A chest pain center construction method and system
By constructing a causal directed acyclic graph and an improved neural network model with a loss function, the problem of low resource utilization in the traditional construction of chest pain centers was solved, and more accurate prediction of chest pain center types and rational allocation of resources were achieved.
Patent Information
- Application Number
- CN202610409942.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-31
- Publication Date
- 2026-06-26
AI Technical Summary
Traditional methods for constructing chest pain centers lack data support and precise analysis, resulting in low resource utilization, insufficient accuracy of existing model predictions, and an inability to scientifically and rationally plan the types of chest pain centers.
We construct a causal directed acyclic graph and calculate the average causal effect and causal moderating factor. We improve the cross-entropy loss function and combine it with neural network model training to predict the type of chest pain center.
This improved the model's prediction accuracy and resource utilization, ensuring that the construction of chest pain centers better meets local needs, and enhanced the model's generalization ability and convergence speed.
Smart Images
Figure CN122290924A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of chest pain center construction, and particularly relates to a method and system for constructing a chest pain center. Background Technology
[0002] Chest pain is a common and potentially life-threatening symptom, and rapid, accurate diagnosis and treatment are crucial for improving patient outcomes. Establishing chest pain centers can help optimize the treatment process for chest pain patients and improve efficiency. However, different regions have varying population structures, disease incidence rates, and medical resources, making the scientific and rational planning of chest pain center construction types an urgent issue to be addressed.
[0003] Traditional construction methods often rely on experience to determine the type of chest pain center, lacking data support and precise analysis. They fail to comprehensively consider the characteristic factors of the target area, potentially leading to the construction of mismatched chest pain centers in unsuitable areas and ineffective resource utilization. Existing models are mostly based on superficial data correlations, without in-depth exploration of the causal relationships between evaluation indicators. They cannot fully grasp the complex influencing factors of chest pain center construction, thus limiting prediction accuracy. The basic loss functions commonly used in traditional prediction models are not optimized to suit the actual scenario of chest pain center construction, failing to adequately measure the deviation between model predictions and reality, resulting in slow model convergence and insufficient generalization ability.
[0004] Therefore, there is an urgent need to develop a method and system for the construction of chest pain centers that can improve the accuracy of model prediction results, avoid the blindness of traditional experience-based decision-making, make the construction of chest pain centers more in line with local needs, and improve resource utilization. Summary of the Invention
[0005] To address the aforementioned technical problems, this invention provides a method and system for constructing chest pain centers, which can improve the accuracy of model prediction results, avoid the blindness of traditional experience-based decision-making, make the construction of chest pain centers more in line with local needs, and improve resource utilization.
[0006] This invention provides a method for constructing a chest pain center, the method comprising the following steps: S1. Obtain evaluation indicators for each region of the existing chest pain center; S2. Construct a dataset based on the evaluation indicators and the corresponding types of chest pain centers; among which, the types of chest pain centers include primary care and standard versions; S3. Construct the first loss function based on the evaluation indicators; S4. Construct a neural network model and train the neural network model using the dataset until the first loss function converges to obtain a chest pain center construction type prediction model. S5. Obtain the evaluation indicators for the target area and input them into the chest pain center construction type prediction model to obtain the type of chest pain center that should be built in the target area.
[0007] Furthermore, in S1, the evaluation indicators include total population, proportion of elderly population, morbidity rate, mortality rate, and total number of existing medical institutions.
[0008] Furthermore, in S3, the first loss function is constructed based on the evaluation metrics, including: S31. Construct a causal directed acyclic graph based on the evaluation indicators, and calculate the average causal effect of each causal path; S32. Calculate the causal moderating factor based on the average causal effect; S33. Improve the cross-entropy loss function based on the causal adjustment factor to obtain the first loss function.
[0009] Furthermore, in S31, constructing a causal directed acyclic graph based on evaluation metrics includes: S311. Using each evaluation index as a causal node, initially establish undirected edges between every two causal nodes to obtain an initial undirected graph. S312. Perform conditional independence checks on each pair of causal nodes, delete the undirected edges between each pair of conditionally independent causal nodes, and obtain the causal skeleton graph. S313. Determine the direction of the undirected edge between each pair of causal nodes in the causal skeleton graph according to the orientation rule to obtain a causal directed acyclic graph.
[0010] Furthermore, in S31, the average causal effect of each causal path is calculated using the following formula: ; Among them, X j Y represents the evaluation index used as the causal variable in the j-th causal path. j ACE represents the evaluation index used as the outcome variable in the j-th causal path. Xj→Yj Do(X) represents the average causal effect of the j-th causal path. j =x j1 ) indicates that X j The value is set to x j1 Intervention operations, E[Y j |do(X j =x j1 )] indicates that X j The value is set to x j1 At that time, Y j The expected value.
[0011] Furthermore, in S32, the causal moderating factor is calculated based on the average causal effect, using the following formula: ; Among them, w j Let $\mathbf{j}$ represent the causal moderating factor of the j-th causal path, and $\mathbf{ACE}$ represent the maximum value of the average causal effect among all causal paths.
[0012] Furthermore, in S33, the cross-entropy loss function is improved based on the causal adjustment factor to obtain the first loss function, calculated as follows: ; Where L'(θ) represents the first loss function, i represents the i-th region in the dataset, n represents the total number of regions in the dataset, and y i This represents the actual type of chest pain center constructed in the i-th region. denoted by λ, which represents the type of chest pain center constructed in the i-th region predicted by the model, and d, which represents the total number of causal paths.
[0013] The present invention also provides a chest pain center construction system for performing the above-described chest pain center construction method, characterized in that the system includes the following modules: The data acquisition module is used to acquire evaluation indicators for each region of existing chest pain centers. The dataset construction module, connected to the data acquisition module, is used to construct datasets based on evaluation indicators and the corresponding chest pain center types; among which, the chest pain center types include primary care and standard versions; The first loss function construction module is connected to the data acquisition module and is used to construct the first loss function based on the evaluation index. The model building module, connected to the dataset building module and the first loss function building module, is used to build a neural network model. The neural network model is trained using the dataset until the first loss function converges, resulting in a chest pain center construction type prediction model. The output module, connected to the model building module, is used to obtain various evaluation indicators for the target area and input them into the chest pain center construction type prediction model to obtain the type of chest pain center that should be built in the target area.
[0014] The embodiments of the present invention have the following technical effects: This invention constructs a dataset by acquiring evaluation indicators from various regions of existing chest pain centers and trains a predictive model using a neural network model. It comprehensively considers multiple evaluation indicators such as total population, proportion of elderly population, and incidence rate, enabling a more comprehensive analysis of the actual situation in different regions. Based on this objective data, it predicts the type of chest pain center that should be built in the target region, avoiding the blindness of traditional experience-based decision-making and making the construction of chest pain centers more aligned with local needs, thus improving resource utilization. In constructing the first loss function, a detailed causal directed acyclic graph is built, and the average causal effect and causal moderating factor are calculated. This delves into the causal relationships between various evaluation indicators, rather than merely focusing on surface data correlations. It clearly understands how changes in one indicator affect other indicators and ultimately the choice of chest pain center type, improving the model's predictive accuracy. The cross-entropy loss function is improved using the causal moderating factor, resulting in a first loss function that better reflects the actual situation of chest pain center construction prediction. This better measures the deviation between the model's predictions and the actual situation, accelerating the model's convergence speed during neural network training and further improving the model's generalization ability and predictive accuracy. Attached Figure Description
[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0016] Figure 1 This is a flowchart of a method for constructing a chest pain center provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of a chest pain center construction system provided in an embodiment of the present invention. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0018] This invention provides a method for constructing a chest pain center. Figure 1 This is a flowchart of a method for constructing a chest pain center according to an embodiment of the present invention. See also: Figure 1 The method includes the following steps: S1. Obtain evaluation indicators for each region of the existing chest pain center.
[0019] In some embodiments, the evaluation indicators may include total population, proportion of elderly population, morbidity rate, mortality rate, and total number of existing medical institutions; wherein: The total population refers to the total number of people in the region. A larger population per month indicates a greater demand for medical services. The elderly population ratio refers to the proportion of people aged 65 and above in a region. The incidence of acute cardiovascular events is higher among the elderly population, thus creating a greater demand for higher-level chest pain centers. Incidence rates can include the annual incidence of acute coronary syndrome and the annual incidence of ST-segment elevation myocardial infarction. A high incidence rate indicates a need for stronger emergency response capabilities and more comprehensive facilities. In particular, patients with ST-segment elevation myocardial infarction require rapid and effective treatment, which means that a higher-level chest pain center is needed. Mortality rate refers to the average annual mortality rate of cardiovascular diseases. A high mortality rate indicates that the current emergency medical system and resource allocation in the region are inadequate, and a higher-level chest pain center is needed to improve the treatment effect. The total number of existing medical institutions refers to the total number of existing medical institutions in the region. Regions with a small number of medical institutions need to build higher-level chest pain centers to improve the allocation of medical resources.
[0020] S2. Construct a dataset based on the evaluation indicators and the corresponding chest pain center types.
[0021] Chest pain centers are categorized into basic and standard versions.
[0022] In some embodiments, the dataset can be represented by a matrix D, where each row represents a region and the column vectors include evaluation metrics and type labels for chest pain centers. For example, the primary care version can be represented as 0, and the standard version as 1. The expression for matrix D is as follows: ; Where P represents the total population, E represents the proportion of the elderly population, I represents the incidence rate, M represents the mortality rate, F represents the total number of existing medical institutions, T represents the type label of chest pain centers, and n represents the total number of all regions in the dataset.
[0023] In some embodiments, the establishment criteria for a primary-level chest pain center may include: Staffing: At least 5 cardiovascular specialists will be on staff; monthly training sessions will be conducted for all staff to ensure that all medical personnel are familiar with the procedures for treating acute chest pain.
[0024] Equipment and facilities: Must have the capability and related equipment for thrombolytic therapy (such as thrombolytic drugs, portable electrocardiographs, etc.); equipped with basic emergency equipment such as portable electrocardiographs, multi-functional monitors, defibrillators, etc.
[0025] Pre-hospital emergency care system: Establish a cooperative relationship with the local 120 emergency medical service system to ensure that the time from receiving the order to dispatching the ambulance does not exceed 3 minutes; the ambulance should be equipped with basic monitoring and resuscitation conditions (such as electrocardiograph, multi-function monitor, defibrillator, etc.).
[0026] Community Education and Collaboration: Conduct free clinics and health consultations on cardiovascular disease prevention and control in at least two communities each year; implement training programs for at least five primary healthcare institutions in the region and update the training content regularly.
[0027] In some embodiments, the establishment criteria for a standard chest pain center may include: Staffing: At least 10 cardiovascular specialists will be on staff; quarterly training sessions will be conducted to ensure that all medical staff are familiar with the latest guidelines for the treatment of acute chest pain.
[0028] Equipment and facilities: Must have PPCI capability (catheterization lab and related equipment); equipped with advanced emergency medical equipment such as portable ventilators, monitoring equipment with remote real-time transmission function, temporary cardiac pacemakers, and cardiopulmonary resuscitation machines.
[0029] Pre-hospital emergency medical system: achieve seamless connection with the 120 emergency medical service system, ensuring that the time from receiving the order to dispatch does not exceed 2 minutes; the ambulance should be equipped with advanced monitoring and resuscitation conditions (such as portable ventilators, monitoring equipment with remote real-time transmission function, etc.).
[0030] Community education and collaboration: monthly lectures or health consultation events in multiple communities; comprehensive training programs for at least 10 primary healthcare institutions in the region, with regular updates to the training content.
[0031] S3. Construct the first loss function based on the evaluation index.
[0032] In some embodiments, S3 includes the following sub-steps: S31. Construct a causal directed acyclic graph based on the evaluation indicators, and calculate the average causal effect of each causal path.
[0033] In some embodiments, constructing a causal directed acyclic graph based on evaluation metrics includes: S311. Using each evaluation index as a causal node, initially establish undirected edges between every two causal nodes to obtain an initial undirected graph.
[0034] S312. Perform conditional independence checks on each pair of causal nodes, delete the undirected edges between each pair of conditionally independent causal nodes, and obtain the causal skeleton graph.
[0035] In some embodiments, a significance level (e.g., α=0.05) is determined, and a conditional independence test is performed. For each pair of causal nodes, their set of adjacent causal nodes is determined, and a conditional independence test is performed using statistical methods (e.g., the Pearson correlation coefficient method) to determine whether the pair of causal nodes is conditionally independent given the conditions of their adjacent causal nodes. If so, the undirected edge between the pair of causal nodes is deleted, and all causal nodes are traversed to obtain a causal skeleton graph, where the remaining edges represent direct dependencies.
[0036] S313. Determine the direction of the undirected edge between each pair of causal nodes in the causal skeleton graph according to the orientation rule to obtain a causal directed acyclic graph.
[0037] In some embodiments, determining the causal skeleton graph according to the orientation rules can first determine all v-structures (i.e., the structure A→B←C) in the causal skeleton graph and fix the structure of these causal nodes; then, the direction of the edges between other causal nodes is further determined by the Meek rule, and it is ensured that there are no cyclic paths in the graph, thus obtaining a causal directed acyclic graph.
[0038] For example, a causal directed acyclic graph may contain paths such as: P→I←E; This path indicates the impact of total population and the proportion of elderly population on the incidence rate. Both total population and the proportion of elderly population directly affect the incidence rate, suggesting that regions with a larger population or a higher proportion of elderly population may have a higher incidence of acute chest pain. I→M; This path represents the impact of morbidity on mortality. Morbidity directly affects mortality, which means that areas with a high incidence of acute chest pain may also have a higher mortality rate. F→M; This path represents the impact of the total number of existing medical institutions on the mortality rate. The total number of existing medical institutions directly affects the mortality rate, indicating that regions with a larger number of medical institutions are able to provide better emergency medical services, thereby reducing the mortality rate.
[0039] In some embodiments, the average causal effect of each causal path is calculated using the following formula: ; Among them, X j Y represents the evaluation index used as the causal variable in the j-th causal path. j ACE represents the evaluation index used as the outcome variable in the j-th causal path. Xj→Yj Do(X) represents the average causal effect of the j-th causal path. j =x j1 ) indicates that X j The value is set to xj1 Intervention operations, E[Y j |do(X j =x j1 )] indicates that X j The value is set to x j1 At that time, Y j The expected value.
[0040] S32. Calculate the causal moderating factor based on the average causal effect.
[0041] In some embodiments, the calculation formula is as follows: ; Among them, w j Let $\mathbf{j}$ represent the causal moderating factor of the j-th causal path, and $\mathbf{ACE}$ represent the maximum value of the average causal effect among all causal paths.
[0042] S33. Improve the cross-entropy loss function based on the causal adjustment factor to obtain the first loss function.
[0043] In some embodiments, the calculation formula is as follows: ; Where L'(θ) represents the first loss function, i represents the i-th region in the dataset, n represents the total number of regions in the dataset, and y i This represents the actual type of chest pain center constructed in the i-th region. denoted by λ, which represents the type of chest pain center constructed in the i-th region predicted by the model; λ represents the regularization parameter, which can be determined by tuning on the validation set; and d represents the total number of causal paths.
[0044] Traditional machine learning models primarily rely on statistical correlations, which may overlook underlying causal relationships. Introducing causal modifiers can help models better capture these causal relationships, thereby reducing bias caused by ignoring causal structures. Since causal relationships are generally more stable than statistical correlations, models built on causal relationships often exhibit better robustness and generalization ability when data distributions change. By improving the loss function, models can better identify different types of chest pain centers, helping to ensure more rational resource allocation and improve the quality and efficiency of healthcare services.
[0045] By introducing the causal adjustment factor, the loss function not only considers statistical correlation but also incorporates causal relationships, making the loss function more comprehensive and balanced. This helps the model converge faster and achieve better performance. Furthermore, the causal adjustment factor is similar to a regularization term, which can prevent overfitting to some extent and improve the model's generalization ability.
[0046] S4. Construct a neural network model and train it using the dataset until the first loss function converges to obtain a prediction model for the construction type of chest pain centers.
[0047] In some embodiments, the neural network model can be a multilayer perceptron (MLP) model. The dataset is divided into training and validation sets. The number of neurons in the input layer is determined based on the number of evaluation metrics; in this embodiment, there are 5 evaluation metrics, so the number of neurons in the input layer is 5. The number of hidden layers and neurons can be determined experimentally on the validation set; for example, two hidden layers can be set, with 10 and 5 neurons in each layer, respectively. The output layer has 1 neuron and is used to output the predicted chest pain center type (0 or 1). The model's weights and bias parameters are initialized using a random initialization method.
[0048] The neural network model is trained using a training set. During training, evaluation metrics data from the training set are input into the model, and the model outputs the predicted chest pain center type. The loss between the model's predicted value and the true value is calculated based on the first loss function. The model's weights and bias parameters are adjusted using the backpropagation algorithm to reduce the loss value until the first loss function converges, thus obtaining the chest pain center construction type prediction model.
[0049] S5. Obtain the evaluation indicators for the target area and input them into the chest pain center construction type prediction model to obtain the type of chest pain center that should be built in the target area.
[0050] This invention constructs a dataset by acquiring evaluation indicators from various regions of existing chest pain centers and trains a predictive model using a neural network model. It comprehensively considers multiple evaluation indicators such as total population, proportion of elderly population, and incidence rate, enabling a more comprehensive analysis of the actual situation in different regions. Based on this objective data, it predicts the type of chest pain center that should be built in the target region, avoiding the blindness of traditional experience-based decision-making and making the construction of chest pain centers more aligned with local needs, thus improving resource utilization. In constructing the first loss function, a detailed causal directed acyclic graph is built, and the average causal effect and causal moderating factor are calculated. This delves into the causal relationships between various evaluation indicators, rather than merely focusing on surface data correlations. It clearly understands how changes in one indicator affect other indicators and ultimately the choice of chest pain center type, improving the model's predictive accuracy. The cross-entropy loss function is improved using the causal moderating factor, resulting in a first loss function that better reflects the actual situation of chest pain center construction prediction. This better measures the deviation between the model's predictions and the actual situation, accelerating the model's convergence speed during neural network training and further improving the model's generalization ability and predictive accuracy.
[0051] This invention also provides a chest pain center construction system for performing the aforementioned chest pain center construction method. Figure 2 This is a schematic diagram of a chest pain center construction system provided in an embodiment of the present invention. See also: Figure 2 The system includes the following modules: The data acquisition module is used to acquire evaluation indicators for each region of existing chest pain centers. The dataset construction module, connected to the data acquisition module, is used to construct datasets based on evaluation indicators and the corresponding chest pain center types; among which, the chest pain center types include primary care and standard versions; The first loss function construction module is connected to the data acquisition module and is used to construct the first loss function based on the evaluation index. The model building module, connected to the dataset building module and the first loss function building module, is used to build a neural network model. The neural network model is trained using the dataset until the first loss function converges, resulting in a chest pain center construction type prediction model. The output module, connected to the model building module, is used to obtain various evaluation indicators for the target area and input them into the chest pain center construction type prediction model to obtain the type of chest pain center that should be built in the target area.
[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.
Claims
1. A method for constructing a chest pain center, characterized in that, The method includes the following steps: S1. Obtain evaluation indicators for each region of the existing chest pain center; S2. Construct a dataset based on the evaluation indicators and the corresponding types of chest pain centers; wherein, the types of chest pain centers include primary care and standard versions; S3. Construct the first loss function based on the evaluation indicators; S4. Construct a neural network model and train the neural network model using the dataset until the first loss function converges to obtain a chest pain center construction type prediction model. S5. Obtain the evaluation indicators of the target area and input them into the chest pain center construction type prediction model to obtain the type of chest pain center that should be built in the target area.
2. The method for constructing a chest pain center according to claim 1, characterized in that, In S1, the evaluation indicators include total population, proportion of elderly population, morbidity rate, mortality rate, and total number of existing medical institutions.
3. The method for constructing a chest pain center according to claim 1, characterized in that, In step S3, constructing the first loss function based on the evaluation metrics includes: S31. Construct a causal directed acyclic graph based on the evaluation indicators, and calculate the average causal effect of each causal path; S32. Calculate the causal moderating factor based on the average causal effect; S33. Improve the cross-entropy loss function based on the causal adjustment factor to obtain the first loss function.
4. The method for constructing a chest pain center according to claim 3, characterized in that, In step S31, constructing a causal directed acyclic graph based on evaluation metrics includes: S311. Using each evaluation index as a causal node, initially establish undirected edges between every two causal nodes to obtain an initial undirected graph. S312. Perform conditional independence checks on each pair of causal nodes, delete the undirected edges between each pair of conditionally independent causal nodes, and obtain the causal skeleton graph. S313. Determine the direction of the undirected edge between each pair of causal nodes in the causal skeleton graph according to the orientation rule to obtain a causal directed acyclic graph.
5. A method for constructing a chest pain center according to claim 3, characterized in that, In step S31, the average causal effect of each causal path is calculated using the following formula: ； Among them, X j Y represents the evaluation index used as the causal variable in the j-th causal path. j ACE represents the evaluation index used as the outcome variable in the j-th causal path. Xj→Yj Do(X) represents the average causal effect of the j-th causal path. j =x j1 ) indicates that X j The value is set to x j1 Intervention operations, E[Y j |do(X j =x j1 )] indicates that X j The value is set to x j1 At that time, Y j The expected value.
6. A method for constructing a chest pain center according to claim 5, characterized in that, In step S32, the causal moderating factor is calculated based on the average causal effect, using the following formula: ； Among them, w j Let $\mathbf{j}$ represent the causal moderating factor of the j-th causal path, and $\mathbf{ACE}$ represent the maximum value of the average causal effect among all causal paths.
7. A method for constructing a chest pain center according to claim 6, characterized in that, In step S33, the cross-entropy loss function is improved based on the causal adjustment factor to obtain the first loss function, calculated as follows: ； Where L'(θ) represents the first loss function, i represents the i-th region in the dataset, n represents the total number of regions in the dataset, and y i This represents the actual type of chest pain center constructed in the i-th region. denoted by λ, which represents the type of chest pain center constructed in the i-th region predicted by the model, and d, which represents the total number of causal paths.
8. A chest pain center construction system, used to execute the chest pain center construction method according to any one of claims 1-7, characterized in that, The system includes the following modules: The data acquisition module is used to acquire evaluation indicators for each region of existing chest pain centers. A dataset construction module, connected to the data acquisition module, is used to construct a dataset based on evaluation indicators and the corresponding chest pain center type; wherein, the chest pain center type includes a primary care version and a standard version; The first loss function construction module is connected to the data acquisition module and is used to construct the first loss function based on the evaluation index. The model building module, connected to the dataset building module and the first loss function building module, is used to build a neural network model and train the neural network model using the dataset until the first loss function converges, thereby obtaining a chest pain center construction type prediction model. The output module, connected to the model building module, is used to obtain various evaluation indicators of the target area and input them into the chest pain center construction type prediction model to obtain the type of chest pain center that should be built in the target area.