A method and system for identifying abdominal pain caused by non-abdominal organs

Abstract

The present invention provides a method and system for identifying abdominal pain caused by non-abdominal organs, relating to the technical field of abdominal pain data analysis and processing, including: establishing a multi-system symptom database, wherein the data stored in the multi-system symptom database includes patient abdominal pain symptom data and multiple system symptom data, wherein the system symptom data includes data of the cardiovascular system, respiratory system, and nervous system; training a judgment model to perform correlation analysis on the data in the multi-system symptom database, wherein the judgment model is used to identify the correlation between the patient's abdominal pain symptoms and various system symptoms; based on the patient data collected in real time, the judgment model outputs the correlation between the patient's abdominal pain and the cardiovascular system, respiratory system, or nervous system. The present invention has the advantages of being able to establish a data relationship between abdominal pain symptoms and other system symptoms and achieve more accurate data analysis of abdominal pain inducements.

Description

Technical Field

[0001] The present invention relates to the technical field of abdominal pain data analysis and processing, and in particular to a method and system for identifying abdominal pain caused by non-abdominal organs. Background Art

[0002] Abdominal pain is a common clinical symptom with complex and diverse causes. It may be caused not only by diseases of intra-abdominal organs but also by diseases of non-abdominal organs.

[0003] To assist medical personnel, data monitoring is often used to analyze whether non-abdominal organs are the cause of abdominal pain. However, existing technologies present difficulties in identifying non-abdominal organ causes of abdominal pain. For one thing, abdominal pain caused by non-abdominal organ diseases is often accompanied by symptoms from other systems, such as the cardiovascular, respiratory, and nervous systems. Furthermore, certain endocrine and autoimmune diseases can also cause abdominal pain. Consequently, existing technologies struggle to correlate abdominal pain symptoms with symptoms from other systems when using intelligent numerical analysis.

[0004] Therefore, it is necessary to optimize the method of abdominal pain data analysis, establish the data relationship between abdominal pain symptoms and other systemic symptoms, and achieve more accurate abdominal pain cause data analysis. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for identifying abdominal pain caused by non-abdominal organs, which can establish a data relationship between abdominal pain symptoms and other system symptoms, and achieve more accurate abdominal pain cause data analysis.

[0006] The present invention is achieved through the following technical solutions:

[0007] A method for identifying abdominal pain caused by non-intra-abdominal organs, comprising the following steps:

[0008] Establishing a multi-system symptom database, wherein the data stored in the multi-system symptom database includes abdominal pain symptom data of the patient and multiple system symptom data, wherein the system symptom data includes data of the cardiovascular system, respiratory system, and nervous system;

[0009] The training judgment model performs correlation analysis on the data in the multi-system symptom database. The judgment model is used to identify the correlation between the patient's abdominal pain symptoms and various system symptoms;

[0010] Based on the real-time collected patient data, the judgment model outputs the correlation between the patient's abdominal pain and the cardiovascular system, respiratory system or nervous system.

[0011] Preferably, the method for establishing a multi-system symptom database is:

[0012] Collect data on the patient's abdominal pain symptoms, including location, intensity, and duration of abdominal pain;

[0013] Collecting the patient's system symptom data, including electrocardiogram, blood pressure, body temperature, cough data, respiratory data, headache data, and dizziness data;

[0014] Perform data storage and standardization processing.

[0015] Preferably, the normalization method is Z-Score normalization or Min-Max normalization.

[0016] Preferably, the method for performing association analysis on the data in the multi-system symptom database by the training judgment model is:

[0017] Performing feature extraction based on the abdominal pain symptom data and the systemic symptom data through an input layer to obtain a feature vector;

[0018] Setting a hidden layer to extract deep information of the feature vector;

[0019] The correlation is outputted through an output layer.

[0020] Preferably, the method for extracting features based on the abdominal pain symptom data and the systemic symptom data through the input layer is:

[0021] The abdominal pain symptom data include abdominal pain location, intensity and duration;

[0022] Extract abdominal pain feature data :

[0023] ;

[0024] in, are weights and are optimized through training, and They are the highest intensity of abdominal pain per unit time and the longest duration of a single episode, is the number of the location where the abdominal pain occurs, and e is a natural constant;

[0025] The system symptom data includes electrocardiogram, blood pressure, body temperature, cough data, respiratory data, headache data and dizziness data;

[0026] Extract cardiovascular system characteristic data :

[0027] ;

[0028] ;

[0029] ;

[0030] ;

[0031] in, are weights and are optimized through training, 、 and They are electrocardiogram extraction data, blood pressure extraction data and body temperature extraction data, is the heart rate, is heart rate variability, are systolic and diastolic blood pressure, T is body temperature, is the body temperature threshold, e is a natural constant;

[0032] Extract respiratory system feature data :

[0033] ;

[0034] in, are weights and are optimized through training, and cough frequency and respiratory rate, respectively;

[0035] Extracting neural system feature data :

[0036] ;

[0037] in, are weights and are optimized through training, and They are the number of headaches per unit time and the maximum duration of a single headache, and They are the number of dizziness occurrences per unit time and the maximum duration of a single dizziness;

[0038] The feature vector for:

[0039] .

[0040] Preferably, the hidden layer adopts a fully connected layer, Process and output the processed .

[0041] Preferably, the method of outputting the correlation through the output layer is:

[0042] ;

[0043] ;

[0044] ;

[0045] in, 、 and are the correlation parameters between cardiovascular system, respiratory system and nervous system and abdominal pain, 、 、 、 、 and is the training weight, 、 、 、 、 and is the training bias;

[0046] Get judgment parameters :

[0047] ;

[0048] in, is the preset threshold, and max means finding the maximum value;

[0049] like It is determined that there is a correlation between the patient's abdominal pain symptoms and the systemic symptoms; otherwise, there is no correlation and the output correlation is 0;

[0050] If there is a relationship, get 、 and The maximum value among them is output, and the correlation of the system symptoms corresponding to the maximum value is output, and the correlation is the maximum value.

[0051] Preferably, when training the judgment model, the learning rate is dynamically adjusted by:

[0052] Initialize the iteration number , initialize the model parameters when the number of iterations s is 0 , the first-order moment estimate when the number of initial iterations s is 0 , the second-order moment estimate when the number of initial iterations s is 0 ;

[0053] in, 、 and represent the initial values ​​of model parameters, first-order moment estimates, and second-order moment estimates, respectively;

[0054] Iterate as follows:

[0055] Compute the gradient for iteration s :

[0056] ;

[0057] in, To find the gradient function, is the loss function;

[0058] Update the first-order and second-order moment estimates:

[0059] ;

[0060] ;

[0061] in, and They are the first-order moment estimated exponential decay rate and the second-order moment estimated exponential decay rate respectively;

[0062] The first-order moment estimate and the second-order moment estimate are bias-corrected to obtain the corrected first-order moment estimate and second-order moment estimates :

[0063] ;

[0064] ;

[0065] Update the parameters:

[0066] ;

[0067] ;

[0068] in, is the learning rate for the number of iterations s, To avoid constants with denominators equal to 0, is the initial learning rate, is the decay rate, and T is the decay period.

[0069] Preferably, the first-order moment estimates the exponential decay rate is 0.9, the second-order moment estimate has an exponential decay rate is 0.999, the attenuation rate is 0.95.

[0070] The present invention further provides a method and system for identifying abdominal pain caused by non-intra-abdominal organs, which is applied to the above-mentioned method for identifying abdominal pain caused by non-intra-abdominal organs, and comprises:

[0071] Database module: establishing a multi-system symptom database, wherein the data stored in the multi-system symptom database includes patient abdominal pain symptom data and multiple system symptom data, wherein the system symptom data includes data of the cardiovascular system, respiratory system and nervous system;

[0072] Model training module: The training judgment model performs correlation analysis on the data in the multi-system symptom database. The judgment model is used to identify the correlation between the patient's abdominal pain symptoms and various system symptoms;

[0073] Identification module: Based on the real-time collected patient data, the judgment model outputs the correlation between the patient's abdominal pain and the cardiovascular system, respiratory system or nervous system.

[0074] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0075] By analyzing the correlation between a patient's abdominal pain symptoms and cardiovascular, respiratory, and nervous system symptoms, the present invention can identify potential links between non-abdominal organ diseases and abdominal pain. Automatic analysis is achieved through training judgment models, improving the accuracy of data analysis, providing a reliable basis for medical staff and reducing their workload.

[0076] The present invention helps to identify non-intra-abdominal causes of abdominal pain at an early stage, improves the timeliness of data identification and auxiliary efficiency, and can achieve more accurate personalized diagnosis based on analysis of a multi-system symptom database to adapt to the symptom manifestations of different patients.

[0077] When establishing a judgment model, the present invention uses easily accessible data from various aspects to extract features and accurately analyze the relationship between abdominal pain and the cardiovascular, respiratory, and nervous systems, thereby improving the accuracy of etiology identification. At the same time, it has strong scalability, facilitating the expansion of analysis of diseases in other systems.

[0078] The present invention adopts a dynamic learning rate when training the judgment model. As the training progresses, when the model training tends to be stable, the learning rate is gradually reduced to prevent the model from skipping the optimal solution when approaching the optimal solution, thereby improving the training accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 This is a flow chart of a method for identifying abdominal pain caused by non-abdominal organs provided in Example 1 of the present invention;

[0080] Figure 2 This is a schematic diagram of the principle of a system for identifying abdominal pain caused by non-abdominal organs provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0081] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0082] Example 1

[0083] This embodiment provides a method for identifying abdominal pain caused by non-abdominal organs. Figure 1 , including the following steps:

[0084] Establishing a multi-system symptom database, wherein the data stored in the multi-system symptom database includes abdominal pain symptom data of the patient and multiple system symptom data, wherein the system symptom data includes data of the cardiovascular system, respiratory system, and nervous system;

[0085] The training judgment model performs correlation analysis on the data in the multi-system symptom database. The judgment model is used to identify the correlation between the patient's abdominal pain symptoms and various system symptoms;

[0086] Based on the real-time collected patient data, the judgment model outputs the correlation between the patient's abdominal pain and the cardiovascular system, respiratory system or nervous system.

[0087] The method of this embodiment aims to improve the ability to identify abdominal pain caused by non-intra-abdominal organ diseases, reduce misdiagnosis, and optimize the diagnosis and treatment process through intelligent analysis. This method combines a multi-system symptom database and an intelligent judgment model to perform a multi-dimensional analysis of the patient's abdominal pain symptoms, and accurately judge whether the abdominal pain is caused by cardiovascular system, respiratory system or nervous system diseases. This embodiment first constructs a multi-system symptom database, which stores a large amount of patient data, including abdominal pain symptom data and multiple system symptom data. Then, based on the database, the judgment model is trained to conduct an in-depth analysis of the potential correlation between abdominal pain symptoms and system symptoms. Finally, the existing data to be analyzed can be processed by the trained judgment model. The solution based on this embodiment can provide an auxiliary solution for intelligent solutions for the accurate identification of the cause of abdominal pain, reduce the risk of misdiagnosis, optimize medical decision-making and its efficiency, and improve the treatment effect of patients. If it is determined that the patient's abdominal pain is caused by the cardiovascular system, respiratory system or nervous system, it means that the abdominal pain is caused by non-intra-abdominal organs.

[0088] In this embodiment, the method for establishing a multi-system symptom database is:

[0089] Collect data on the patient's abdominal pain symptoms, including location, intensity, and duration of abdominal pain;

[0090] Collecting the patient's system symptom data, including electrocardiogram, blood pressure, body temperature, cough data, respiratory data, headache data, and dizziness data;

[0091] Perform data storage and standardization processing.

[0092] Specifically, the normalization method is Z-Score normalization or Min-Max normalization.

[0093] The Z-Score standardization is based on the mean and standard deviation std of the given data group. The value newData after data standardization is:

[0094] ;

[0095] Min-Max normalization is a linear transformation of the given data. It is based on the maximum value max_d and the minimum value min_d of the given data to perform data normalization operations. The value newData after data normalization is:

[0096] .

[0097] It should be noted that when establishing a multi-system symptom database, patient data can be collected through questionnaires, medical records, etc.

[0098] As a preferred solution of this embodiment, the method for performing association analysis on the data in the multi-system symptom database by the training judgment model is:

[0099] Performing feature extraction based on the abdominal pain symptom data and the systemic symptom data through an input layer to obtain a feature vector;

[0100] Setting a hidden layer to extract deep information of the feature vector;

[0101] The correlation is outputted through an output layer.

[0102] Furthermore, the method for extracting features based on the abdominal pain symptom data and the system symptom data through the input layer is:

[0103] The abdominal pain symptom data include abdominal pain location, intensity and duration;

[0104] Extract abdominal pain feature data :

[0105] ;

[0106] in, are weights and are optimized through training, and They are the highest intensity of abdominal pain per unit time and the longest duration of a single episode, is the number of the location where the abdominal pain occurs, and e is a natural constant;

[0107] When extracting abdominal pain feature data, computing efficiency and cost-effectiveness are improved by using simple and representative data.

[0108] The system symptom data includes electrocardiogram, blood pressure, body temperature, cough data, respiratory data, headache data and dizziness data;

[0109] Extract cardiovascular system characteristic data :

[0110] ;

[0111] ;

[0112] ;

[0113] ;

[0114] in, are weights and are optimized through training, 、 and They are electrocardiogram extraction data, blood pressure extraction data and body temperature extraction data, is the heart rate, is heart rate variability, are systolic and diastolic blood pressure, T is body temperature, is the body temperature threshold, e is a natural constant;

[0115] Extract respiratory system feature data :

[0116] ;

[0117] in, are weights and are optimized through training, and cough frequency and respiratory rate, respectively;

[0118] Extracting neural system feature data :

[0119] ;

[0120] in, are weights and are optimized through training, and They are the number of headaches per unit time and the maximum duration of a single headache, and They are the number of dizziness occurrences per unit time and the maximum duration of a single dizziness;

[0121] The feature vector for:

[0122] .

[0123] As an optional solution, the hidden layer adopts a fully connected layer, respectively Process and output the processed .

[0124] Then, the method of outputting the correlation through the output layer is:

[0125] ;

[0126] ;

[0127] ;

[0128] in, 、 and are the correlation parameters between cardiovascular system, respiratory system and nervous system and abdominal pain, 、 、 、 、 and is the training weight, 、 、 、 、 and is the training bias;

[0129] Get judgment parameters :

[0130] ;

[0131] in, is the preset threshold, and max means finding the maximum value;

[0132] like It is determined that there is a correlation between the patient's abdominal pain symptoms and the systemic symptoms; otherwise, there is no correlation and the output correlation is 0;

[0133] If there is a relationship, get 、 and The maximum value among them is used to output the correlation of the system symptoms corresponding to the maximum value, where the correlation is the maximum value. After a large amount of training and actual recognition, the threshold value of the correlation can be set as extremely high, relatively high or normal according to the empirical value.

[0134] In the above scheme, this embodiment combines multiple physiological parameters such as electrocardiogram, blood pressure, body temperature, cough, respiratory rate, headache, dizziness, etc. to achieve multi-angle health status assessment, which is more comprehensive than the traditional single symptom analysis method.

[0135] Cardiovascular data is quite diverse and complex. This example uses a simplified method to extract more features in advance to ensure the sensitivity of the model. Specifically, the cardiovascular system feature data is extracted through electrocardiogram, blood pressure and body temperature. When extracting ECG features, heart rate and heart rate variability are combined. Heart rate variability is processed using an exponential transformation, and ECG features are expressed as the product of the indexed heart rate variability and the heart rate. Heart rate is an indicator of short-term dynamic changes, while heart rate variability reflects the long-term stability of the heart rhythm. After indexation, the impact of heart rate variability on the overall heart rate is nonlinearly amplified, better integrating short-term heart rate fluctuations with long-term trends, improving the ability to identify abnormal ECG signals. When extracting blood pressure-related features, systolic, diastolic, and pulse pressure are considered. Specifically, the sum of systolic and diastolic pressures is extracted, and the product of the difference between the indexed pulse pressure and a preset pulse pressure threshold is extracted. The sum of systolic and diastolic pressures can more comprehensively characterize an individual's overall blood pressure level, increasing sensitivity to blood pressure abnormalities. The difference between systolic and diastolic pressures, or pulse pressure, reflects arterial compliance (elasticity) and vascular resistance. After the difference between pulse pressure and pulse pressure threshold is indexed, the abnormality can also be amplified to improve data sensitivity; the body temperature feature is directly obtained by comparing the body temperature with the conventional body temperature threshold, which can be set to 37 degrees Celsius. Finally, the three parameters extracted from the electrocardiogram, blood pressure and body temperature are weighted and summed to obtain the cardiovascular system characteristic data. .

[0136] On the other hand, respiratory system feature data is extracted through cough frequency and respiratory rate When the weighted sum is directly performed; the nervous system feature data is extracted through the headache and dizziness related data When analyzing the headache frequency and the maximum duration of a single headache, as well as the dizziness frequency and the maximum duration of a single headache, the sensitivity to the duration of discomfort symptoms was enhanced, and the abdominal pain feature data was extracted. The same principle applies when considering the highest intensity and the longest duration of abdominal pain in a unit time. It should be noted that the above unit time can be set to one day, two days or three days, etc.

[0137] Based on the extracted feature vectors, each feature data and abdominal pain feature data can be optimized through training. The similarity between them is calculated based on the maximum similarity and the preset threshold To determine whether abdominal pain is associated with any systemic symptoms. The working principle of the output layer is clear, which improves the stability and interpretability of the output layer.

[0138] Through extensive data training, we determine the appropriate neural network and its parameters (such as weights and biases) to achieve the closest simulation of complex tasks. To achieve this goal, we first need a metric to measure how closely the current model simulates the complex task. This metric is the loss function, which measures the difference between the model's predictions and the actual results, or the error. In layman's terms, the loss function is like a scoring system, indicating how well or poorly the model performs. By minimizing the loss function, we can continuously optimize the model, making its predictions more accurate. Furthermore, training data can account for 80% of the total data, with the remaining 20% ​​serving as test data.

[0139] During the training phase, when training the judgment model, the learning rate is dynamically adjusted as follows:

[0140] Initialize the iteration number , initialize the model parameters when the number of iterations s is 0 , the first-order moment estimate when the number of initial iterations s is 0 , the second-order moment estimate when the number of initial iterations s is 0 ;

[0141] in, 、 and represent the initial values ​​of model parameters, first-order moment estimates, and second-order moment estimates, respectively;

[0142] Iterate as follows:

[0143] Compute the gradient for iteration s :

[0144] ;

[0145] in, To find the gradient function, is the loss function;

[0146] Update the first-order and second-order moment estimates:

[0147] ;

[0148] ;

[0149] in, and They are the first-order moment estimated exponential decay rate and the second-order moment estimated exponential decay rate respectively;

[0150] The first-order moment estimate and the second-order moment estimate are bias-corrected to obtain the corrected first-order moment estimate and second-order moment estimates :

[0151] ;

[0152] ;

[0153] Update the parameters:

[0154] ;

[0155] ;

[0156] in, is the learning rate for the number of iterations s, To avoid constants with denominators equal to 0, is the initial learning rate, is the decay rate, and T is the decay period.

[0157] At this time, the first-order moment estimates the exponential decay rate is 0.9, the second-order moment estimate has an exponential decay rate is 0.999, the attenuation rate is 0.95.

[0158] The learning rate determines the step size for parameter updates during model training. A dynamic learning rate adjustment strategy is used. A higher learning rate is set at the beginning of training to enable the model to quickly approach the optimal solution. As training progresses and the model stabilizes, the learning rate is gradually reduced to prevent the model from skipping the optimal solution as it approaches it. For example, an exponentially decaying learning rate can be used, where the learning rate decreases exponentially with increasing training rounds. In practice, a reasonable learning rate range can be selected: [0.00001, 0.0001, 0.001, 0.01, 0.1]. Using cross-validation, the model's performance metrics (such as accuracy and loss) are evaluated on a validation set for different initial learning rates. The optimal initial learning rate is ultimately selected, resulting in an initial learning rate of 0.001.

[0159] During actual training, the model is tested and evaluated to ensure high recognition accuracy. This can be done using metrics such as accuracy or the F1 score based on precision and recall.

[0160] The method for obtaining the accuracy is:

[0161] Accuracy=(TP+TN) / (TP+TN+FP+FN);

[0162] Among them, TP is the true positive example, that is, the number of samples correctly predicted by the model as the positive class (for example, correctly predicting that abdominal pain is caused by intra-abdominal organs); TN is the true negative example, that is, the number of samples correctly predicted by the model as the negative class (for example, correctly predicting that abdominal pain is caused by non-intra-abdominal organs); FP is the false positive example, that is, the number of samples incorrectly predicted by the model as the negative class as the positive class (for example, incorrectly predicting that abdominal pain is caused by intra-abdominal organs, when in fact non-intra-abdominal organs cause abdominal pain); FN is the false negative example, that is, the number of samples incorrectly predicted by the model as the negative class (for example, incorrectly predicting that abdominal pain is caused by non-intra-abdominal organs, when in fact abdominal pain is caused by intra-abdominal organs).

[0163] Accuracy indicates the proportion of correctly predicted examples among all predictions. It's an intuitive metric, but can be misleading in cases of class imbalance. For example, if a dataset contains far more examples of abdominal pain caused by intra-abdominal organs than by non-intra-abdominal organs, the model may have a high accuracy even if it always predicts normal, but in reality, the model may not be ideal.

[0164] The method to obtain the precision is:

[0165] Precision=TP / (TP+FP);

[0166] Precision focuses on the proportion of samples predicted by the model as positive that are actually positive. In fault prediction, it indicates the proportion of devices that actually malfunction among those predicted to have abdominal pain caused by non-intra-abdominal organs. A higher precision indicates a more accurate prediction of whether abdominal pain is caused by an intra-abdominal organ. However, it does not account for cases where abdominal pain caused by an intra-abdominal organ is actually present but not predicted.

[0167] The method for obtaining the recall rate Recall is:

[0168] Recall = TP / (TP+FN);

[0169] Recall indicates the proportion of actual positive examples correctly predicted by the model. In predicting the organ causing abdominal pain, it reflects the model's ability to correctly identify the organ causing the pain. A higher recall indicates a more comprehensive coverage of the biological organs causing abdominal pain. However, the model does not consider cases where abdominal pain caused by an intra-abdominal organ is mistakenly identified as a non-intra-abdominal organ.

[0170] Finally, the F1 score is calculated based on the precision and recall obtained previously:

[0171] F1=(Precision* Recall) / (Precision+ Recall);

[0172] The F1 score is the harmonic mean of precision and recall, taking both into account. High precision and recall result in a high F1 score, striking a balance between the two and suitable for situations where both are important. For abdominal pain organ prediction, if you want to accurately predict whether an intra-abdominal organ is causing the pain while also identifying the underlying organ as closely as possible, the F1 score is a good evaluation metric.

[0173] Finally, it is worth noting that this solution is highly scalable and can easily expand the abdominal pain recognition of more systemic symptoms. That is, the systemic symptoms can be expanded according to needs. For example, the data of the endocrine system and the immune system can be taken into consideration, and the relevant data can be directly collected. The response features can be extracted by the same method to increase the relevance of the output. For example, the relevance of the endocrine system and the immune system are 、 , and then the same method can be used to make judgments, for example:

[0174] Get judgment parameters :

[0175] ;

[0176] in, is the preset threshold, and max means finding the maximum value;

[0177] like It is determined that there is a correlation between the patient's abdominal pain symptoms and the systemic symptoms; otherwise, there is no correlation and the output correlation is 0;

[0178] If there is a relationship, get 、 、 、 The maximum value among them is used to output the correlation of the system symptoms corresponding to the maximum value, where the correlation is the maximum value. After a large amount of training and actual recognition, the threshold value of the correlation can be set as extremely high, relatively high or normal according to the empirical value.

[0179] Example 2

[0180] The present invention also provides a method and system for identifying abdominal pain caused by non-abdominal organs. Figure 2 A method for identifying abdominal pain caused by non-abdominal organs, applied to the above embodiment, comprises:

[0181] Database module: establishing a multi-system symptom database, wherein the data stored in the multi-system symptom database includes patient abdominal pain symptom data and multiple system symptom data, wherein the system symptom data includes data of the cardiovascular system, respiratory system and nervous system;

[0182] Model training module: The training judgment model performs correlation analysis on the data in the multi-system symptom database. The judgment model is used to identify the correlation between the patient's abdominal pain symptoms and various system symptoms;

[0183] Identification module: Based on the real-time collected patient data, the judgment model outputs the correlation between the patient's abdominal pain and the cardiovascular system, respiratory system or nervous system.

[0184] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for identifying abdominal pain caused by non-abdominal organs, characterized in that: The following steps are involved: Establishing a multi-system symptom database, wherein the data stored in the multi-system symptom database includes abdominal pain symptom data of the patient and multiple system symptom data, wherein the system symptom data includes data of the cardiovascular system, respiratory system, and nervous system; The training judgment model performs correlation analysis on the data in the multi-system symptom database. The judgment model is used to identify the correlation between the patient's abdominal pain symptoms and various system symptoms; Based on the real-time collected patient data, the judgment model outputs the correlation between the patient's abdominal pain and the cardiovascular system, respiratory system or nervous system; The method for performing association analysis on the data in the multi-system symptom database by the training judgment model is as follows: Performing feature extraction based on the abdominal pain symptom data and the systemic symptom data through an input layer to obtain a feature vector; Setting a hidden layer to extract deep information of the feature vector; Outputting the correlation through an output layer; The method for extracting features based on the abdominal pain symptom data and the system symptom data through the input layer is: Extract abdominal pain feature data : ； in, are weights and are optimized through training, and They are the highest intensity of abdominal pain per unit time and the longest duration of a single episode, is the number of the location where the abdominal pain occurs, and e is a natural constant; Extract cardiovascular system characteristic data : ； ； ； ； in, are weights and are optimized through training, 、 and They are electrocardiogram extraction data, blood pressure extraction data and body temperature extraction data, is the heart rate, is heart rate variability, are systolic and diastolic blood pressure, T is body temperature, is the body temperature threshold, e is a natural constant; Extract respiratory system feature data : ； in, are weights and are optimized through training, and cough frequency and respiratory rate, respectively; Extracting neural system feature data : ； in, are weights and are optimized through training, and They are the number of headaches per unit time and the maximum duration of a single headache, and They are the number of dizziness occurrences per unit time and the maximum duration of a single dizziness; The feature vector for: ； The hidden layer adopts a fully connected layer, Process and output the processed ; The method of outputting the correlation through the output layer is: ； ； ； in, 、 and are the correlation parameters between cardiovascular system, respiratory system and nervous system and abdominal pain, 、 、 、 、 and is the training weight, 、 、 、 、 and is the training bias; Get judgment parameters : ； in, is the preset threshold, and max means finding the maximum value; like It is determined that there is a correlation between the patient's abdominal pain symptoms and the systemic symptoms; otherwise, there is no correlation and the output correlation is 0; If there is a relationship, get 、 and The maximum value among them is output, and the correlation of the system symptoms corresponding to the maximum value is output, and the correlation is the maximum value.

2. The method for identifying abdominal pain caused by non-abdominal organs according to claim 1, characterized in that: The method for establishing a multi-system symptom database is: Collect data on the patient's abdominal pain symptoms, including location, intensity, and duration of abdominal pain; Collecting the patient's system symptom data, including electrocardiogram, blood pressure, body temperature, cough data, respiratory data, headache data, and dizziness data; Perform data storage and standardization processing.

3. The method for identifying abdominal pain caused by non-abdominal organs according to claim 2, characterized in that: The method of the normalization process is Z-Score normalization or Min-Max normalization.

4. The method for identifying abdominal pain caused by non-abdominal organs according to claim 1, characterized in that: When training the judgment model, the learning rate is dynamically adjusted as follows: Initialize the iteration number , initialize the model parameters when the number of iterations s is 0 , the first-order moment estimate when the number of initial iterations s is 0 , the second-order moment estimate when the number of initial iterations s is 0 ; in, 、 and represent the initial values ​​of model parameters, first-order moment estimates, and second-order moment estimates, respectively; Iterate as follows: Compute the gradient for iteration s : ； in, To find the gradient function, is the loss function; Update the first-order and second-order moment estimates: ； ； in, and They are the first-order moment estimated exponential decay rate and the second-order moment estimated exponential decay rate respectively; The first-order moment estimate and the second-order moment estimate are bias-corrected to obtain the corrected first-order moment estimate and second-order moment estimates : ； ； Update the parameters: ； ； in, is the learning rate for the number of iterations s, To avoid constants with denominators equal to 0, is the initial learning rate, is the decay rate, and T is the decay period.

5. The method for identifying abdominal pain caused by non-abdominal organs according to claim 4, characterized in that: The first-order moment estimates the exponential decay rate is 0.9, the second-order moment estimate has an exponential decay rate is 0.999, the attenuation rate is 0.

95.

6. A method and system for identifying abdominal pain caused by non-abdominal organs, applied to the method for identifying abdominal pain caused by non-abdominal organs according to any one of claims 1 to 5, characterized in that: include: Database module: establishing a multi-system symptom database, wherein the data stored in the multi-system symptom database includes patient abdominal pain symptom data and multiple system symptom data, wherein the system symptom data includes data of the cardiovascular system, respiratory system and nervous system; Model training module: The training judgment model performs correlation analysis on the data in the multi-system symptom database. The judgment model is used to identify the correlation between the patient's abdominal pain symptoms and various system symptoms; Identification module: Based on the real-time collected patient data, the judgment model outputs the correlation between the patient's abdominal pain and the cardiovascular system, respiratory system or nervous system.

Images