter, |A ij | represents the subclass trajectory set A ij The number of comorbidity trajectories.

4. The comorbidity trajectory clustering method according to claim 1, characterized in that: Perform binary nested clustering on each subset of trajectories to be clustered, including: S41, for each trajectory subset to be clustered, remove the hub comorbidity trajectory in the trajectory subset to be clustered; if there are unconnected comorbidity trajectories in the trajectory subset to be clustered, remove the unconnected comorbidity trajectories; S42, performing binary clustering on each subset of trajectories to be clustered after removing unconnected comorbid trajectories to obtain two subclass trajectory sets corresponding to each subset of trajectories to be clustered; S43, obtaining the attribute of each comorbidity trajectory in the subclass trajectory set based on the connectivity of the disease chain in the subclass trajectory set, and judging whether the subclass trajectory set is a trajectory subset to be clustered based on the attribute of the comorbidity trajectory; and taking the trajectory subset not to be clustered as a cluster set in the clustering result; S44, if there is a subset of trajectories to be clustered, return to step S41, otherwise, end clustering and execute step S45; S45. If there are unconnected comorbidity trajectories, for each unconnected comorbidity trajectory, the cluster to which the unconnected comorbidity trajectory belongs is obtained based on the connectivity between the unconnected comorbidity trajectory and other comorbidity trajectories.

5. The comorbidity trajectory clustering method according to claim 4, characterized in that: Based on the connectivity between the unconnected comorbidity trajectory and other comorbidity trajectories, the cluster to which the unconnected comorbidity trajectory belongs is obtained, including: The correlation between the unconnected comorbidity trajectory and other comorbidity trajectories is calculated by calculating the correlation between the unconnected comorbidity trajectory and each cluster, and the unconnected comorbidity trajectory is assigned to the cluster with the largest correlation.

6. The comorbidity trajectory clustering method according to claim 1, characterized in that: Correlations between comorbidity trajectories were calculated as follows: in, represents the ith comorbidity trajectory V i and the jth comorbidity trajectory V j The correlation between represents the ith comorbidity trajectory V i and the jth comorbidity trajectory V j The number of diseases included in common, represents the ith comorbidity trajectory V i and the jth comorbidity trajectory V j The total number of diseases included.

7. A comorbidity trajectory clustering system, characterized in that: Includes the following modules: A correlation calculation module is used to obtain comorbidity trajectory data and calculate the correlation between comorbidity trajectories; An initial clustering module is used to construct an initial trajectory set, and to perform binary clustering on the initial trajectory set based on the correlation between the comorbidity trajectories to obtain two sub-class trajectory sets; for each sub-class trajectory set, the attributes of each comorbidity trajectory in the sub-class trajectory set are obtained based on the connectivity of the disease chain in the sub-class trajectory set, and whether the sub-class trajectory set is a trajectory subset to be clustered is determined based on the attributes of the comorbidity trajectory; the attributes include hubs and non-hubs; A binary nested clustering module is used to perform binary nested clustering on each subset of trajectories to be clustered if there are any subsets of trajectories to be clustered; The clustering results of comorbidity trajectories were obtained.

8. The comorbidity trajectory clustering system according to claim 7, characterized in that: Based on the connectivity of the disease chain in the subclass trajectory set, the attributes of each comorbidity trajectory in the subclass trajectory set are obtained, including: Calculate the connectivity of each disease chain in the subclass trajectory set, and obtain the attributes of each disease chain based on the connectivity of each disease chain; For each comorbidity trajectory in the subclass trajectory set, if the comorbidity trajectory covers the hub disease chain, then the comorbidity trajectory is a hub comorbidity trajectory, otherwise the comorbidity trajectory is a non-hub comorbidity trajectory.

9. The comorbidity trajectory clustering system according to claim 8, characterized in that: Calculate the connectivity of each disease chain in the subclass trajectory set, and obtain the properties of each disease chain based on the connectivity of each disease chain, including: For each disease chain in the subclass trajectory set, the connectivity of the disease chain is the number of comorbidity trajectories in the subclass trajectory set that contain at least one disease in the disease chain; If the connectivity of the disease chain is d e ≥q×(|A ij |), then the disease chain is a hub disease chain, otherwise the disease chain is a non-hub disease chain, q represents the proportion parameter, |A ij | represents the subclass trajectory set A ij The number of comorbidity trajectories.

10. The comorbidity trajectory clustering system according to claim 7, characterized in that: Perform binary nested clustering on each subset of trajectories to be clustered, including: S41, for each trajectory subset to be clustered, remove the hub comorbidity trajectory in the trajectory subset to be clustered; if there are unconnected comorbidity trajectories in the trajectory subset to be clustered, remove the unconnected comorbidity trajectories; S42, performing binary clustering on each subset of trajectories to be clustered after removing unconnected comorbid trajectories to obtain two subclass trajectory sets corresponding to each subset of trajectories to be clustered; S43, obtaining the attribute of each comorbidity trajectory in the subclass trajectory set based on the connectivity of the disease chain in the subclass trajectory set, and judging whether the subclass trajectory set is a trajectory subset to be clustered based on the attribute of the comorbidity trajectory; and taking the trajectory subset not to be clustered as a cluster set in the clustering result; S44, if there is a subset of trajectories to be clustered, return to step S41, otherwise, end clustering and execute step S45; S45. If there are unconnected comorbidity trajectories, for each unconnected comorbidity trajectory, the cluster to which the unconnected comorbidity trajectory belongs is obtained based on the connectivity between the unconnected comorbidity trajectory and other comorbidity trajectories.

Images