Clustering method and device for people dying from cardiovascular and cerebrovascular diseases based on DNN and consensus clustering

Abstract

The present invention discloses a method and device for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering, which relates to the field of biological data processing, including: obtaining indicators for clustering, performing multivariate COX regression on the indicators, and screening out analysis indicators; constructing and training a population classification model based on a deep neural network to obtain a trained population classification model; inputting the analysis indicators of each individual in the population to be analyzed into the trained population classification model to obtain classification results to determine the surviving population and the population who died of cardiovascular and cerebrovascular diseases in the population to be analyzed; obtaining the analysis indicators of each individual in the population who died of cardiovascular and cerebrovascular diseases through the last hidden layer output of the trained population classification model, clustering the population who died of cardiovascular and cerebrovascular diseases using consensus clustering based on the advanced features, and obtaining clustering results. The present invention solves the problem of difficulty in identifying different subsets of people who died of cardiovascular and cerebrovascular diseases.

Description

Technical Field

[0001] The present invention relates to the field of biological data processing, and in particular to a method and device for clustering people who died from cardiovascular and cerebrovascular diseases based on DNN and consensus clustering. Background Art

[0002] Cardiovascular and cerebrovascular diseases (collectively referred to as CVDs) constitute a major public health problem, are the leading cause of death worldwide, and significantly reduce quality of life. CVDs are often diagnosed through imaging studies, and the treatment of certain diseases is also guided by these imaging techniques. Although imaging is very useful in screening and monitoring, it often cannot elucidate the underlying causes of CVDs or the pathophysiological differences in individual patients. CVDs are a series of highly heterogeneous diseases that are closely related to diabetes, hypertension, obesity, genetic factors, and other causes. However, in clinical practice, the specific characteristics and causes of each person who dies from CVD may not be fully understood, and there are also differences in survival among people who die from CVD. Summary of the invention

[0003] The purpose of this application is to propose a method and device for clustering people who died from cardiovascular and cerebrovascular diseases based on DNN and consensus clustering in response to the above-mentioned technical problems.

[0004] In a first aspect, the present invention provides a method for clustering people who died from cardiovascular and cerebrovascular diseases based on DNN and consensus clustering, comprising the following steps:

[0005] Obtain indicators for clustering, including demographic indicators, laboratory indicators and questionnaire indicators, select covariates and independent variables from the indicators, perform multivariate COX regression through covariates and independent variables, and screen out analysis indicators, which are indicators related to the death outcome of cardiovascular and cerebrovascular diseases;

[0006] Constructing and training a crowd classification model based on a deep neural network to obtain a trained crowd classification model, wherein the trained crowd classification model includes an input layer, at least one hidden layer, and an output layer;

[0007] Inputting the analysis index of each individual in the population to be analyzed into the trained population classification model, sequentially passing through the input layer, at least one hidden layer and the output layer to obtain the classification result, and determining the surviving population and the population that died of cardiovascular and cerebrovascular diseases in the population to be analyzed according to the classification result;

[0008] The analytical indicators of each individual among the people who died of cardiovascular and cerebrovascular diseases are obtained by outputting high-level features from the last hidden layer of the trained population classification model. Based on the high-level features, consensus clustering is used to cluster the people who died of cardiovascular and cerebrovascular diseases to obtain clustering results.

[0009] Preferably, demographic indicators include: age, gender, race, marital status, and education level;

[0010] Laboratory indicators included: BMI, waist circumference (WC), C-reactive protein (CRP), triglycerides (TG), low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), total cholesterol (TC), glycosylated hemoglobin (HbA1c), fasting blood glucose (FPG), insulin (INS), systolic blood pressure (SBP), diastolic blood pressure (DBP), triglyceride-glucose index (TyG), white blood cell count (WBC), basophil count (BA#), basophil percentage (BA%), eosinophil count (EO#), eosinophil percentage (BA%), lymphocyte count (LY#), lymphocyte percentage (LY%), neutrophil count (NE#), neutrophil percentage (NE%), monocyte count (MO#), monocyte percentage (MO%), mean red blood cell count (RBC), mean leukocyte count (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean red blood cell count (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte MCH, MCV, RBC, HCT, HGB, RDW, PLT, MPV, NLR, ALB, ALT, AST, ALP, BUN, Ca, HCO, GGT, Fe, LDH, P, TP, TBIL, UA, CRE, CL, OSm, GLO, CK;

[0011] The questionnaire indicators included: whether there was hypertension (HTN), high cholesterol (HLP), diabetes (DM), ischemic heart failure (CHF), heart attack (HA), coronary heart disease (CAD), angina pectoris (AP), stroke (Stroke), smoking and drinking status and physical activity index (MET).

[0012] As a preference, covariates and independent variables are selected from the indicators, and multivariate COX regression is performed through the covariates and independent variables to screen out analysis indicators, specifically including:

[0013] Age, gender, race, marital status, education level, smoking and drinking habits were used as covariates, and the remaining 62 indicators were used as independent variables for multivariate COX regression to obtain the p value corresponding to each indicator. Indicators with p values ​​less than 0.05 were selected as analysis indicators, and a total of 46 analysis indicators were included.

[0014] Preferably, the number of neurons in the input layer and the last hidden layer are 64 and 8 respectively.

[0015] As a preferred method, consensus clustering is used based on advanced features to cluster the population that died of cardiovascular and cerebrovascular diseases, and the clustering results are obtained, specifically including:

[0016] The high-level features obtained by outputting the last hidden layer of the trained population classification model through the analytical indicators of each individual in the population who died of cardiovascular and cerebrovascular diseases were input into the ConsensusClusterPlus of the R package. The K-mean clustering method was adopted, in which the K value was set to 2-6, and the Euclidean distance was selected for measurement. After 1000 iterations, the consensus matrix diagram, the consistency cumulative distribution function diagram and the consistency Delta diagram corresponding to each K value were obtained. The contour width was calculated based on the consensus matrix obtained from the consensus matrix diagram, and the silhouette coefficient diagram was drawn. The optimal K value was determined based on the consistency cumulative distribution function diagram, the consistency Delta diagram and the silhouette coefficient diagram. The population who died of cardiovascular and cerebrovascular diseases were divided into K clusters according to the optimal K value, and the clustering result was obtained.

[0017] As a preference, it also includes:

[0018] The survival status of individuals in different clusters was analyzed through survival curves. The unit of the survival curve was month, and the outcome event was death due to cardiovascular and cerebrovascular diseases.

[0019] The indicators of individuals in different clusters were analyzed through cluster heat maps, box plots and bar charts.

[0020] In a second aspect, the present invention provides a device for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering, comprising:

[0021] An indicator screening module is configured to obtain indicators for clustering, including demographic indicators, laboratory indicators and questionnaire indicators, select covariates and independent variables from the indicators, perform multivariate COX regression through the covariates and independent variables, and screen out analysis indicators, which are indicators related to the death outcome of cardiovascular and cerebrovascular diseases;

[0022] A model building module is configured to build and train a crowd classification model based on a deep neural network to obtain a trained crowd classification model, wherein the trained crowd classification model includes an input layer, at least one hidden layer and an output layer;

[0023] A classification module is configured to input the analysis index of each individual in the population to be analyzed into a trained population classification model, sequentially through an input layer, at least one hidden layer and an output layer, to obtain a classification result, and determine the surviving population and the population that died of cardiovascular and cerebrovascular diseases in the population to be analyzed according to the classification result;

[0024] The clustering module is configured to obtain the analytical indicators of each individual in the population who died of cardiovascular and cerebrovascular diseases through the high-level features output by the last hidden layer of the trained population classification model, and cluster the population who died of cardiovascular and cerebrovascular diseases using consensus clustering based on the high-level features to obtain clustering results.

[0025] In a third aspect, the present invention provides an electronic device comprising one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner in the first aspect.

[0026] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any implementation manner in the first aspect.

[0027] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described in any implementation manner in the first aspect.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] (1) The clustering method for people who died from cardiovascular and cerebrovascular diseases based on DNN and consensus clustering proposed in the present invention first uses multivariate COX regression to screen indicators related to the outcome of death from cardiovascular and cerebrovascular diseases, so that the indicators input into the DNN model are all related to death from cardiovascular and cerebrovascular diseases, reducing the interference of other irrelevant indicators.

[0030] (2) The method for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering proposed in the present invention uses high-level features obtained by analyzing indicators through the last hidden layer output of a trained population classification model to perform consensus clustering. The trained population classification model uses DNN, which can capture the complex relationship between multiple features.

[0031] (3) The clustering method for people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering proposed in the present invention clusters people who died of cardiovascular and cerebrovascular diseases into k clusters of people. There are significant differences between the survival curves of each cluster of people, and the UMAP dimensionality reduction diagram also shows that the three clusters are well distinguished. Some indicators between each cluster of people have significant differences. Although they all died of cardiovascular and cerebrovascular diseases, there are still differences in pathological or physiological changes between different subsets. This will help promote the development of precision medicine, as well as pay close attention to the characteristics of the population at an early stage, intervene in time, and improve survival. At the same time, the effect of multiple indicators on identifying different subsets is much better than that of a single indicator, which helps to better discover people with different characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 A schematic diagram of a process of clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering in an embodiment of the present application;

[0034] Figure 2 A UMAP dimensionality reduction graph of high-level features obtained by outputting the last hidden layer of a trained population classification model for analysis indicators of the surviving population and the population dying from cardiovascular and cerebrovascular diseases in the clustering method for the population dying from cardiovascular and cerebrovascular diseases based on DNN and consensus clustering in an embodiment of the present application;

[0035] Figure 3 A schematic diagram of survival curves of various clusters of people who died of cardiovascular and cerebrovascular diseases in a method for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering in an embodiment of the present application;

[0036] Figure 4 A clustering heat map of indicators of each cluster of people who died of cardiovascular and cerebrovascular diseases in the clustering method of people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering in an embodiment of the present application;

[0037] Figure 5 A box plot of the BUN index of each cluster of people who died of cardiovascular and cerebrovascular diseases in the clustering method of people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering in an embodiment of the present application;

[0038] Figure 6A UMAP dimensionality reduction graph of high-level features obtained by outputting the last hidden layer of a trained population classification model for analysis indicators of each cluster of people who died of cardiovascular and cerebrovascular diseases in a clustering method for people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering in an embodiment of the present application;

[0039] Figure 7 This is a result graph of mapping the BUN value in the clustering method for people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering in an embodiment of the present application on a UMAP dimensionality reduction graph of high-level features obtained by outputting the last hidden layer of a trained population classification model for each cluster of people who died of cardiovascular and cerebrovascular diseases;

[0040] Figure 8 This is a schematic diagram of a device for clustering people who died of cardiovascular diseases based on DNN and consensus clustering according to an embodiment of the present application;

[0041] Fig. 9 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Figure 1 A method for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering provided in an embodiment of the present application is shown, comprising the following steps:

[0044] S1. Obtain indicators for clustering, including demographic indicators, laboratory indicators and questionnaire indicators. Select covariates and independent variables from the indicators, perform multivariate COX regression through covariates and independent variables, and screen out analysis indicators. The analysis indicators are indicators related to the death outcome of cardiovascular and cerebrovascular diseases.

[0045] In a specific embodiment, the demographic indicators include: age, gender, race, marital status, and education level;

[0046] Laboratory indicators included: BMI, waist circumference (WC), C-reactive protein (CRP), triglycerides (TG), low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), total cholesterol (TC), glycosylated hemoglobin (HbA1c), fasting blood glucose (FPG), insulin (INS), systolic blood pressure (SBP), diastolic blood pressure (DBP), triglyceride-glucose index (TyG), white blood cell count (WBC), basophil count (BA#), basophil percentage (BA%), eosinophil count (EO#), eosinophil percentage (BA%), lymphocyte count (LY#), lymphocyte percentage (LY%), neutrophil count (NE#), neutrophil percentage (NE%), monocyte count (MO#), monocyte percentage (MO%), mean red blood cell count (RBC), mean leukocyte count (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean red blood cell count (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte count (RBC), mean leukocyte percentage (RBC), mean leukocyte MCH, MCV, RBC, HCT, HGB, RDW, PLT, MPV, NLR, ALB, ALT, AST, ALP, BUN, Ca, HCO, GGT, Fe, LDH, P, TP, TBIL, UA, CRE, CL, OSm, GLO, CK;

[0047] The questionnaire indicators included: whether there was hypertension (HTN), high cholesterol (HLP), diabetes (DM), ischemic heart failure (CHF), heart attack (HA), coronary heart disease (CAD), angina pectoris (AP), stroke (Stroke), smoking and drinking status and physical activity index (MET).

[0048] In a specific embodiment, covariates and independent variables are selected from the indicators, and multivariate COX regression is performed through the covariates and independent variables to screen out analysis indicators, specifically including:

[0049] Age, gender, race, marital status, education level, smoking and drinking habits were used as covariates, and the remaining 62 indicators were used as independent variables for multivariate COX regression to obtain the p value corresponding to each indicator. Indicators with p values ​​less than 0.05 were selected as analysis indicators, and a total of 46 analysis indicators were included.

[0050] Specifically, the corresponding data were obtained from the National Health and Nutrition Examination Survey (NHANES) public database, including demographic indicators, laboratory indicators, and questionnaire indicators. After excluding indicators with too many missing values, there were a total of 69 indicators. Among them, TyG = Ln [TG (mg / dL) × FPG (mg / dL) / 2]; NLR = NE% / LY; MET (MET-min / week) = weekly frequency × duration × MET value of each physical activity.

[0051] Multivariate COX regression was used to select the characteristics related to the outcome of death from cardiovascular and cerebrovascular diseases. Age, gender, race, marital status, education level, smoking and drinking were used as covariates, and the remaining 62 indicators were used as independent variables for multivariate COX regression. p<0.05 indicated that the indicator was related to the outcome of death from cardiovascular and cerebrovascular diseases. Finally, 46 indicators had p<0.05, of which 29 indicators were risk factors and 17 indicators were protective factors. Therefore, these indicators were used as analysis indicators for the next step of analysis. The 46 analysis indicators include: BMI, waist circumference (WC), C-reactive protein (CRP), low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), total cholesterol (TC), glycosylated hemoglobin (HbA1c), fasting blood glucose (FPG), insulin (INS), systolic blood pressure (SBP), triglyceride-glucose index (TyG), white blood cell count (WBC), eosinophil percentage (EO%), lymphocyte count (LY#), lymphocyte percentage (LY%), neutrophil count (NE#), neutrophil percentage (NE%), monocyte count (MO#), red blood cell count (RBC), red blood cell ratio The blood pressure was 404,647,854,,734,764,874 respectively, and the blood pressure was 247,627,734,874 respectively. The blood pressure was 404,647,854,734,874 respectively, and the blood pressure was 247,627,734,874 respectively.

[0052] First, multivariate COX regression is used to screen indicators related to the outcome of death due to cardiovascular and cerebrovascular diseases, so that the indicators included in the deep neural network are all related to the outcome of death due to cardiovascular and cerebrovascular diseases, reducing the interference of other irrelevant indicators.

[0053] S2, constructing and training a crowd classification model based on a deep neural network to obtain a trained crowd classification model, wherein the trained crowd classification model includes an input layer, at least one hidden layer and an output layer.

[0054] In a specific embodiment, the number of neurons in the input layer and the last hidden layer are 64 and 8 respectively.

[0055] Specifically, the 46 selected analysis indicators are used as input data of the deep neural network (DNN) to distinguish between the surviving population and the population that died of cardiovascular and cerebrovascular diseases. The population classification model uses a deep neural network, which includes an input layer, at least one hidden layer and an output layer. The input layer, hidden layer and output layer all use fully connected layers. The input layer and hidden layer use the dropout method. In the process of forward propagation, the activation value of a certain neuron stops working with a certain probability, which can make the model more generalized.

[0056] During the training process of the population classification model, training data is constructed based on 46 analysis indicators of each individual in the surviving population and 46 analysis indicators of each individual in the population who died of cardiovascular and cerebrovascular diseases. The population classification model is trained using the training data. The 46 analysis indicators are input into the population classification model, and the classification result is obtained by passing through the input layer, at least one hidden layer and the output layer in sequence. The classification result is 1 or 0, where 0 represents the surviving population and 1 represents the population who died of cardiovascular and cerebrovascular diseases. A loss function is constructed according to the classification result and the actual classification result. The parameters of the population classification model are optimized and adjusted based on the loss function to obtain a trained population classification model.

[0057] Furthermore, the number of neurons in the last hidden layer of the deep neural network is set to 8, 16 and 32 neurons respectively, and the AUC is used to evaluate the comprehensive performance of the model under the three settings. The AUC results are 0.8888, 0.8876 and 0.8872 respectively. Therefore, the setting corresponding to the highest AUC is selected, and the number of neurons in the last hidden layer is set to 8. Taking a three-layer deep neural network as an example, the deep neural network includes an input layer, a hidden layer and an output layer connected in sequence, and the number of neurons in the input layer is set to 64, the number of neurons in the hidden layer is set to 8, and the number of neurons in the output layer is set to 1. In other embodiments, the number of hidden layers and the number of neurons in the remaining hidden layers can also be adjusted according to actual conditions.

[0058] S3, inputting the analysis index of each individual in the population to be analyzed into the trained population classification model, sequentially passing through the input layer, at least one hidden layer and the output layer to obtain the classification result, and determining the surviving population and the population that died of cardiovascular and cerebrovascular diseases in the population to be analyzed according to the classification result.

[0059] Specifically, since the embodiment of the present application only clusters the population that died of cardiovascular and cerebrovascular diseases, it is necessary to use a trained population classification model to distinguish the surviving population from the population that died of cardiovascular and cerebrovascular diseases in the population to be analyzed. The analysis index of each individual in the population to be analyzed is input into the trained population classification model, and the classification result is output, which can distinguish the surviving population from the population that died of cardiovascular and cerebrovascular diseases in the population to be analyzed.

[0060] 46 analysis indicators are input into the input layer and processed by each hidden layer. The output result of the last hidden layer represents the high-level features extracted from the input data. These high-level features are usually information after multiple layers of nonlinear transformation, which can better capture the complex patterns of the data. The closer to the output layer, the better it can reflect the final prediction results. The above AUC results show that the distinction between the surviving population and the population that died of cardiovascular and cerebrovascular diseases is good. In order to further observe and verify the distinction between the surviving population and the population that died of cardiovascular and cerebrovascular diseases in low-dimensional space using the high-level features output by the last hidden layer in the trained population classification model. In one embodiment, the total number of individuals in the population to be analyzed, including the surviving population and the population that died of cardiovascular and cerebrovascular diseases, is 37,006. The 46 analysis indicators of the 37,006 individuals in the population to be analyzed are input into the trained population classification model, and the high-level features output by the last hidden layer are 8×37,006-dimensional results. The 8×37,006-dimensional results are subjected to UMAP dimensionality reduction. Because the number of surviving people is much greater than the number of people that died of cardiovascular and cerebrovascular diseases, the surviving population is randomly downsampled to the same number of people in the two categories. Figure 2 ,The UMAP dimension reduction results show that the survival population and the death population due to cardiovascular ,cerebrovascular disease ,determination is good in low-dimensional space. Therefore, the high-level features obtained from the ,last hidden layer output in the trained population classification model are selected as the ,input of the next clustering algorithm.

[0061] S4, obtaining the analytical indicators of each individual among the people who died of cardiovascular and cerebrovascular diseases through the high-level features output by the last hidden layer of the trained population classification model, clustering the people who died of cardiovascular and cerebrovascular diseases by consensus clustering based on the high-level features to obtain the clustering results.

[0062] In a specific embodiment, consensus clustering is used based on advanced features to cluster people who died of cardiovascular and cerebrovascular diseases, and clustering results are obtained, which specifically include:

[0063] The high-level features obtained by outputting the last hidden layer of the trained population classification model through the analytical indicators of each individual in the population who died of cardiovascular and cerebrovascular diseases were input into the ConsensusClusterPlus of the R package. The K-mean clustering method was adopted, in which the K value was set to 2-6, and the Euclidean distance was selected for measurement. After 1000 iterations, the consensus matrix diagram, the consistency cumulative distribution function diagram and the consistency Delta diagram corresponding to each K value were obtained. The contour width was calculated based on the consensus matrix obtained from the consensus matrix diagram, and the silhouette coefficient diagram was drawn. The optimal K value was determined based on the consistency cumulative distribution function diagram, the consistency Delta diagram and the silhouette coefficient diagram. The population who died of cardiovascular and cerebrovascular diseases were divided into K clusters according to the optimal K value, and the clustering result was obtained.

[0064] Specifically, the trained population classification model is used to distinguish the surviving population and the population that died of cardiovascular and cerebrovascular diseases in the population to be analyzed. Here, only all individuals in the population that died of cardiovascular and cerebrovascular diseases need to be clustered. The number of individuals in the population that died of cardiovascular and cerebrovascular diseases is 1416. Therefore, the analysis index of each individual in the population that died of cardiovascular and cerebrovascular diseases is an 8×1416-dimensional high-level feature obtained by the last hidden layer output of the trained population classification model. The 8×1416-dimensional result is input into the R package ConsensusClusterPlus, and the K-mean clustering method is used. The Euclidean distance is selected to measure the distance, and 1000 iterations are performed. The K value is set to 2~6, and then the consensus matrix diagram, the consistency cumulative distribution function (CDF) diagram and the consistency Delta diagram corresponding to each K value are obtained. Then, the contour width is calculated and the contour coefficient diagram is drawn according to the consensus matrix obtained in the above consensus matrix diagram. Finally, the optimal K value is determined according to the above result diagram. Finally, in this embodiment, the best K value is determined to be 3, that is, the clustering result is 3 categories, so the people who died of cardiovascular and cerebrovascular diseases can be divided into 3 clusters.

[0065] In the examples of the present application, the distinguishing effects of several single important indicators and 46 analytical indicators are also compared. The results show that the AUC of the 46 analytical indicators is much higher than the AUC of a single important indicator, indicating that multiple indicators and the interrelationships between multiple indicators can better identify different subsets of people who died of cardiovascular and cerebrovascular diseases.

[0066] In a specific embodiment, it also includes:

[0067] The survival status of individuals in different clusters was analyzed through survival curves. The unit of the survival curve was month, and the outcome event was death due to cardiovascular and cerebrovascular diseases.

[0068] The indicators of individuals in different clusters were analyzed through cluster heat maps, box plots and bar charts.

[0069] Specifically, after the optimal K value is determined, the optimal number of clusters is also determined, and then the survival curve is used to analyze the survival of each cluster population. The follow-up time is from the interview date to the end of the outcome event, in months, and the outcome event is death due to cardiovascular and cerebrovascular diseases. Figure 3 It was found that the survival conditions of the three clusters were significantly different, and there were significant differences between the survival curves of each cluster. The Log-rank test results showed that P<0.001.

[0070] Cluster heatmaps, box plots, and bar plots are further used to explore the differences in indicators between clusters. The input data for the cluster heatmap is the original values ​​of the above 46 analysis indicators. The results are as follows: Figure 4 As shown in the figure, the initial observation is that the current medical history is quite different, and other indicators are also somewhat different. Then the next step is to cluster the indicators according to the group: the continuous indicators are drawn with a box plot. If the indicator meets the ANOVA conditions, ANOVA is used to perform an overall difference analysis first. Otherwise, the Kruskal-Wallis test is used. If P<0.05, a pairwise comparison is performed, and Bonferroni is used to correct the p value. Taking blood urea nitrogen (BUN) as an example, the box plot is as follows Figure 5 As shown. The bar graphs were drawn for the categorical indicators, and the chi-square test was used. The steps were the same as above. The above results show the differences in the current medical history, glucose metabolism and lipid metabolism between different subsets. Figure 6 , this application also performs UMAP dimensionality reduction on the 8×1416-dimensional results, and displays the distribution effect of K clusters of people who died of cardiovascular and cerebrovascular diseases in the low-dimensional space. The distribution effect is consistent with the clustering effect. It can be clearly seen that it is divided into K clusters, and from left to right, they are the subset with the best survival curve performance to the subset with the worst survival curve performance. Then, some important eigenvalues ​​are mapped to the UMAP dimensionality reduction diagram, refer to Figure 7 The result of blood urea nitrogen (BUN) in the UMAP dimensionality reduction diagram has the same trend as the box plot. Therefore, the UMAP dimensionality reduction diagram also shows that the three clusters of people are well distinguished, and some indicators between each cluster have significant differences. The results show that although they all die from cardiovascular and cerebrovascular diseases, there are still differences in pathological or physiological changes between different subsets. This will help promote the development of precision medicine and pay close attention to the indicators of the population at an early stage, intervene in time, and improve survival.

[0071] The above steps S1-S4 do not necessarily represent the order between the steps, but are represented by step symbols, and the order between the steps can be adjusted.

[0072] Further references Figure 8 As an implementation of the methods shown in the above figures, the present application provides an embodiment of a device for clustering people who died of cardiovascular diseases based on DNN and consensus clustering. Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0073] The embodiment of the present application provides a device for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering, including:

[0074] The indicator screening module 1 is configured to obtain indicators for clustering, including demographic indicators, laboratory indicators and questionnaire indicators, select covariates and independent variables from the indicators, perform multivariate COX regression through the covariates and independent variables, and screen out analysis indicators, which are indicators related to the death outcome of cardiovascular and cerebrovascular diseases;

[0075] A model building module 2 is configured to build and train a crowd classification model based on a deep neural network to obtain a trained crowd classification model, wherein the trained crowd classification model includes an input layer, at least one hidden layer and an output layer;

[0076] The classification module 3 is configured to input the analysis index of each individual in the population to be analyzed into the trained population classification model, sequentially through the input layer, at least one hidden layer and the output layer, to obtain the classification result, and determine the surviving population and the population who died of cardiovascular and cerebrovascular diseases in the population to be analyzed according to the classification result;

[0077] Clustering module 4 is configured to obtain the analysis indicators of each individual in the population who died of cardiovascular and cerebrovascular diseases through the high-level features output by the last hidden layer of the trained population classification model, and cluster the population who died of cardiovascular and cerebrovascular diseases using consensus clustering based on the high-level features to obtain clustering results.

[0078] Fig. 9 Schematic diagram of the hardware structure of the electronic device provided by the embodiment of the present invention. Fig. 9 As shown, the electronic device of this embodiment includes: a processor 901 and a memory 902; wherein the memory 902 is used to store computer-executable instructions; the processor 901 is used to execute the computer-executable instructions stored in the memory to implement the various steps performed by the electronic device in the above embodiment. For details, please refer to the relevant description in the above method embodiment.

[0079] Optionally, the memory 902 may be independent or integrated with the processor 901 .

[0080] When the memory 902 is independently provided, the electronic device further includes a bus 903 for connecting the memory 902 and the processor 901 .

[0081] The embodiment of the present invention further provides a computer storage medium, in which computer execution instructions are stored. When the processor 901 executes the computer execution instructions, the above method is implemented.

[0082] The embodiment of the present invention further provides a computer program product, including a computer program. When the computer program is executed by the processor 901, the above method is implemented.

[0083] In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0084] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to implement the solution of this embodiment.

[0085] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each module may exist physically separately, or two or more modules may be integrated into one unit. The unit formed by the above modules may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0086] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor 901 to perform some steps of the methods of various embodiments of the present application.

[0087] It should be understood that the processor 901 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or the processor 901 may be any conventional processor 901, etc. The steps of the method disclosed in the invention may be directly embodied in the hardware processor 901 for execution, or may be executed by a combination of hardware and software modules in the processor 901.

[0088] The memory 902 may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk memory, and may also be a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk, or an optical disk.

[0089] The bus 903 may be an Industry Standard Architecture (ISA), a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus 903 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus 903 in the drawings of the present application is not limited to only one bus 903 or one type of bus 903.

[0090] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general or special purpose computer.

[0091] An exemplary storage medium is coupled to the processor 901, so that the processor 901 can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor 901. The processor 901 and the storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor 901 and the storage medium can also exist as discrete components in an electronic device or a main control device.

[0092] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A clustering method for people dying from cardiovascular and cerebrovascular diseases based on DNN and consensus clustering, characterized in that: The following steps are involved: Obtaining indicators for clustering, the indicators include demographic indicators, laboratory indicators and questionnaire indicators, selecting covariates and independent variables from the indicators, performing multivariate COX regression through the covariates and independent variables, and screening out analysis indicators, the analysis indicators are indicators related to the death outcome of cardiovascular and cerebrovascular diseases; Constructing and training a crowd classification model based on a deep neural network to obtain a trained crowd classification model, wherein the trained crowd classification model includes an input layer, at least one hidden layer and an output layer; Inputting the analysis index of each individual in the population to be analyzed into the trained population classification model, sequentially passing through the input layer, at least one hidden layer and the output layer to obtain a classification result, and determining the surviving population and the population that died of cardiovascular and cerebrovascular diseases in the population to be analyzed according to the classification result; Obtaining the analysis index of each individual in the population of people who died of cardiovascular and cerebrovascular diseases by the high-level features output by the last hidden layer of the trained population classification model, clustering the population of people who died of cardiovascular and cerebrovascular diseases by consensus clustering based on the high-level features, and obtaining clustering results, specifically including: The high-level features obtained by outputting the last hidden layer of the trained population classification model through the analysis indicators of each individual in the population that died of cardiovascular and cerebrovascular diseases are input into the ConsensusClusterPlus of the R package, and the K-means clustering method is adopted, wherein the K value is set to 2-6, and the Euclidean distance is selected for measurement, and 1000 iterations are performed to obtain the consensus matrix diagram, the consistency cumulative distribution function diagram, and the consistency Delta diagram corresponding to each K value, calculate the contour width according to the consensus matrix obtained from the consensus matrix diagram, and draw the contour coefficient diagram, determine the optimal K value according to the consistency cumulative distribution function diagram, the consistency Delta diagram, and the contour coefficient diagram, and divide the population that died of cardiovascular and cerebrovascular diseases into K clusters according to the optimal K value, so as to obtain the clustering result.

2. The method for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering according to claim 1, characterized in that: The demographic indicators include: age, gender, race, marital status, and education level; The laboratory parameters included: BMI, waist circumference, C-reactive protein, triglycerides, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, total cholesterol, glycosylated hemoglobin, fasting blood glucose, insulin, systolic blood pressure, diastolic blood pressure, triglyceride-glucose index, white blood cell count, basophil count, basophil percentage, eosinophil count, eosinophil percentage, lymphocyte count, lymphocyte percentage, neutrophil count, neutrophil percentage, monocyte count, count, percentage of monocytes, mean corpuscular hemoglobin content, mean corpuscular volume, red blood cell count, hematocrit, hemoglobin, red blood cell volume distribution width, platelet count, mean platelet volume, neutrophil-to-lymphocyte ratio, albumin, alanine aminotransferase, aspartate aminotransferase, alkaline phosphatase, urea nitrogen, calcium, bicarbonate, gamma-glutamyl transferase, iron, lactate dehydrogenase, phosphorus, total protein, total bilirubin, uric acid, creatinine, sodium, potassium, chloride, osmotic pressure, globulin, creatine kinase; The questionnaire indicators included: whether there was high blood pressure, high cholesterol, diabetes, ischemic heart failure, heart attack, coronary heart disease, angina pectoris, stroke, smoking and drinking status, and physical activity index.

3. The method for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering according to claim 2, characterized in that: Covariates and independent variables are selected from the above indicators, and multivariate COX regression is performed through the covariates and independent variables to screen out analysis indicators, including: Age, gender, race, marital status, education level, smoking and drinking habits among the indicators were used as covariates, and the remaining 62 indicators were used as independent variables for multivariate COX regression to obtain the p value corresponding to each indicator, and the indicators with p value less than 0.05 were selected as analysis indicators. There were 46 analysis indicators in total.

4. The method for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering according to claim 1, characterized in that: The number of neurons in the input layer and the last hidden layer are 64 and 8 respectively.

5. The method for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering according to claim 1, characterized in that: Also includes: The survival status of individuals in different clusters was analyzed by survival curves, where the unit of the survival curves was months and the outcome event was death due to cardiovascular and cerebrovascular diseases; The indicators of individuals in different clusters were analyzed through cluster heat maps, box plots and bar charts.

6. A device for clustering people who died of cardiovascular and cerebrovascular diseases based on DNN and consensus clustering, characterized in that: include: An indicator screening module is configured to obtain indicators for clustering, wherein the indicators include demographic indicators, laboratory indicators and questionnaire indicators, select covariates and independent variables from the indicators, perform multivariate COX regression through the covariates and independent variables, and screen out analysis indicators, wherein the analysis indicators are indicators related to the death outcome of cardiovascular and cerebrovascular diseases; A model building module is configured to build and train a crowd classification model based on a deep neural network to obtain a trained crowd classification model, wherein the trained crowd classification model includes an input layer, at least one hidden layer and an output layer; a classification module configured to input the analysis index of each individual in the population to be analyzed into the trained population classification model, sequentially through the input layer, at least one hidden layer and the output layer, to obtain a classification result, and determine the surviving population and the population that died of cardiovascular and cerebrovascular diseases in the population to be analyzed according to the classification result; The clustering module is configured to obtain the high-level features of the last hidden layer output of the trained population classification model through the analysis index of each individual in the population of deaths due to cardiovascular and cerebrovascular diseases, and cluster the population of deaths due to cardiovascular and cerebrovascular diseases by consensus clustering based on the high-level features to obtain clustering results, which specifically include: The high-level features obtained by outputting the last hidden layer of the trained population classification model through the analysis indicators of each individual in the population that died of cardiovascular and cerebrovascular diseases are input into the ConsensusClusterPlus of the R package, and the K-means clustering method is adopted, wherein the K value is set to 2-6, and the Euclidean distance is selected for measurement, and 1000 iterations are performed to obtain the consensus matrix diagram, the consistency cumulative distribution function diagram, and the consistency Delta diagram corresponding to each K value, calculate the contour width according to the consensus matrix obtained from the consensus matrix diagram, and draw the contour coefficient diagram, determine the optimal K value according to the consistency cumulative distribution function diagram, the consistency Delta diagram, and the contour coefficient diagram, and divide the population that died of cardiovascular and cerebrovascular diseases into K clusters according to the optimal K value, so as to obtain the clustering result.

7. An electronic device comprising: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images