Diagnostic model for sudden cardiac death and construction method thereof

Abstract

The invention relates to the field of medicine, and provides a sudden cardiac death diagnosis model and a construction method thereof, and the method comprises the steps: collecting data samples, and constructing a training set based on the data samples, the training set comprises diagnosis features, and the diagnosis features represent that the data samples in the training set are sudden cardiac death; performing data interpolation and variable screening on the diagnosis features in the training set; a logistic regression model is constructed based on the training set, independent variables of the logistic regression model are diagnosis features, and dependent variables of the logistic regression model are diagnosis results of samples corresponding to the diagnosis features; and constructing a diagnosis calculator based on the logistic regression model, wherein the diagnosis calculator inputs the received characteristic values into the logistic regression model to diagnose sudden cardiac death. The sudden cardiac death diagnosis method and device are used for overcoming the defect that a diagnosis result is not accurate enough mainly depending on subjective experience of a forensic when sudden cardiac death diagnosis is carried out at present, and the sudden cardiac death diagnosis method and device can carry out accurate diagnosis on sudden cardiac death.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a diagnostic model for sudden cardiac death and a construction method thereof. Background Art

[0002] Sudden cardiac death is an unpredictable, acute natural death caused by cardiovascular disease. About 5 million cases of sudden cardiac death occur worldwide each year, accounting for about 50% of deaths from cardiovascular disease. Due to the suddenness and unpredictability of sudden cardiac death, its diagnosis is also a very difficult task.

[0003] In current diagnostic practice, the diagnosis of sudden cardiac death in corpses is mainly made based on the subjective experience of the forensic doctor after the autopsy. However, such results are highly accidental and may be affected by a variety of subjective and objective factors, resulting in inaccurate diagnostic results. Summary of the invention

[0004] The present invention provides a diagnostic model for sudden cardiac death and a construction method thereof, so as to solve the problem that the current diagnosis of sudden cardiac death mainly relies on the subjective experience of forensic doctors to give diagnostic results, and the results are not accurate enough. The scheme of the present application can construct a logistic regression model based on sample data, and obtain the input feature value through a diagnostic calculator and call the logistic regression model, so as to accurately diagnose sudden cardiac death.

[0005] The present invention provides a method for constructing a diagnostic model for sudden cardiac death, comprising:

[0006] Collecting data samples, and constructing a training set based on the data samples, wherein the training set includes diagnostic features, and the diagnostic features characterize that the data samples in the training set are sudden cardiac death;

[0007] Performing data interpolation and variable screening on the diagnostic features in the training set;

[0008] Building a logistic regression model based on the training set, wherein the independent variable of the logistic regression model is the diagnostic feature, and the dependent variable of the logistic regression model is the diagnostic result of the sample corresponding to the diagnostic feature;

[0009] A diagnostic calculator is constructed based on the logistic regression model, and the diagnostic calculator inputs the received feature values ​​into the logistic regression model to diagnose sudden cardiac death.

[0010] According to the method for constructing a diagnostic model for sudden cardiac death provided by the present invention, the diagnostic features include anatomical data and pathological analysis results of the data sample;

[0011] The anatomical data includes the heart weight of the data sample, and / or the left ventricular wall thickness, and / or the right ventricular wall thickness, and / or the mitral valve circumference, and / or the aortic valve circumference, and / or the tricuspid valve circumference, and / or the pulmonary valve circumference, and / or the stenosis degree of the left anterior descending artery, and / or the stenosis degree of the left circumflex artery, and / or the stenosis degree of the right coronary artery;

[0012] The pathological analysis results include myocardial infarction, and / or cardiomyopathy, and / or conduction system disease, and / or valvulitis, and / or epicarditis, and / or myocarditis, and / or endocarditis and pericarditis.

[0013] The diagnostic features also include the height of the data sample, and / or the thickness of the subcutaneous fat of the abdominal wall, and / or the gender, and / or the age.

[0014] According to the method for constructing a diagnostic model for sudden cardiac death provided by the present invention, the anatomical data includes the stenosis degree of the left anterior descending artery, the stenosis degree of the left circumflex artery, the stenosis degree of the right coronary artery, the left ventricular wall thickness and the circumference of the pulmonary valve of the data sample.

[0015] The pathological analysis results include myocardial infarction, cardiomyopathy, valvulitis, pericarditis, myocarditis, and conduction system lesions.

[0016] According to the method for constructing a diagnostic model for sudden cardiac death provided by the present invention, the diagnostic feature also includes the age of the data sample.

[0017] According to the method for constructing a diagnostic model for sudden cardiac death provided by the present invention, the data interpolation and variable screening of the diagnostic features in the training set include:

[0018] Using the Lost Forest interpolation method to interpolate data on the diagnostic features in the training set;

[0019] The LASSO regression method with ten-fold cross validation and the recursive feature elimination method based on random forest were used to screen the variables of the diagnostic features in the training set.

[0020] According to the method for constructing a diagnostic model for sudden cardiac death provided by the present invention, the LASSO regression method with ten-fold cross validation and the recursive feature elimination method based on random forest are used to perform variable screening on the diagnostic features in the training set, comprising:

[0021] Using the LASSO regression method with ten-fold cross validation to screen the diagnostic features in the training set to obtain a first variable set;

[0022] Using a recursive feature elimination method based on random forest to perform variable screening on the diagnostic features in the training set to obtain a second variable set;

[0023] The first variable set and the second variable set are intersected to obtain the result of variable screening.

[0024] According to the method for constructing a diagnostic model for sudden cardiac death provided by the present invention, the diagnostic calculator inputs the received characteristic values ​​into the logistic regression model to diagnose sudden cardiac death, including:

[0025] Obtain the diagnostic characteristics of the body to be tested through the input box;

[0026] Inputting the diagnostic characteristics of the corpse to be tested into the logistic regression model;

[0027] Calculating the diagnostic probability corresponding to the diagnostic feature through the Norman diagram derived from the logistic regression model, wherein the diagnostic probability reflects the probability that the corpse to be tested is diagnosed as sudden cardiac death;

[0028] If the diagnosis probability corresponding to the diagnostic feature is higher than the pre-calculated critical probability value, it is determined that the corpse to be tested is suffering from sudden cardiac death.

[0029] According to the method for constructing a diagnostic model for sudden cardiac death provided by the present invention, the critical probability value is calculated in the following manner:

[0030] Calculating the Youden Index of the training set;

[0031] The critical probability value is determined based on the maximum value of the Youden Index.

[0032] The present invention also provides a system for constructing a diagnostic model for sudden cardiac death, comprising:

[0033] A data acquisition module, used for collecting data samples and constructing a training set based on the data samples, wherein the training set includes diagnostic features, and the diagnostic features indicate that the data samples in the training set are sudden cardiac death;

[0034] A data processing module, used for performing data interpolation and variable screening on the diagnostic features in the training set;

[0035] A model building module, used to build a logistic regression model based on the training set, wherein the independent variable of the logistic regression model is the diagnostic feature, and the dependent variable of the logistic regression model is the diagnostic result of the sample corresponding to the diagnostic feature;

[0036] The model diagnosis module is used to construct a diagnosis calculator based on the logistic regression model, and the diagnosis calculator inputs the received feature values ​​into the logistic regression model to diagnose sudden cardiac death.

[0037] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for constructing a diagnostic model for sudden cardiac death as described above is implemented.

[0038] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for constructing a diagnostic model for sudden cardiac death.

[0039] The present invention also provides a computer program product, including a computer program, which implements any of the above-mentioned methods for constructing a diagnostic model for sudden cardiac death when executed by a processor.

[0040] In the construction method provided by the present invention, a training set can be constructed with data samples characterizing sudden cardiac death. By constructing a special training set in this way, the accuracy of the training process can be improved. In addition, data interpolation and variable screening can be used to improve the completeness of the data in the training set and simplify the training set to simplify the difficulty of training. After the training set is processed, a logistic regression model and a diagnostic calculator can be constructed. After the diagnostic calculator obtains various characteristic values ​​of the corpse input by the operator through a graphical user interface, they can be input into the logistic regression model. In the logistic regression model, a diagnosis result of sudden cardiac death of the corpse can be obtained based on the input various diagnostic characteristic values. In this process, forensic personnel do not need to make judgments based on experience, which basically avoids the phenomenon that the diagnosis result is inaccurate due to the influence of external subjective and objective factors, and can significantly improve the accuracy of diagnosis of sudden cardiac death. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are 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.

[0042] Figure 1 This is one of the flow charts of the method for constructing a diagnostic model for sudden cardiac death provided by the present invention;

[0043] Figure 2 is a schematic diagram of a Norman diagram provided by an embodiment of the present invention;

[0044] Figure 3 is a schematic diagram of the structure of a system for constructing a diagnostic model for sudden cardiac death provided by an embodiment of the present invention;

[0045] Figure 4 A schematic diagram of performance test results provided by an embodiment of the present invention;

[0046] Figure 5 A schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of 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.

[0048] Figure 1 This is one of the flow charts of the method for constructing a diagnostic model for sudden cardiac death provided by the present invention.

[0049] like Figure 1 As shown, this embodiment provides a method for constructing a diagnostic model for sudden cardiac death, comprising:

[0050] Step 101, collecting data samples, and constructing a training set based on the data samples, wherein the training set includes diagnostic features, and the diagnostic features indicate that the data samples in the training set are sudden cardiac death;

[0051] Step 102, performing data interpolation and variable screening on the diagnostic features in the training set;

[0052] Step 103, constructing a logistic regression model based on the training set, wherein the independent variable of the logistic regression model is the diagnostic feature, and the dependent variable of the logistic regression model is the diagnostic result of the sample corresponding to the diagnostic feature;

[0053] Step 104: construct a diagnosis calculator based on the logistic regression model, and the diagnosis calculator inputs the received feature values ​​into the logistic regression model to diagnose sudden cardiac death.

[0054] In implementation, the data sample in this embodiment may be a cause of death identification result obtained by forensic identification. Furthermore, the cause of death identification result of the data sample shows that the cause of death of the corpse is sudden cardiac death.

[0055] In practical applications, the data samples collected in this embodiment may be several death cause identification cases collected by the Forensic Identification Center of Sun Yat-sen University in the past period of time. The diagnostic features may include characteristic data closely related to the diagnosis of sudden cardiac death, such as the morphology of cardiac anatomy, histological parameters, and demographic information. Through these diagnostic features, the various characteristics of sudden cardiac death can be comprehensively summarized, so that the trained logistic regression model can be more accurate.

[0056] The diagnostic features in this embodiment refer to features that can be used to diagnose sudden cardiac death, or features that are helpful for the diagnosis of sudden cardiac death.

[0057] Data interpolation refers to the process of inferring missing data values ​​through specific algorithms when the values ​​of certain data points are known. This operation can improve the integrity of the data. A complete training set can provide more comprehensive information, reduce analytical bias caused by incomplete data, and improve the reliability and accuracy of the constructed logistic regression model.

[0058] In practical applications, the constructed logistic regression model can use the diagnostic features of the training set as independent variables, and the dependent variable can be the diagnostic result of sudden cardiac death. Specifically, the diagnostic result can be the probability of dying from sudden cardiac death. For example, by inputting the various diagnostic features of a sample into the logistic regression model, the logistic regression model can determine that the probability of the sample dying from sudden cardiac death is 78%, or determine the probability is 12%, and so on.

[0059] A diagnostic calculator is also constructed in the scheme of this embodiment. The diagnostic calculator may have a graphical user interface through which the operator's input can be accepted. The forensic operator can input various diagnostic features of the corpse to be tested in the diagnostic calculator. In actual applications, the graphical user interface of the diagnostic calculator may indicate the diagnostic features that need to be input by components such as sliding bars, drop-down boxes, and buttons. In implementation, the forensic operator may, based on actual conditions, input only known diagnostic features into the diagnostic calculator when some diagnostic features are unavailable, and perform a conservative diagnosis of sudden cardiac death based on the input diagnostic features. It should be noted that the more and more comprehensive the diagnostic features input, the more accurate the diagnostic result will naturally be. However, this method in the scheme of this embodiment can cope with various special situations. For example, in actual applications, the case information is incomplete or the specimen is partially damaged, resulting in the inability to collect certain feature information. In this case, the diagnosis of sudden cardiac death can also be performed, thereby improving the scope of application of the diagnostic model.

[0060] The following is an introduction to how to use the diagnostic calculator:

[0061] First, enter the diagnostic features through the drop-down box and slider in the diagnostic calculator;

[0062] Click the "Predict" button in the diagnostic calculator to output the probability value of the corresponding sample being sudden cardiac death and its 95% confidence interval;

[0063] In the "Model Performance Overview" page of the diagnostic calculator, you can find the cut-off value closest to the output probability, and determine the confidence of diagnosing the sample as sudden cardiac death based on the parameters corresponding to the cut-off value. The corresponding parameters may include accuracy, sensitivity, specificity, positive predictive value, negative predictive value, etc.

[0064] In addition, in the construction method provided in this embodiment, a training set can be constructed with data samples that characterize sudden cardiac death. By constructing a special training set in this way, the accuracy of the training process can be improved. In addition, the completeness of the data in the training set can be improved by data interpolation. After the training set processing is completed, a logistic regression model and a diagnostic calculator can be constructed. After the diagnostic calculator obtains the various feature values ​​of the corpse input by the operator through a graphical user interface, it can be input into the logistic regression model. In the logistic regression model, a diagnosis result of sudden cardiac death of the corpse can be obtained based on the input feature values. In this process, forensic personnel do not need to make judgments based on experience, which basically avoids the phenomenon of inaccurate diagnostic results caused by external subjective and objective factors, and can significantly improve the diagnostic accuracy of sudden cardiac death.

[0065] In an exemplary embodiment, the diagnostic features include anatomical data and pathological analysis results of the data sample;

[0066] The anatomical data includes the heart weight of the data sample, and / or the left ventricular wall thickness, and / or the right ventricular wall thickness, and / or the mitral valve circumference, and / or the aortic valve circumference, and / or the tricuspid valve circumference, and / or the pulmonary valve circumference, and / or the stenosis degree of the left anterior descending artery, and / or the stenosis degree of the left circumflex artery, and / or the stenosis degree of the right coronary artery;

[0067] The pathological analysis results include myocardial infarction, and / or cardiomyopathy, and / or conduction system disease, and / or valvulitis, and / or epicarditis, and / or myocarditis, and / or endocarditis and pericarditis.

[0068] In practical applications, the degree of stenosis refers to the percentage of the stenotic part of the lumen in the lumen, that is, the degree of stenosis = 1 - the percentage of the normal part of the lumen in the lumen.

[0069] In an exemplary embodiment, the diagnostic features further include height, and / or thickness of abdominal wall subcutaneous fat, and / or gender, and / or age of the data sample.

[0070] Left ventricular wall thickness (LVWT), right ventricular wall thickness (RVWT), circumference of mitral annulus (cMA), circumference of aortic annulus (cAA), circumference of tricuspid annulus (cTA), circumference of pulmonary annulus (cPA), as well as the degree of stenosis of the left anterior descending branch (LAD), the degree of stenosis of the left circumflex branch (LCX), the degree of stenosis of the right coronary artery (RCA), the thickness of subcutaneous fat of the abdominal wall, gender, and age are all key information for diagnosing sudden cardiac death.

[0071] In addition, myocardial infarction (MI), cardiomyopathy (CM), conduction systematic lesions (CSL), valvulitis, epicarditis, myocarditis, endocarditis and pericarditis in the pathological analysis results are also within the diagnostic scope of sudden cardiac death.

[0072] In the solution of this embodiment, the anatomical data, pathological analysis result data and demographic data can more comprehensively summarize the diagnostic characteristics of sudden cardiac death and improve the accuracy of diagnosis.

[0073] In an exemplary embodiment, the anatomical data includes the stenosis degree of the left anterior descending artery, the stenosis degree of the left circumflex artery, the stenosis degree of the right coronary artery, the left ventricular wall thickness, and the circumference of the pulmonary valve of the data sample.

[0074] The pathological analysis results include myocardial infarction, cardiomyopathy, valvulitis, pericarditis, myocarditis, and conduction system lesions.

[0075] The demographic data includes age.

[0076] The anatomical data, pathological analysis results and demographic data provided in this embodiment are certain more representative diagnostic features screened and determined from a number of diagnostic features. That is to say, the diagnostic features provided in this embodiment are more helpful for the diagnosis of sudden cardiac death, or in other words, the diagnostic features provided in this embodiment have a greater proportional weight when diagnosing sudden cardiac death.

[0077] The diagnostic features provided in this embodiment can assist in the diagnosis of sudden cardiac death.

[0078] In an exemplary embodiment, the data interpolation and variable screening of the diagnostic features in the training set include:

[0079] Using the Lost Forest interpolation method to interpolate data on the diagnostic features in the training set;

[0080] The LASSO regression method with ten-fold cross validation and the recursive feature elimination method based on random forest were used to screen the variables of the diagnostic features in the training set.

[0081] Random forest is an algorithm that integrates multiple trees through the idea of ​​ensemble learning. The basic unit is the decision tree, and its essence belongs to the ensemble learning method in machine learning. The essence of the random forest algorithm is a classifier integration algorithm based on decision trees. Random forest randomizes the column variables and row observations of the data set, generates multiple classification numbers, and finally summarizes the results of multiple classification numbers. When performing data interpolation, the random forest algorithm reduces the amount of calculation while significantly improving the prediction accuracy compared to the traditional neural network algorithm. Moreover, the algorithm is insensitive to multicollinearity and is relatively robust to missing data and unbalanced data, and can adapt well to data sets with up to thousands of explanatory variables.

[0082] LASSO regression, the full name of which is Least Absolute Shrinkage and Selection Operator, is a penalized regression method. It adds an L1 regularization term to the loss function to reduce some coefficients to 0, thereby performing feature selection and model simplification. The role of the L1 regularization term is to penalize the absolute value of the coefficients in the model, which helps to reduce the complexity of the model and prevent overfitting. Ten-fold cross validation is a method to evaluate model performance. It divides the data set into 10 parts, uses 9 of them as training sets and 1 as a test set in turn, and conducts 10 experiments. Each experiment will produce a corresponding accuracy or error rate, and finally take the average of these 10 results as an estimate of the accuracy of the algorithm. This method can reduce the risk of overfitting and improve the generalization ability of the model.

[0083] In the scheme of this embodiment, the data set is interpolated by the Lost Forest algorithm, and variable screening is completed by combining the recursive feature elimination method based on random forest and the LASSO regression method with ten-fold cross validation, which can combine the advantages of the two algorithms and ultimately improve the performance of the diagnostic model.

[0084] In an exemplary embodiment, the LASSO regression method with ten-fold cross validation and the recursive feature elimination method based on random forest are used to perform variable screening on the diagnostic features in the training set, including:

[0085] Using the LASSO regression method with ten-fold cross validation to screen the diagnostic features in the training set to obtain a first variable set;

[0086] Using a recursive feature elimination method based on random forest to perform variable screening on the diagnostic features in the training set to obtain a second variable set;

[0087] The first variable set and the second variable set are intersected to obtain the result of variable screening.

[0088] In the scheme of this embodiment, when performing variable screening, the first variable set and the second variable set can be obtained based on the LASSO regression method of ten-fold cross validation and the recursive feature elimination method based on random forest, that is, two algorithms are used to perform variable screening respectively, and then the results obtained by the two screenings are comprehensively summarized to obtain the final variable screening result. The method for summarizing the two algorithms can be to take the intersection, and the process of taking the intersection is equivalent to further variable screening, that is to say, the final variable screening result is the data screened by the two algorithms at the same time. This method can improve the stability of variable screening.

[0089] In an exemplary embodiment, the diagnostic calculator inputs the received feature value into the logistic regression model to diagnose sudden cardiac death, including:

[0090] Obtain diagnostic characteristics of the body to be tested;

[0091] Inputting the diagnostic characteristics of the corpse to be tested into the logistic regression model;

[0092] Calculating the diagnostic probability corresponding to the diagnostic feature through the Norman diagram in the logistic regression model, wherein the diagnostic probability reflects the probability that the corpse to be tested is diagnosed as sudden cardiac death;

[0093] If the diagnosis probability corresponding to the diagnostic feature is higher than the pre-calculated critical probability value, it is determined that the corpse to be tested is suffering from sudden cardiac death.

[0094] Figure 2 It is a schematic diagram of a Norman diagram provided by an embodiment of the present invention.

[0095] like Figure 2 As shown, the Norman diagram is a tool derived from the logistic regression model that uses images to calculate probabilities. It is a two-dimensional image. That is to say, the Norman diagram in this embodiment can intuitively represent the probability of the body to be tested being diagnosed with sudden cardiac death.

[0096] In an exemplary embodiment, the critical probability value is calculated as follows:

[0097] Calculating the Youden Index of the training set;

[0098] The critical probability value is determined based on the maximum value of the Youden Index.

[0099] In implementation, the Youden index, also known as the correct index, is a method to evaluate the authenticity of a screening test. In practical applications, the Youden index is the sum of sensitivity and specificity minus one, which indicates the total ability of the screening method to detect true patients and non-patients. The larger the Youden index, the better the effect of the screening experiment and the greater its authenticity.

[0100] Sensitivity, also known as sensitivity, refers to the proportion of people who are actually sick that can be correctly identified as patients in the screening method; specificity refers to the proportion of people who are actually healthy that can be correctly identified as non-patients in the screening method.

[0101] In this embodiment, the critical probability value can be determined according to the calculated maximum value of the Youden index, and the determined critical probability value can be 0.240, that is, 24%, that is, the predicted probability higher than 24% through the Norman diagram is a high probability, and the predicted probability lower than 24% is a low probability.

[0102] The following is a description of the system for constructing a diagnostic model for sudden cardiac death provided by the present invention. The system for constructing a diagnostic model for sudden cardiac death described below and the method for constructing a diagnostic model for sudden cardiac death described above can be referenced to each other.

[0103] Figure 3 It is a structural schematic diagram of a system for constructing a diagnostic model for sudden cardiac death provided by an embodiment of the present invention.

[0104] like Figure 3 As shown, the system for constructing a diagnostic model for sudden cardiac death provided in this embodiment includes:

[0105] A data collection module 301 is used to collect data samples and construct a training set based on the data samples, wherein the training set includes diagnostic features, and the diagnostic features indicate that the data samples in the training set are sudden cardiac death;

[0106] A data processing module 302 is used to perform data interpolation and variable screening on the diagnostic features in the training set;

[0107] A model building module 303 is used to build a logistic regression model based on the training set, wherein the independent variable of the logistic regression model is the diagnostic feature, and the dependent variable of the logistic regression model is the diagnostic result of the sample corresponding to the diagnostic feature;

[0108] The model diagnosis module 304 is used to construct a diagnosis calculator based on the logistic regression model, and the diagnosis calculator inputs the received feature values ​​into the logistic regression model to diagnose sudden cardiac death.

[0109] In an exemplary embodiment, the data processing module is further used to: perform data interpolation on the diagnostic features in the training set using a Lost Forest interpolation method;

[0110] The LASSO regression method with ten-fold cross validation and the recursive feature elimination method based on random forest were used to screen the variables of the diagnostic features in the training set.

[0111] In an exemplary embodiment, the data processing module is further used to: use a LASSO regression method with ten-fold cross validation to perform variable screening on the diagnostic features in the training set to obtain a first variable set;

[0112] Using a recursive feature elimination method based on random forest to perform variable screening on the diagnostic features in the training set to obtain a second variable set;

[0113] The first variable set and the second variable set are intersected to obtain the result of variable screening.

[0114] In an exemplary embodiment, the model diagnosis module is further used to: obtain diagnostic features of the body to be tested through the input module;

[0115] Inputting the diagnostic characteristics of the corpse to be tested into the logistic regression model;

[0116] Calculating the diagnostic probability corresponding to the diagnostic feature through the Norman diagram derived from the logistic regression model, wherein the diagnostic probability reflects the probability that the corpse to be tested is diagnosed as sudden cardiac death;

[0117] If the diagnosis probability corresponding to the diagnostic feature is higher than the pre-calculated critical probability value, it is determined that the corpse to be tested is suffering from sudden cardiac death.

[0118] In an exemplary embodiment, the model diagnosis module is further used to: calculate the Youden index of the training set;

[0119] The critical probability value is determined based on the maximum value of the Youden Index.

[0120] The specific implementation method of the system for constructing a diagnostic model for sudden cardiac death provided in this embodiment can be implemented with reference to the above embodiments, and will not be described in detail here.

[0121] The solution of the present application also provides a method for evaluating the diagnostic performance of the diagnostic model, as shown below:

[0122] The constructed training set can be 2284 sample data collected by the Forensic Identification Center of Sun Yat-sen University from 2017 to 2023, and evaluated by the receiver operator characteristic curve (ROC) and area under the curve (AUC) and related statistical parameters (accuracy, sensitivity and specificity) used in the external test set. The external test set is composed of forensic identification centers from 5 universities in 4 provinces, including Southern Medical University (741 cases), China Medical University (567 cases), Jinan University (216 cases), Xi'an Jiaotong University (181 cases), and Guizhou Medical University (178 cases).

[0123] Figure 4 A schematic diagram of performance test results provided by an embodiment of the present invention.

[0124] like Figure 4 As shown, the AUC of the Norman diagram diagnostic model in the Forensic Identification Center of Sun Yat-sen University (training set) is 0.894 (95% confidence interval [Confidence interval, CI]: 0.879-0.908), indicating that it has a high diagnostic performance in the training set. At the same time, the AUC of the Norman diagram in five external test sets including Southern Medical University, China Medical University, Jinan University, Xi'an Jiaotong University, and Guizhou Medical University are 0.871 (95% CI: 0.839-0.903), 0.930 (95% CI: 0.906-0.953), 0.904 (95% CI: 0.865-0.943), 0.878 (95% CI: 0.827-0.929) and 0.834 (95% CI: 0.757-0.912), respectively, indicating that the model has the characteristics of high robustness.

[0125] As shown in Table 1, when the Norman diagram prediction probability is cut off at 0.240, the accuracy, sensitivity and specificity of the model for sudden cardiac death are 0.820, 0.876 and 0.797 in the Sun Yat-sen University training set, which also indicates that the model has good diagnostic performance. At the same time, the accuracy of the Norman diagram in five external test sets, including Southern Medical University, China Medical University, Jinan University, Xi'an Jiaotong University, and Guizhou Medical University, are 0.703, 0.824, 0.787, 0.729 and 0.798, respectively, the sensitivity is 0.873, 0.947, 0.926, 0.936 and 0.738, and the specificity is 0.655, 0.786, 0.741, 0.657 and 0.816, which verifies the generalizability of the model.

[0126] Table 1 Statistical parameters of the Norman graph in the training set and the other five external test sets

[0127]

[0128] At the same time, 4 senior forensic pathologists and 4 junior forensic pathologists can be invited to participate in the human-machine performance comparison and human-machine fusion experiment. As shown in Table 2, the ability of the Norman chart to diagnose sudden cardiac death is similar to that of senior forensic pathologists who know the systemic lesions when only considering cardiac lesions, and is significantly better than that of junior pathologists. In the human-machine fusion experiment, as shown in Table 3, with the assistance of the Norman chart, the AUC of 1 senior pathologist and all junior pathologists has been significantly improved, the sensitivity of 2 senior pathologists and 3 junior pathologists has also been significantly improved, and the specificity of 2 junior pathologists has been significantly improved. Overall, the AUC and sensitivity of pathologists have been significantly improved before and after human-machine fusion (P = 0.004 and P = 0.01).

[0129] Table 2 Comparison of diagnostic performance between Norman chart and senior pathologists

[0130]

[0131]

[0132] Table 3 Comparison of performance before and after human-machine fusion

[0133]

[0134] Among them, P in Table 2 # The p values ​​are obtained by Delong's test, and the p values ​​are given based on the McNemar test.

[0135] The data in the table marked with * indicate that the calculated P value is less than 0.05, which means that there is a statistically significant difference.

[0136] P in Table 3 # The value is obtained by Delong's test on the ROC curve before and after human-machine integration, P § The values ​​are given by McNemar test for sensitivity and specificity before and after human-machine fusion.

[0137] AUC in Tables 2 and 3 is the area under receiver-operating characteristic curve; Nom is the nomogram; SP is the senior pathologist; JP is the junior pathologist.

[0138] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute a method for constructing a diagnostic model for sudden cardiac death, the method comprising:

[0139] Collecting data samples, and constructing a training set based on the data samples, wherein the training set includes diagnostic features, and the diagnostic features characterize that the data samples in the training set are sudden cardiac death;

[0140] Performing data interpolation and variable screening on the diagnostic features in the training set;

[0141] Building a logistic regression model based on the training set, wherein the independent variable of the logistic regression model is the diagnostic feature, and the dependent variable of the logistic regression model is the diagnostic result of the sample corresponding to the diagnostic feature;

[0142] A diagnostic calculator is constructed based on the logistic regression model, and the diagnostic calculator inputs the received feature values ​​into the logistic regression model to diagnose sudden cardiac death.

[0143] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0144] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the method for constructing a diagnostic model for sudden cardiac death provided by the above methods, the method comprising:

[0145] Collecting data samples, and constructing a training set based on the data samples, wherein the training set includes diagnostic features, and the diagnostic features characterize that the data samples in the training set are sudden cardiac death;

[0146] Performing data interpolation and variable screening on the diagnostic features in the training set;

[0147] Building a logistic regression model based on the training set, wherein the independent variable of the logistic regression model is the diagnostic feature, and the dependent variable of the logistic regression model is the diagnostic result of the sample corresponding to the diagnostic feature;

[0148] A diagnostic calculator is constructed based on the logistic regression model, and the diagnostic calculator inputs the received feature values ​​into the logistic regression model to diagnose sudden cardiac death.

[0149] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the method for constructing a diagnostic model for sudden cardiac death provided by the above methods, the method comprising:

[0150] Collecting data samples, and constructing a training set based on the data samples, wherein the training set includes diagnostic features, and the diagnostic features characterize that the data samples in the training set are sudden cardiac death;

[0151] Performing data interpolation and variable screening on the diagnostic features in the training set;

[0152] Building a logistic regression model based on the training set, wherein the independent variable of the logistic regression model is the diagnostic feature, and the dependent variable of the logistic regression model is the diagnostic result of the sample corresponding to the diagnostic feature;

[0153] A diagnostic calculator is constructed based on the logistic regression model, and the diagnostic calculator inputs the received feature values ​​into the logistic regression model to diagnose sudden cardiac death.

[0154] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0155] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0156] 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 make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a diagnostic model for sudden cardiac death, characterized in that: include: Collecting data samples, and constructing a training set based on the data samples, wherein the training set includes diagnostic features, and the diagnostic features characterize that the data samples in the training set are sudden cardiac death; Performing data interpolation and variable screening on the diagnostic features in the training set; Building a logistic regression model based on the training set, wherein the independent variable of the logistic regression model is the diagnostic feature, and the dependent variable of the logistic regression model is the diagnostic result of the sample corresponding to the diagnostic feature; A diagnostic calculator is constructed based on the logistic regression model, and the diagnostic calculator inputs the received feature values ​​into the logistic regression model to diagnose sudden cardiac death.

2. The method for constructing a diagnostic model for sudden cardiac death according to claim 1, characterized in that: The diagnostic features include anatomical data and pathological analysis results of the data sample; The anatomical data includes the heart weight of the data sample, and / or the left ventricular wall thickness, and / or the right ventricular wall thickness, and / or the mitral valve circumference, and / or the aortic valve circumference, and / or the tricuspid valve circumference, and / or the pulmonary valve circumference, and / or the stenosis degree of the left anterior descending artery, and / or the stenosis degree of the left circumflex artery, and / or the stenosis degree of the right coronary artery; The pathological analysis results include myocardial infarction, and / or cardiomyopathy, and / or conduction system disease, and / or valvulitis, and / or epicarditis, and / or myocarditis, and / or endocarditis and pericarditis.

3. The method for constructing a diagnostic model for sudden cardiac death according to claim 2, characterized in that: The anatomical data includes the stenosis degree of the left anterior descending artery, the stenosis degree of the left circumflex artery, the stenosis degree of the right coronary artery, age, left ventricular wall thickness and pulmonary valve circumference of the data sample. The pathological analysis results include myocardial infarction, cardiomyopathy, valvulitis, pericarditis, myocarditis, and conduction system lesions.

4. The method for constructing a diagnostic model for sudden cardiac death according to claim 3, characterized in that: The diagnostic features also include the height of the data sample, and / or the thickness of the subcutaneous fat of the abdominal wall, and / or the gender, and / or the age.

5. The method for constructing a diagnostic model for sudden cardiac death according to claim 1, characterized in that: The data interpolation and variable screening of the diagnostic features in the training set include: Using the Lost Forest interpolation method to interpolate data on the diagnostic features in the training set; The LASSO regression method with ten-fold cross validation and the recursive feature elimination method based on random forest were used to screen the variables of the diagnostic features in the training set.

6. The method for constructing a diagnostic model for sudden cardiac death according to claim 5, characterized in that: The LASSO regression method using ten-fold cross validation and the recursive feature elimination method based on random forest are used to screen the diagnostic features in the training set, including: Using the LASSO regression method with ten-fold cross validation to screen the diagnostic features in the training set to obtain a first variable set; Using a recursive feature elimination method based on random forest to perform variable screening on the diagnostic features in the training set to obtain a second variable set; The first variable set and the second variable set are intersected to obtain the result of variable screening.

7. The method for constructing a diagnostic model for sudden cardiac death according to claim 1, characterized in that: The diagnostic calculator inputs the received characteristic values ​​into the logistic regression model to diagnose sudden cardiac death, including: Obtain the diagnostic characteristics of the body to be tested through the input box; Inputting the diagnostic characteristics of the corpse to be tested into the logistic regression model; Calculating the diagnostic probability corresponding to the diagnostic feature through the Norman diagram derived from the logistic regression model, wherein the diagnostic probability reflects the probability that the corpse to be tested is diagnosed as sudden cardiac death; If the diagnosis probability corresponding to the diagnostic feature is higher than the pre-calculated critical probability value, it is determined that the corpse to be tested is suffering from sudden cardiac death.

8. The method for constructing a diagnostic model for sudden cardiac death according to claim 7, characterized in that: The critical probability value is calculated as follows: Calculating the Youden Index of the training set; The critical probability value is determined based on the maximum value of the Youden Index.

9. A system for constructing a diagnostic model for sudden cardiac death, characterized in that: include: A data acquisition module, used for collecting data samples and constructing a training set based on the data samples, wherein the training set includes diagnostic features, and the diagnostic features indicate that the data samples in the training set are sudden cardiac death; A data processing module, used for performing data interpolation and variable screening on the diagnostic features in the training set; A model building module, used to build a logistic regression model based on the training set, wherein the independent variable of the logistic regression model is the diagnostic feature, and the dependent variable of the logistic regression model is the diagnostic result of the sample corresponding to the diagnostic feature; The model diagnosis module is used to construct a diagnosis calculator based on the logistic regression model, and the diagnosis calculator inputs the received feature values ​​into the logistic regression model to diagnose sudden cardiac death.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for constructing a diagnostic model for sudden cardiac death as described in any one of claims 1-8 is implemented.

Images