Airway obstruction intervention recommendation method and system
By establishing an airway obstruction intervention decision model and utilizing real-time diagnostic data and severity assessment models, a rapid and effective airway obstruction intervention plan is provided, which solves the problem of delayed airway obstruction intervention caused by the shortage of professional medical personnel and improves the treatment effect in emergency situations.
Patent Information
- Application Number
- CN202111454140.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-01
- Publication Date
- 2026-02-03
- Estimated Expiration
- 2041-12-01
AI Technical Summary
In emergency situations, a shortage of professional medical personnel can lead to delays in airway obstruction intervention, affecting the patient's treatment outcome.
By establishing an airway obstruction intervention decision-making model, extracting key indicator feature data from real-time diagnostic data, and combining it with an airway obstruction severity assessment model, we can provide rapid and effective intervention recommendations.
It assists emergency medical personnel in quickly and accurately intervening in airway obstruction, improving the ability to treat patients in emergency situations.
Smart Images
Figure CN114068016B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing, and more specifically to a method and system for recommending interventions for airway obstruction. Background Technology
[0002] Emergency medicine is a comprehensive clinical discipline based on the development of modern medical science, utilizing clinical medical treatment measures, and studying and managing the timely, rapid, and effective treatment of acute illnesses from a holistic perspective. With the advancement of medical science, emergency medicine has become an independent, new, and comprehensive medical discipline, and is one of the fastest-growing clinical disciplines. Emergency medicine differs from conventional outpatient medicine, demanding higher standards in diagnosis and treatment. Doctors need to make accurate diagnoses and provide rapid treatment within a short timeframe, using limited information about the patient's condition. Therefore, providing emergency physicians with rapid, efficient, and accurate methods for analyzing patient conditions is of great significance.
[0003] Airway obstruction is a major cause of traumatic death, and timely and effective airway management is crucial to reducing mortality. The core issue in trauma airway management is the early identification of airway obstruction and accurate assessment of its cause. The core techniques of airway management are protecting the airway, opening the airway, and establishing artificial ventilation, with the fundamental goal of ensuring adequate ventilation and oxygenation. In emergency situations, the need for airway management must be decided as early as possible. However, for patients with spontaneous breathing, this decision is often significantly delayed, and severe airway obstruction poses a significant threat to the patient's life. Early and accurate airway identification and effective intervention can often lead to a rapid recovery. However, a current problem is the shortage of qualified medical personnel. Therefore, how to assist emergency medical personnel in rapidly and effectively intervening in airway obstruction is a significant challenge facing the field. Summary of the Invention
[0004] This invention provides a recommended method and system for airway obstruction intervention, in order to assist relevant medical personnel in quickly and effectively managing airway obstruction in patients and improve the ability to provide emergency care under sudden circumstances.
[0005] Therefore, the present invention provides the following technical solution:
[0006] A recommended method for airway obstruction intervention, the method comprising:
[0007] Obtain real-time diagnostic data for the current patient;
[0008] Extract key indicator feature data from the real-time diagnostic data;
[0009] Using the key indicator feature data and the pre-established airway obstruction intervention decision model, a recommended airway obstruction intervention plan is determined.
[0010] Optionally, the method further includes:
[0011] Determine the severity score of the patient's airway obstruction;
[0012] Using the key indicator feature data, the patient's airway obstruction severity score, and the pre-established airway obstruction intervention decision model, a recommended airway obstruction intervention plan is determined.
[0013] Optionally, the method further includes:
[0014] A model for assessing the severity of airway obstruction is pre-built based on patient medical data sources;
[0015] The severity score of airway obstruction in the patient was determined using the airway obstruction severity assessment model.
[0016] Optionally, the airway obstruction severity assessment model constructed based on patient medical data sources includes:
[0017] Collect patient medical data sources and extract patient visit-related data from the medical data sources;
[0018] Correlation analysis was performed on the patient visit data to obtain a set of key indicators;
[0019] The indicators in the key indicator set are normalized and missing values are filled to obtain the key indicator sample set.
[0020] Determine the airway intervention label for each sample in the key indicator sample set and the airway severity corresponding to the airway intervention label;
[0021] An airway obstruction severity assessment model was trained based on the key indicator sample set and the airway severity.
[0022] Optionally, the correlation analysis of the patient's medical visit data to obtain a set of key indicators includes:
[0023] Correlation analysis was performed on the patient's test indicators to obtain a set of candidate key indicators;
[0024] For each candidate indicator in the candidate key indicator set, sort the correlation coefficients between the candidate indicators and the severity of airway obstruction by absolute value, select candidate indicators whose absolute correlation coefficients are greater than or equal to a set threshold and add them to the recommended indicator set; otherwise, add them to the remaining indicator set.
[0025] Based on medical knowledge, the importance of each indicator in the recommended indicator set is classified, and then the indicators in the classified recommended indicator set are added to the key indicator set.
[0026] Select a subset of indicators from the remaining indicator set and add them to the key indicator set;
[0027] Add a mental state label and an agitation label to each training sample in the set of key indicators.
[0028] Optionally, the standardization process for the indicators in the key indicator set includes:
[0029] If the indicator is within the normal range, then the indicator value is set to 0;
[0030] If the indicator is higher than the normal range and lower than the extreme danger upper limit, then the indicator value will be mapped to a value in the range of 0 to 1;
[0031] If the indicator is higher than or equal to the extreme danger upper limit value, then the indicator value is set to 1;
[0032] If the indicator is below the normal range and above the extremely dangerous lower limit, then the indicator value will be mapped to a value in the range of 0 to -1.
[0033] If the indicator is lower than or equal to the extreme danger lower limit, then the indicator is set to -1.
[0034] Optionally, the method further includes establishing an airway obstruction intervention decision model in the following manner:
[0035] A random forest is constructed from multiple decision trees. The Gini coefficient is used as the criterion for measuring the contribution of features in each decision tree. The decision result of each decision tree is the probability of recommending an airway intervention plan.
[0036] The random forest is used as the airway obstruction intervention decision model, and the output of the airway obstruction intervention decision model is the decision result of the multiple decision trees.
[0037] An airway obstruction intervention recommendation system, the system comprising:
[0038] The data acquisition module is used to acquire the current patient's real-time diagnostic data;
[0039] The feature extraction module is used to extract key indicator feature data from the real-time diagnostic data;
[0040] The recommendation module is used to determine recommended intervention schemes for airway obstruction by utilizing the key indicator feature data and the pre-established airway obstruction intervention decision model.
[0041] Optionally, the system further includes:
[0042] The assessment module is used to determine the severity score of the patient's airway obstruction.
[0043] The recommendation module is used to determine the recommended airway obstruction intervention plan by utilizing the key indicator feature data, the patient's airway obstruction severity score, and the pre-established airway obstruction intervention decision model.
[0044] Optionally, the system further includes:
[0045] The first model building module is used to pre-build an airway obstruction severity assessment model based on patient medical data sources;
[0046] The assessment module is used to determine the severity score of the patient's airway obstruction using the airway obstruction severity assessment model.
[0047] Optionally, the first model building module includes:
[0048] The data acquisition unit is used to collect patient medical data sources and extract patient visit-related data from the medical data sources.
[0049] The data analysis unit is used to perform correlation analysis on the patient's medical visit-related data to obtain a set of key indicators;
[0050] The data processing unit is used to perform normalization processing and missing value filling on the indicators in the key indicator set to obtain a key indicator sample set.
[0051] The label determination unit is used to determine the airway intervention label and the airway severity corresponding to the airway intervention label for each sample in the key indicator sample set.
[0052] The training unit is used to train an airway obstruction severity assessment model based on the key indicator sample set and the airway severity.
[0053] Optionally, the data analysis unit includes:
[0054] The data parsing subunit is used to perform correlation analysis on the patient's test indicators to obtain a set of candidate key indicators.
[0055] The sorting subunit is used to sort each candidate indicator in the candidate key indicator set according to the absolute value of its correlation coefficient with the severity of airway obstruction, and select candidate indicators whose absolute value of correlation coefficient is greater than or equal to a set threshold to add to the recommended indicator set; otherwise, it is added to the remaining indicator set.
[0056] The filtering subunit is used to classify the importance of each indicator in the recommended indicator set based on medical knowledge, and then add the graded indicators in the recommended indicator set to the key indicator set; and select some indicators from the remaining indicator set to add to the key indicator set.
[0057] A label-adding subunit is used to add a mental state label and an agitation label to each training sample in the key indicator set.
[0058] Optionally, the system further includes:
[0059] The second model construction module is used to construct a random forest from multiple decision trees. The criterion for measuring the contribution of features in each decision tree is the Gini coefficient, and the decision result of each decision tree is the recommendation probability of an airway intervention plan. The random forest is used as an airway obstruction intervention decision model, and the output of the airway obstruction intervention decision model is the decision result of the multiple decision trees.
[0060] The airway obstruction intervention recommendation method and system provided in this invention pre-establishes an airway obstruction intervention decision model, obtains the patient's real-time diagnostic data and airway obstruction severity score, extracts key indicator feature data from the real-time diagnostic data, inputs the key indicator feature data and the patient's airway obstruction severity score into the airway obstruction intervention decision model, and obtains an airway obstruction intervention recommendation plan based on the model's output. This allows for the rapid provision of recommended intervention plans for the patient's airway obstruction, assisting relevant medical personnel in making rapid and effective treatments for the patient's airway obstruction and improving the emergency treatment capabilities under sudden conditions. Attached Figure Description
[0061] Figure 1 This is a flowchart of the recommended method for airway obstruction intervention according to an embodiment of the present invention;
[0062] Figure 2 This is a schematic diagram of the airway obstruction intervention decision model in an embodiment of the present invention;
[0063] Figure 3 This is a schematic diagram of the airway obstruction severity assessment model in an embodiment of the present invention;
[0064] Figure 4 This is a flowchart of constructing an airway obstruction severity assessment model in an embodiment of the present invention;
[0065] Figure 5 This is a flowchart of filling missing values for indicators in the key indicator set in an embodiment of the present invention;
[0066] Figure 6 This is a flowchart illustrating the process of determining the severity score of a patient's airway obstruction using an airway obstruction severity assessment model, as described in this embodiment of the invention.
[0067] Figure 7 This is a schematic diagram of a structure of the airway obstruction intervention recommendation system according to an embodiment of the present invention;
[0068] Figure 8 This is another structural schematic diagram of the airway obstruction intervention recommendation system according to an embodiment of the present invention;
[0069] Figure 9 This is a schematic diagram of the structure of the first model construction module in an embodiment of the present invention. Detailed Implementation
[0070] This invention provides a method and system for recommending airway obstruction intervention. By pre-establishing an airway obstruction intervention decision model, it acquires the patient's immediate diagnostic data and airway obstruction severity score, extracts key indicator feature data from the immediate diagnostic data, and inputs the key indicator feature data and the patient's airway obstruction severity score into the airway obstruction intervention decision model. Based on the model's output, it obtains an airway obstruction intervention recommendation plan, thereby quickly providing a recommended intervention plan for the patient's airway obstruction. This assists relevant medical personnel in rapidly and effectively managing the patient's airway obstruction, improving the emergency response capability under sudden conditions.
[0071] In one non-limiting embodiment, the airway obstruction intervention recommendation method of the present invention can acquire the current patient's real-time diagnostic data; extract key indicator feature data from the real-time diagnostic data; and determine the recommended airway obstruction intervention plan using the key indicator feature data and a pre-established airway obstruction intervention decision model.
[0072] In another non-limiting embodiment, the airway obstruction intervention recommendation method of the present invention can further determine the severity score of the patient's airway obstruction, and use the key indicator feature data, the severity score of the patient's airway obstruction, and the pre-established airway obstruction intervention decision model to determine the recommended airway obstruction intervention plan.
[0073] It should be noted that, in the embodiments of the present invention, the airway obstruction intervention decision model can be constructed using random forest, based on key indicator feature data, or based on key indicator feature data and the patient's airway obstruction severity score. The only difference is the input parameters of the decision tree.
[0074] The second option described above will be explained in detail below.
[0075] like Figure 1 The diagram shows a flowchart of a recommended method for airway obstruction intervention according to an embodiment of the present invention, which includes the following steps:
[0076] Step 101: Obtain the current patient's real-time diagnostic data.
[0077] The real-time diagnostic data includes: basic vital signs, test data, and consciousness status labels.
[0078] Step 102: Extract key indicator feature data from the real-time diagnostic data.
[0079] Similar to the extraction of key indicator sets when establishing the airway obstruction severity assessment model, key indicator feature data are extracted from the current patient's immediate diagnostic data. These key indicator features mainly include the indicators shown in Table 1.
[0080] Table 1:
[0081]
[0082] Step 103: Determine the severity score of the patient's airway obstruction.
[0083] In one non-limiting embodiment, the severity score of the patient's airway obstruction can be assigned by the medical personnel based on experience, using the patient's current real-time diagnostic data.
[0084] In another non-limiting embodiment, automatic scoring can be achieved based on an airway obstruction severity assessment model pre-built from patient medical data sources. The method for establishing the airway obstruction severity assessment model and the process of scoring using this model will be described in detail later.
[0085] Step 104: Input the key indicator feature data and the score into the pre-established airway obstruction intervention decision model, and obtain the recommended airway obstruction intervention plan based on the output of the airway obstruction intervention decision model.
[0086] The airway obstruction intervention decision model can adopt a random forest structure, such as... Figure 2 As shown. Specifically, it consists of multiple decision trees ( Figure 2 The diagram illustrates the construction of a random forest using T decision trees. The criterion for measuring the contribution of features in each decision tree is the Gini coefficient, and the decision result of each decision tree is the probability of recommending an airway intervention plan. The random forest is used as an airway obstruction intervention decision model, and the output of the airway obstruction intervention decision model is the decision result of the multiple decision trees.
[0087] Furthermore, to prevent overfitting of each decision tree, the depth of each decision tree can be limited to 5. The final output is the decision results of T decision trees. The following formula gives the recommended probability of each airway intervention:
[0088]
[0089] Where T represents the number of decision trees, Pi represents the probability of predicting the i-th airway intervention, and DT t(x) represents the output result of the t-th decision tree with x as input. When the result is i, increment by 1; otherwise, increment by 0.
[0090] The above formula can be used to convert the results of T trees in a random forest into the recommended probability of each airway intervention. The higher the recommended probability of a certain intervention, the more likely the patient is to need that airway intervention for treatment.
[0091] When using the above airway obstruction intervention decision model to recommend airway obstruction intervention programs, the model input includes key patient indicator data and the patient's airway obstruction severity score, with the input multidimensional vector x = <x1,x2,...,x n S>, where x1,…,x n denoted as n standardized key indicator feature data of the patient, and S represents the airway obstruction severity score; the output of the model is the recommendation probability of each intervention recommendation, such as: the probability of endotracheal intubation is 60%, and the probability of nasal cannula oxygen inhalation is 40%.
[0092] The present invention can better assist medical personnel in quickly and effectively managing airway obstruction in patients, thereby improving the ability to provide emergency care in sudden situations.
[0093] As mentioned earlier, in practical applications, the severity of airway obstruction in patients can be scored manually or automatically based on an airway obstruction severity assessment model. The second method will be explained in detail below.
[0094] The airway obstruction severity assessment model employs a machine learning model, which uses a multilayer perceptron structure, such as... Figure 3 As shown, the model includes an input layer, three hidden layers, and an output layer. The three hidden layers are fully connected to the output layer, and a random deactivation mechanism is used between adjacent hidden layers. The activation function is calculated using the Rectified Linear Unit (ReLU) activation function. The ReLU activation function typically refers to a nonlinear function represented by the ramp function and its variants.
[0095] like Figure 4 The diagram shown is a flowchart of constructing an airway obstruction severity assessment model in an embodiment of the present invention, including the following steps:
[0096] Step 401: Collect patient medical data sources and extract patient visit-related data from the medical data sources.
[0097] The patient medical data source may include, but is not limited to, any one or more of the following: emergency database, nursing system, hospital database.
[0098] The patient's medical data may include, but is not limited to, any one or more of the following: basic patient information (such as age, gender, etc.), patient airway intervention records (such as nasal cannula oxygen therapy, face mask oxygen therapy, oropharyngeal airway or endotracheal intubation), patient test indicators (such as vital signs, complete blood count, blood gas analysis, coagulation function, emergency biochemistry, etc.), and patient's state of consciousness information (such as mental status, whether agitated, etc.).
[0099] Step 402: Perform correlation analysis on the patient's medical data to obtain a set of key indicators.
[0100] In practice, Pearson correlation analysis technology can be used to analyze the set of test indicators such as vital signs, blood routine, blood gas analysis, and coagulation function, and combined with prior medical knowledge, to form a set of candidate key indicators.
[0101] The Pearson correlation coefficient is a value between -1 and 1, describing the degree of linear association between two variables. The closer the absolute value of the correlation coefficient is to 1, the stronger the correlation. If the correlation coefficient is 0, it means that there is no linear association between the two variables.
[0102] The formula for calculating the Pearson correlation coefficient is as follows:
[0103]
[0104] In the formula, X and Y represent two variables, cov(X,Y) is the covariance of variables X and Y, and σx and σy are the variances of variables X and Y respectively. Y Let X and Y be the variances of variables X and Y, respectively.
[0105] The above formula can be decomposed into its rightmost form. Let X and Y represent the average values of variables X and Y respectively, and let Xi and Yi represent the i-th sample values of X and Y respectively, for a total of n samples.
[0106] For each candidate indicator in the candidate key indicator set, sort the correlation coefficients between the candidate indicator and the severity of airway obstruction by absolute value, select indicators whose absolute correlation coefficients are greater than or equal to the threshold t and add them to the recommended indicator set; otherwise, add them to the remaining indicator set.
[0107] The indicator weights calculated using the Pearson correlation coefficient can, to some extent, describe the importance of indicators in airway obstruction, but they also have certain biases. Analysis of patient medical records revealed that most patients with airway obstruction had multiple injuries, often accompanied by head, chest, and abdominal injuries. Due to these multiple injuries, changes in physiological indicators caused by other conditions can influence the assessment of airway obstruction. For example, red blood cell count has a high correlation weight, but its impact on the airway is not significant; setting a high weight for this indicator can lead to poor model generalization. Therefore, when selecting the key indicator set, it is necessary to revise the indicator weights given by the correlation analysis based on prior medical knowledge, lowering the weights of some indicators with low medical relevance to the condition. Thus, the establishment of the key indicator set first involves classifying the importance of each indicator in the recommended indicator set based on medical knowledge, then adding the classified recommended indicator set to the key indicator set; finally, the remaining indicators are screened, selecting those with significant medical relevance to the condition to be added to the key indicator set.
[0108] For example, based on the ranking results of the Pearson correlation coefficient and medical knowledge, all indicators in the key indicator set are divided into three levels, each with a different weight. Indicators within the same level have the same weight. The key indicator set is shown in Table 1 above.
[0109] In addition, numerous experiments have shown that physiological phenomena such as the patient's level of coma and agitation can influence the assessment of airway obstruction. Therefore, labels for mental status and agitation have been added to the key indicators to describe the patient's current mental state.
[0110] Step 403: Standardize the data in the key indicator set and fill in missing values to obtain the key indicator sample set.
[0111] For different patients, due to individual differences such as age, gender, and physical condition, various physiological indicators will vary within a certain range. Taking respiratory rate as an example: 60-100 is considered normal, while only below 50 or above 120 is considered serious. Because the judgment range of the normal range is relatively large, individual differences can have a significant impact on the model. Therefore, based on existing medical prior knowledge, standardized methods should be used to process the data to reduce the impact of individual differences on the model.
[0112] The specific standardization process is as follows:
[0113] If a certain indicator is within the normal range, then the value of that indicator is 0;
[0114] If the indicator is higher than or equal to the upper limit of extreme danger (there is a risk of death), then the indicator is set to 1;
[0115] If the indicator is higher than the normal range and lower than the extreme danger upper limit, then the indicator value is mapped to the range of 0 to 1;
[0116] If the indicator is lower than or equal to the extremely dangerous lower limit, then the indicator is set to -1;
[0117] If the indicator is below the normal range and above the extremely dangerous lower limit, then the indicator value will be mapped to the range of 0 to -1.
[0118] For example, in the basic vital signs, respiratory rate (RR): when RR < 8, it is recorded as -1, indicating that the RR value is too low and has reached an extremely dangerous level; when RR ≥ 8 and RR < 12, the recorded value is scaled down to between -1 and 0; when RR ≥ 12 and RR < 24, it is considered normal and is recorded as 0; when RR ≥ 24 and RR < 34, the recorded value is scaled down to between 0 and 1; when RR ≥ 34, it is recorded as 1, indicating that the RR value is too high and has reached an extremely dangerous level. For other test indicators, the normal range and the extreme values that are too high or too low can be calculated using statistical methods combined with prior medical knowledge.
[0119] Because patients may not have certain indicators tested for various reasons, some indicator records may be missing. For missing data imputation, after standardizing the patient data, the K-Nearest Neighbor (KNN) method is first used to predict missing values. For patients with a missing indicator k, the KNN imputation method first selects patient data from all patient data where indicator k is not missing and other indicators have relatively few missing values as the training sample set T. The missing indicator k is removed, and the similarity between the patient and each sample in T is calculated. N samples most similar to the patient are selected, and the average value of the missing indicator is calculated as the imputation value. If the patient has too many missing values, making it difficult to compare similarity with other samples, or if there are no N samples similar to the patient in the training samples, then a 0-value imputation method (normal value imputation method) is used. This imputation method can be understood as assuming that if the patient did not have the indicator tested, then the indicator is considered normal. During application, the number of indicators tested increases as the patient progresses, and the prediction accuracy also increases accordingly. For example, in the initial stage of emergency care, only basic vital signs data are available, and other indicators are missing; the prediction accuracy is relatively poor at this time. As more and more indicators are obtained, the accuracy of the prediction will also increase. Therefore, for patients with too many missing values (few detection indicators), the zero-value filling method can make a relatively accurate preliminary diagnosis.
[0120] Figure 5This invention illustrates a flowchart of filling missing values for indicators in the key indicator set, as shown in an embodiment of the invention, including the following steps:
[0121] Step 501, patient indicator k is missing.
[0122] Step 502: Determine if the number of missing indicators for the patient is too high, for example, exceeding 30% of the total number of indicators. If so, proceed to step 503; otherwise, proceed to step 507.
[0123] Step 503: Select patient data with no missing indicator k and few missing values from all patient data to construct training sample set T.
[0124] Step 504: Using all indicators except indicator k, calculate the similarity between the patient and each sample in the training sample set T.
[0125] Step 505: Determine if there are at least N samples that are highly similar to the patient, i.e., the differences in all indicators are small. If so, proceed to step 506; otherwise, proceed to step 507.
[0126] If N is too high, it will be difficult to match a large number of samples; if N is too low, it will be easily affected by individual cases. In experiments, N is generally set to 5-10.
[0127] Step 506: Use the mean of the index k of the N most similar samples as the filling value to fill the patient index k.
[0128] Step 507: Fill the missing indicator k of the patient with 0.
[0129] By filling in the missing indicators for each patient as described above, we can obtain the complete indicators for each patient. The complete indicators for each patient, along with their corresponding mental status and agitation labels, form a key indicator sample set.
[0130] Continue to refer to Figure 4 In step 404, the airway intervention label and the airway severity corresponding to the airway intervention label are determined for each sample in the key indicator sample set.
[0131] The severity of airway involvement is set according to the intervention method used. For example, no airway intervention indicates the lowest severity, and the airway severity is set to 1; the intervention method is nasal cannula oxygen inhalation, and the airway severity is set to 2; the intervention method is endotracheal intubation or tracheotomy, which indicates the highest severity, and is set to 10.
[0132] Step 405: Train an airway obstruction severity assessment model based on the key indicator sample set and the airway severity.
[0133] This airway obstruction severity assessment model can be iteratively trained using the Adam optimizer, with MAE (Mean Absolute Error) as the loss function. The MAE formula is defined as follows:
[0134]
[0135] Where N is the amount of data. This indicates the true severity of the patient's airway obstruction. The MAE value indicates the predicted severity of airway obstruction in patients. By minimizing the MAE value, the model can better fit the training data.
[0136] It should be noted that the patients mentioned in the above model establishment process refer to patients in the general sense, that is, patients with medical records, rather than referring to a specific patient or a certain type of patient.
[0137] The airway obstruction severity assessment method provided in this embodiment of the invention utilizes the aforementioned airway obstruction severity assessment model to assist relevant medical personnel in making a rapid and accurate assessment of the patient's airway obstruction and in providing timely and effective treatment, thereby improving the ability to provide emergency care under sudden circumstances.
[0138] like Figure 6 The diagram shown is a flowchart illustrating the process of determining the severity score of a patient's airway obstruction using an airway obstruction severity assessment model, as described in an embodiment of the present invention. The process includes the following steps:
[0139] Step 601: Obtain the current patient's real-time diagnostic data.
[0140] The real-time diagnostic data includes: basic vital signs, test data, and consciousness status labels.
[0141] Step 602: Extract key indicator feature data from the real-time diagnostic data.
[0142] Similar to the extraction of key indicator sets when establishing the airway obstruction severity assessment model, key indicator feature data are extracted from the current patient's immediate diagnostic data. These key indicator features mainly include some of the indicators shown in Table 1.
[0143] Step 603: Input the key indicator feature data into the airway obstruction severity assessment model, and obtain the airway obstruction severity score of the patient based on the output of the airway obstruction severity assessment model.
[0144] The output of the airway obstruction severity assessment model is an airway obstruction severity score, which can be a floating-point number.
[0145] By using the airway obstruction severity assessment model, it is possible to quickly and accurately score the severity of airway obstruction in patients, intelligently assisting emergency medical personnel in making correct assessments of whether patients have airway obstruction symptoms and their severity.
[0146] Accordingly, embodiments of the present invention also provide an airway obstruction intervention recommendation system. Figure 7 A block diagram of the system is shown. This airway obstruction intervention recommendation system includes the following modules:
[0147] Data acquisition module 701 is used to acquire the real-time diagnostic data of the current patient;
[0148] Feature extraction module 702 is used to extract key indicator feature data from the real-time diagnostic data;
[0149] The recommendation module 704 is used to determine the recommended airway obstruction intervention scheme by utilizing the key indicator feature data and the pre-established airway obstruction intervention decision model 700.
[0150] Furthermore, such as Figure 8 As shown, in another embodiment of the airway obstruction intervention recommendation system provided by the present invention, the system may further include: an assessment module 703, used to determine the severity score of the patient's airway obstruction.
[0151] Accordingly, in this embodiment, the recommendation module 704 uses the key indicator feature data, the patient's airway obstruction severity score, and the pre-established airway obstruction intervention decision model to determine the recommended airway obstruction intervention plan.
[0152] In practice, the assessment module 703 can obtain the score obtained by the medical staff based on experience, or achieve automatic scoring based on the airway obstruction severity assessment model pre-constructed based on the patient's medical data source.
[0153] In this embodiment of the invention, the airway obstruction severity assessment model includes an input layer, three hidden layers, and an output layer. The three hidden layers are fully connected to the output layer, and a random deactivation mechanism is used between adjacent hidden layers.
[0154] The airway obstruction severity assessment model can be established by a first model construction module, one specific structure of which is as follows: Figure 8 As shown, it may include the following units:
[0155] The data acquisition unit 801 is used to collect patient medical data sources and extract patient visit-related data from the medical data sources.
[0156] Data analysis unit 802 is used to perform correlation analysis on the patient's medical visit-related data to obtain a set of key indicators;
[0157] The data processing unit 803 is used to perform normalization processing and missing value filling on the indicators in the key indicator set to obtain a key indicator sample set.
[0158] The label determination unit 804 is used to determine the airway intervention label and the airway severity corresponding to the airway intervention label for each sample in the key indicator sample set.
[0159] Training unit 805 is used to train an airway obstruction severity assessment model based on the key indicator sample set and the airway severity.
[0160] Specifically, the data analysis unit may include the following sub-units:
[0161] The data parsing subunit is used to perform correlation analysis on the patient's test indicators to obtain a set of candidate key indicators.
[0162] The sorting subunit is used to sort each candidate indicator in the candidate key indicator set according to the absolute value of its correlation coefficient with the severity of airway obstruction, and select candidate indicators whose absolute value of correlation coefficient is greater than or equal to a set threshold to add to the recommended indicator set; otherwise, it is added to the remaining indicator set.
[0163] The filtering subunit is used to classify the importance of each indicator in the recommended indicator set based on medical knowledge, and then add the graded indicators in the recommended indicator set to the key indicator set; from the remaining indicator set, some indicators that are medically significant for the condition are selected and added to the key indicator set.
[0164] A label-adding subunit is used to add a mental state label and an agitation label to each training sample in the key indicator set.
[0165] In this embodiment of the invention, the structure of the airway obstruction intervention decision model has been described in detail in the previous embodiments of the method of the invention, and will not be repeated here.
[0166] It should be noted that the airway obstruction intervention decision model can be pre-built by the second model building module. Specifically, a random forest is constructed from multiple decision trees. The criterion for measuring the contribution of features in the decision trees is the Gini coefficient, and the decision result of each decision tree is the recommendation probability of an airway intervention plan. The random forest is used as the airway obstruction intervention decision model, and the output of the airway obstruction intervention decision model is the decision result of the multiple decision trees.
[0167] It should be noted that, in specific implementations, the first model construction module and / or the second model construction module can be part of the system of the present invention, or they can be independent of the system of the present invention. This embodiment of the present invention does not limit the scope of the invention in this regard. Furthermore, for more details regarding the principles and working methods of each unit and subunit in the two model construction modules, please refer to the relevant descriptions in the preceding embodiments of the method of the present invention, which will not be repeated here.
[0168] The airway obstruction severity assessment system provided in this invention pre-constructs an airway obstruction severity assessment model based on patient medical data sources. It extracts key indicator features from the patient's current real-time diagnostic data, then inputs these features into the airway obstruction severity assessment model. Based on the model's output, it obtains the patient's airway obstruction severity score. This intelligently assists emergency medical personnel in quickly and accurately determining whether a patient has airway obstruction, guiding rescue personnel to rapidly and correctly assess and effectively manage airway obstruction, thus improving the efficiency of airway obstruction treatment.
[0169] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. Furthermore, the system embodiments described above are merely illustrative. The modules and units described as separate components may or may not be physically separate; that is, they may be located on a single network unit or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0170] Those skilled in the art will understand that all or part of the steps in the above method implementation can be accomplished by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc.
[0171] Accordingly, embodiments of the present invention also provide a system for recommending methods for airway obstruction intervention. This system is an electronic device, such as a mobile terminal, computer, tablet device, personal digital assistant, etc. The electronic device may include one or more processors and a memory; wherein the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the methods described in the preceding embodiments.
[0172] The embodiments of the present invention have been described in detail above. Specific implementation methods have been used to illustrate the present invention. The descriptions of the embodiments above are only for the purpose of helping to understand the methods and systems of the present invention, and are merely some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention, and the content of this specification should not be construed as a limitation of the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A recommended method for airway obstruction intervention, characterized in that, The method includes: Acquire real-time diagnostic data of the current patient; the real-time diagnostic data includes: basic vital signs, test data, and consciousness status labels; Extract key indicator feature data from the real-time diagnostic data; The method further includes: A model for assessing the severity of airway obstruction is pre-built based on patient medical data sources; The airway obstruction severity assessment model was used to determine the severity score of the patient's airway obstruction. The airway obstruction severity assessment model constructed based on patient medical data sources includes: Collect patient medical data sources and extract patient visit-related data from the medical data sources; the patient medical data sources include: emergency database, nursing system, and hospital database; Correlation analysis was performed on the patient visit data to obtain a set of key indicators; The indicators in the key indicator set are normalized and missing values are filled to obtain the key indicator sample set. Determine the airway intervention label for each sample in the key indicator sample set and the airway severity corresponding to the airway intervention label; An airway obstruction severity assessment model was trained based on the key indicator sample set and the airway severity. The correlation analysis performed on the patient's medical visit data yielded a set of key indicators, including: The patient visit-related data were analyzed using Pearson correlation analysis, and combined with prior medical knowledge, a set of candidate key indicators was obtained. For each candidate indicator in the candidate key indicator set, sort the correlation coefficients between the candidate indicators and the severity of airway obstruction by absolute value, select candidate indicators whose absolute correlation coefficients are greater than or equal to a set threshold and add them to the recommended indicator set; otherwise, add them to the remaining indicator set. Based on medical knowledge, the importance of each indicator in the recommended indicator set is classified, and then the indicators in the classified recommended indicator set are added to the key indicator set. Select a subset of indicators from the remaining indicator set and add them to the key indicator set; Add a mental state label and an agitation label to each training sample in the set of key indicators; The method further includes: Using the key indicator feature data, the patient's airway obstruction severity score, and the pre-established airway obstruction intervention decision model, a recommended airway obstruction intervention plan is determined.
2. The method according to claim 1, characterized in that, The standardization process for the indicators in the key indicator set includes: If the indicator is within the normal range, then the indicator value is set to 0; If the indicator is higher than the normal range and lower than the extreme danger upper limit, then the indicator value will be mapped to a value in the range of 0 to 1; If the indicator is higher than or equal to the extreme danger upper limit value, then the indicator value is set to 1; If the indicator is below the normal range and above the extremely dangerous lower limit, then the indicator value will be mapped to a value in the range of 0 to -1. If the indicator is lower than or equal to the extreme danger lower limit, then the indicator is set to -1.
3. The method according to any one of claims 1 to 2, characterized in that, The method also includes establishing an airway obstruction intervention decision model in the following manner: A random forest is constructed from multiple decision trees. The Gini coefficient is used as the criterion for measuring the contribution of features in each decision tree. The decision result of each decision tree is the probability of recommending an airway intervention plan. The random forest is used as the airway obstruction intervention decision model, and the output of the airway obstruction intervention decision model is the decision result of the multiple decision trees.
4. An airway obstruction intervention recommendation system, characterized in that, The system includes: The data acquisition module is used to acquire the current patient's real-time diagnostic data; the real-time diagnostic data includes: basic vital signs, test data, and consciousness status tags. The feature extraction module is used to extract key indicator feature data from the real-time diagnostic data; The system also includes: The first model building module is used to pre-build an airway obstruction severity assessment model based on patient medical data sources; the patient medical data sources include: emergency database, nursing system, and hospital database. The assessment module is used to determine the severity score of the patient's airway obstruction using the airway obstruction severity assessment model; The first model building module includes: The data acquisition unit is used to collect patient medical data sources and extract patient visit-related data from the medical data sources. The data analysis unit is used to perform correlation analysis on the patient's medical visit-related data to obtain a set of key indicators. The data processing unit is used to perform normalization processing and missing value filling on the indicators in the key indicator set to obtain a key indicator sample set. The label determination unit is used to determine the airway intervention label and the airway severity corresponding to the airway intervention label for each sample in the key indicator sample set. The training unit is used to train an airway obstruction severity assessment model based on the key indicator sample set and the airway severity. The data analysis unit includes: The data parsing subunit analyzes the patient visit-related data using Pearson correlation analysis technology and combines it with prior medical knowledge to obtain a set of candidate key indicators. The sorting subunit is used to sort each candidate indicator in the candidate key indicator set according to the absolute value of its correlation coefficient with the severity of airway obstruction, and select candidate indicators whose absolute value of correlation coefficient is greater than or equal to a set threshold to add to the recommended indicator set; otherwise, it is added to the remaining indicator set. The filtering subunit is used to classify the importance of each indicator in the recommended indicator set based on medical knowledge, and then add the graded indicators in the recommended indicator set to the key indicator set; and select some indicators from the remaining indicator set to add to the key indicator set. A label-adding subunit is used to add a mental state label and an agitation label to each training sample in the key indicator set; The system also includes: The recommendation module is used to determine the recommended airway obstruction intervention plan by utilizing the key indicator feature data, the patient's airway obstruction severity score, and the pre-established airway obstruction intervention decision model.
5. The system according to claim 4, characterized in that, The system also includes: The second model construction module is used to construct a random forest from multiple decision trees. The criterion for measuring the contribution of features in each decision tree is the Gini coefficient, and the decision result of each decision tree is the recommendation probability of an airway intervention plan. The random forest is used as an airway obstruction intervention decision model, and the output of the airway obstruction intervention decision model is the decision result of the multiple decision trees.
Citation Information
Patent Citations
Hypertension drug recommendation system and method for doctor-assisted judgment
CN111986809A