High altitude pulmonary edema severity grading evaluation method and system combining lasso features

Abstract

The present application relates to the technical field of biological information, and more particularly to a method and system for realizing high altitude pulmonary edema severity grading evaluation combined with LASSO features, which comprises: obtaining a clinical data set of patients who have been diagnosed with high altitude pulmonary edema, wherein the clinical data set includes several clinical features selected from a group consisting of vital signs, clinical symptoms and medical history; performing feature selection on the clinical data set by using a LASSO regression model to screen out a core feature subset; constructing and training a random forest classification model based on the feature data corresponding to the core feature subset to obtain a standard random forest classification model; and reducing a preset classification decision threshold when the grading result is determined to be unreliable according to the voting concentration. The present application improves the reliability of the evaluation result.

Description

Technical Field

[0001] This invention relates to the field of bioinformatics, and in particular to a method and system for assessing the severity of high-altitude pulmonary edema by combining LASSO features. Background Technology

[0002] High-altitude pulmonary edema (HAPE) is a life-threatening condition caused by rapid exposure to high altitudes. Its incidence is rising with the increasing popularity of travel and outdoor activities, especially when people ascend rapidly to high altitudes. Timely intervention is crucial for reducing HAPE-related mortality, as prompt treatment can have a significant impact. However, inexperienced physicians may face challenges in timely diagnosis of severely ill patients, leading to delayed treatment and potentially fatal outcomes. Therefore, developing risk-based early screening, risk stratification, and treatment tools can greatly improve patient outcomes and reduce deaths due to delayed treatment. Traditional diagnosis of HAPE relies on clinical symptoms and imaging studies. However, assessing the severity of HAPE is often subjective, and accurate screening tools for identifying severely ill patients are lacking. Recent advances in machine learning have provided new methods for assessing the risk level of HAPE. By integrating various physiological parameters and clinical indicators, machine learning models can provide more accurate and efficient assessments without increasing the burden on patients. Although previous studies have explored the application of machine learning in medical diagnosis, research on risk stratification modeling for patients with HAPE remains limited.

[0003] Chinese Patent Application Publication No. CN117059270A discloses an acute mountain sickness risk assessment system incorporating pseudo-labels based on medical prior knowledge. This system, belonging to the field of acute mountain sickness risk assessment, includes a data acquisition module, a pseudo-label construction module, a self-supervised learning module, and a downstream task risk assessment module. The data acquisition module comprises a first acquisition submodule and a second acquisition submodule. The first acquisition submodule acquires real-time monitored sequence data during intermittent hypoxia training, including human physiological signals such as blood oxygen saturation, heart rate, and respiratory rate. The second acquisition submodule acquires the individual's instantaneous Lake Louise score. This invention, through monitoring physiological signal data, mines the intrinsic attributes of dynamically monitored data, incorporates prior knowledge to label the data with pseudo-labels, and thereby constructs a self-supervised learning framework based on auxiliary tasks. This framework learns the representational information of physiological signals for use in downstream acute mountain sickness risk assessment tasks. However, the aforementioned acute mountain sickness risk assessment system incorporating pseudo-labels based on medical prior knowledge suffers from poor reliability of assessment results due to the lack of consideration for stratifying disease risk based on diagnosed patients and the lack of assurance of the credibility of stratification results in clinical decision-making. Summary of the Invention

[0004] To address this, the present invention provides a method and system for assessing the severity of high-altitude pulmonary edema by combining LASSO features. This overcomes the problems in existing technologies that fail to consider stratification of disease risk based on diagnosed patients and fail to consider the reliability of stratification results in clinical decision-making, thus affecting accurate decision-making and resulting in poor reliability of assessment results.

[0005] To achieve the above objectives, on the one hand, this invention provides a method for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO characteristics, comprising:

[0006] Obtain a clinical dataset of patients diagnosed with high-altitude pulmonary edema, wherein the clinical dataset includes several clinical features selected from a group consisting of vital signs, clinical symptoms and medical history;

[0007] The clinical dataset was subjected to feature selection using the LASSO regression model to identify a core feature subset. Based on the feature data corresponding to the core feature subset, a random forest classification model was constructed and trained to obtain a standard random forest classification model.

[0008] Based on the decision logic of the standard random forest classification model, a corresponding clinical scoring table is generated.

[0009] The standard random forest classification model assesses the severity of high-altitude pulmonary edema based on the current clinical status of the target patient's core feature subset and a preset classification decision threshold, and obtains a classification result. During the assessment process, the voting concentration of the standard random forest classification model is calculated.

[0010] When the classification result is determined to be unreliable based on the voting concentration, the preset classification decision threshold is reduced.

[0011] The grading results that meet the preset standards are output, and the scores corresponding to the core characteristics of the target patient are determined by combining the clinical scoring table to generate a grading assessment report.

[0012] Furthermore, the process of using the LASSO regression model to select features from the clinical dataset and filter out a subset of core features includes:

[0013] Using the severity level of high-altitude pulmonary edema in the patients diagnosed with high-altitude pulmonary edema as the response variable and several of the clinical characteristics as predictors, a LASSO regression model was constructed.

[0014] The optimal regularization parameters for the LASSO regression model were determined through cross-validation.

[0015] The predictor variables with non-zero coefficients under the optimal regularization parameter are selected to form the core feature subset.

[0016] Furthermore, based on the decision logic of the standard random forest classification model, a corresponding clinical scoring table is generated, including:

[0017] Extract the decision nodes and split thresholds of each decision tree in the standard random forest classification model for each feature within the core feature subset;

[0018] Based on the distribution of the splitting thresholds of each feature, the numerical range or state of each feature is divided into several ordered intervals.

[0019] Based on the feature importance score of each feature in the standard random forest classification model, a corresponding score is assigned to each ordered interval of each feature, thereby generating a structured clinical scoring table.

[0020] Furthermore, the core feature subset includes: lung rales, sputum production, heart rate, and blood oxygen saturation.

[0021] Furthermore, the severity of high-altitude pulmonary edema is classified into mild, moderate, and severe.

[0022] Further, the feature is that the clinical dataset is preprocessed, wherein,

[0023] The preprocessing includes removing variables with a missing data percentage greater than or equal to a preset percentage, and estimating variables with a missing data percentage less than a preset percentage using a random forest method.

[0024] Furthermore, following the preprocessing, the process also includes:

[0025] The preprocessed clinical dataset is divided into a training set for model training and a test set for model performance evaluation according to a preset ratio.

[0026] Furthermore, the process of determining the reliability of the grading results based on the voting concentration includes:

[0027] The voting concentration is compared with the preset concentration.

[0028] If the voting concentration is less than the preset concentration, the classification result is determined to be unreliable, and the preset classification decision threshold is reduced according to the difference between the first preset concentration and the voting concentration.

[0029] If the voting concentration is greater than or equal to the preset concentration, the classification result of the standard random forest classification model is directly output.

[0030] Furthermore, the process of obtaining the voting concentration includes:

[0031] The core feature subset of the target patient is input into the standard random forest classification model;

[0032] Obtain the independent predicted category for the patient output by each decision tree in the standard random forest classification model;

[0033] The number of decision trees corresponding to each severity level in the prediction categories output by all decision trees is counted as the number of votes for each level.

[0034] The ratio of the number of votes received by the severity level with the highest number of votes to the total number of decision trees included in the standard random forest classification model is denoted as the voting concentration.

[0035] On the other hand, the present invention also provides a system suitable for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO features, comprising:

[0036] The data acquisition module is used to acquire clinical datasets of patients diagnosed with high-altitude pulmonary edema.

[0037] The feature selection module, which is connected to the data acquisition module, is used to execute the LASSO regression model to select features from the clinical dataset and filter out a subset of core features.

[0038] The model evaluation module, which is connected to the feature selection module, is used to construct a standard random forest classification model based on the core feature subset, and to use the standard random forest classification model to classify the severity of high-altitude pulmonary edema in the target patient.

[0039] A data processing module, which is connected to the model evaluation module, is used to obtain the voting concentration of the standard random forest classification model;

[0040] The control module, which is connected to the model evaluation module and the data processing module respectively, is used to determine whether the grading result of the severity grading of high-altitude pulmonary edema of the target patient is reliable based on the voting concentration.

[0041] An evaluation output module, which is connected to the control module, is used to generate a graded evaluation report.

[0042] Compared with the prior art, the beneficial effects of the present invention are that the core features screened by the present invention through LASSO are all readily available indicators, and the output classification labels and structured clinical scoring tables can help to clearly understand the scoring basis, quickly locate key abnormal indicators, and quickly and accurately identify critically ill patients. At the same time, accurate triage helps to rationally triage patients, thereby improving the treatment efficiency and effectiveness of the medical system.

[0043] Furthermore, this invention uses the LASSO regression model to accurately screen datasets containing a large number of clinical features, effectively eliminating redundant information and irrelevant variables, focusing on core predictive factors, and avoiding interference from irrelevant data on model performance. Based on the screened core feature subset, a random forest classification model is constructed and trained, giving full play to the model's advantages in handling nonlinear relationships and complex feature interactions, accurately achieving three-level stratification of mild, moderate, and severe high-altitude pulmonary edema, thereby ensuring the reliability of the assessment and grading results.

[0044] Furthermore, by specifically defining the core feature subset as lung rales, sputum production, heart rate, and blood oxygen saturation, this invention has extremely high indicator specificity. All features are clinically readily obtainable signs or symptoms, making it suitable for medical conditions in high-altitude areas. This ensures the high feasibility of the entire evaluation method in resource-limited high-altitude emergency environments.

[0045] Furthermore, this invention quantifies the degree of consensus in the random forest model by setting a voting concentration, and optimizes unreliable results by increasing the number of decision trees, thereby filtering out ambiguous results with large disagreements in the decision trees and improving the accuracy of classification. Attached Figure Description

[0046] Figure 1 This is a flowchart illustrating the logic of a method for grading and assessing the severity of high-altitude pulmonary edema based on LASSO features, as described in an embodiment of the present invention.

[0047] Figure 2 This is a flowchart illustrating the process of selecting a subset of core features in an embodiment of the present invention that combines LASSO features to implement a method for grading the severity of high-altitude pulmonary edema.

[0048] Figure 3 This is a flowchart illustrating the process of determining the reliability of the grading results of the method for assessing the severity of high-altitude pulmonary edema by combining LASSO features in an embodiment of the present invention.

[0049] Figure 4 The present invention provides an overall structural diagram of a high-altitude pulmonary edema severity grading and assessment system that incorporates LASSO features. Detailed Implementation

[0050] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0051] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0052] Please see Figure 1 The diagram shown is a logical flowchart of a method for grading the severity of high-altitude pulmonary edema using LASSO features, as described in this embodiment of the invention. The method includes:

[0053] Step S1: Obtain a clinical dataset of patients diagnosed with high-altitude pulmonary edema, wherein the clinical dataset includes several clinical features selected from a group consisting of vital signs, clinical symptoms and medical history.

[0054] Step S2: Use the LASSO regression model to select features from the clinical dataset, filter out the core feature subset, and construct and train a random forest classification model based on the feature data corresponding to the core feature subset to obtain the standard random forest classification model.

[0055] Step S3: Based on the decision logic of the standard random forest classification model, generate the corresponding clinical scoring table;

[0056] Step S4: The standard random forest classification model assesses the severity of high-altitude pulmonary edema based on the current clinical status of the target patient's core feature subset and a preset classification decision threshold, and obtains a classification result. During the assessment process, the voting concentration of the standard random forest classification model is calculated.

[0057] Step S5: When the classification result is determined to be unreliable based on the voting concentration, the preset classification decision threshold is reduced.

[0058] Step S6: Output the grading results that meet the preset standards, and determine the scores corresponding to the core characteristics of the target patient in combination with the clinical scoring table to generate a grading assessment report.

[0059] Specifically, the clinical dataset includes several clinical features selected from a group consisting of vital signs, clinical symptoms, and medical history. These features include, for example, age, sex, race, body mass index, clinical symptoms (including fever, malaise, loss of appetite, nausea, sleep disturbances, dizziness, cough, chest tightness, shortness of breath, palpitations, cyanosis of the lips, and cough); vital signs (including body temperature, heart rate, blood oxygen saturation, systolic blood pressure, and diastolic blood pressure); demographic information (including altitude of residence, altitude of onset, mode of transportation to high-altitude areas, whether bathing and catching a cold were observed after arriving at high altitude, and the number of days / time of the first symptom onset after arriving at high altitude); and other information (including smoking history, alcohol consumption history, hypertension, diabetes, chronic respiratory diseases, and chronic liver disease). The clinical features described in this invention are not limited to the specific items listed in the embodiments, and those skilled in the art can adjust them according to actual needs.

[0060] Specifically, based on comprehensive clinical assessment and imaging reports, the grading criteria for mild, moderate, and severe pulmonary edema in patients with high-altitude pulmonary edema are as follows:

[0061] Mild severity level, bilateral lung marking changes with a few flocculent and patchy shadows;

[0062] Moderate severity level, with flocculent / patchy / exudative / ground-glass opacities in the middle and lower lobes of both lungs, or changes in the inner zone of both lungs;

[0063] Severe severity level, with numerous flocculent / patchy / exudative / ground-glass opacities in both lungs, or clinical assessment indicating critical condition requiring ICU admission.

[0064] Table 1. Clinical characteristics of patients with mild, moderate and severe high-altitude pulmonary edema.

[0065] feature Mild, N=671 Moderate, N=2401 Severe, N=291 p-value gender 0.3 female 18（27%） 45（19%） 7（24%） male 49（73%） 195（81%） 22（76%） age 35(27,42) 31(26,42) 40(27,44) 0.089 height 169(165,172) 170(165,173) 169(163,171) 0.3 weight 74(64,77) 74 (64,77) 75(69,78) 0.5 BMI 24.9(22.9,26.3) 24.8(23.0,26.4) 25.5(23.8,26.4) 0.6 fatigue 15(22%) 60(25%) 7(24%) >0.9 Loss of taste 15(22%) 56(23%) 6(21%) >0.9 nausea 12（18%） 46（19%） 12（41%） 0.017 Sleep disorders 17(25%) 47(20%) 8(28%) 0.4 Dizziness 37(55%) 155(65%) 21(72%) 0.2 cough 57(85%) 205(85%) 25(86%) >0.9 Shortness of breath 34(51%) 125(52%) 16(55%) >0.9 indigestion 17(25%) 54(23%) 7(24%) 0.9 Heart palpitations 12(18%) 46(19%) 4(14%) 0.8 Cyanosis of the lips 11(16%) 36(15%) 2(6.9%) 0.5 sputum >0.9 have 62(93%) 219(91%) 27(93%) none 5(7.5%) 20(8.3%) 2(6.9%) Lung rales 2.2E-16 none 11(16%) 55(23%) 2(6.9%) A few rales at the bases of one or both lungs 41(61%) 124(52%) 3(10%) Moderate rales in the middle or lower lobes of one or both lungs 15(22%) 60(25%) 2(6.9%) Numerous rales throughout the lungs 0(0%) 1 (0.4%) 22(76%) sputum 1.03(0.99,1.38) 1.00 (0.79,1.32) 1.00(1.00,1.22) 0.3 body temperature 36.50(36.30,36.70) 36.50(36.40,36.80) 36.50(36.50,36.80) 0.4 Heart rate 90(80,100) 100(87,119) 100(92,115) ＜0.001 blood oxygen saturation 80(74,86) 74(65,84) 74(63,85) 0.007 systolic blood pressure 122(117,133) 120(111,130) 120(117,130) 0.3 diastolic blood pressure 80(74,88) 80(70,87) 77(70,83) 0.6 CT ＜0.001 normal 7(10%) 2(0.8%) 0(0%) Changes in lung texture 10(15%) 8(3.3%) 2(6.9%) Patchy / flocculent shadows in the lower lobes of both lungs 50(75%) 21(8.8%) 7(24%) other 0(0%) 209(87%) 20(69%)

[0066] As can be seen from the table, there are statistically significant differences (p < 0.05) in features such as nausea, lung rales, heart rate, and blood oxygen saturation among different severity groups. This initially suggests the correlation between these features and disease severity, providing a clinical basis for subsequent feature selection.

[0067] Please see Figure 2 As shown, it is a logical flowchart of the process of selecting a subset of core features in an embodiment of the present invention.

[0068] Specifically, the process of using the LASSO regression model to select features from the clinical dataset and filter out a subset of core features includes:

[0069] Using the severity level of high-altitude pulmonary edema in the patients diagnosed with high-altitude pulmonary edema as the response variable and several of the clinical characteristics as predictors, a LASSO regression model was constructed.

[0070] The optimal regularization parameters for the LASSO regression model were determined through cross-validation.

[0071] The predictor variables with non-zero coefficients under the optimal regularization parameter are selected to form the core feature subset.

[0072] In this embodiment, the method uses the LASSO regression model to accurately screen a dataset containing a large number of clinical features, effectively eliminating redundant information and irrelevant variables, focusing on core predictive factors, and avoiding interference from irrelevant data on model performance. Based on the screened core feature subset, a random forest classification model is constructed and trained, giving full play to the model's advantages in handling nonlinear relationships and complex feature interactions, accurately achieving three-level stratification of mild, moderate, and severe high-altitude pulmonary edema, thereby ensuring the reliability of the assessment and grading results.

[0073] In this embodiment, the core feature subset selected includes lung rales, sputum production, heart rate, and blood oxygen saturation.

[0074] Specifically, a clinical dataset is a sample set of clinical characteristic observations of patients, including vital signs data, clinical symptoms data, and medical history data.

[0075] Specifically, clinical features include blood oxygen saturation, heart rate, respiratory rate, body temperature, and blood pressure.

[0076] Specifically, the LASSO regression model is a classification feature selection model based on L1 regularization.

[0077] Specifically, the corresponding clinical scoring table is generated based on the decision logic of the standard random forest classification model, including:

[0078] Extract the decision nodes and split thresholds of each decision tree in the standard random forest classification model for each feature within the core feature subset;

[0079] Based on the distribution of the splitting thresholds of each feature, the numerical range or state of each feature is divided into several ordered intervals.

[0080] Based on the feature importance score of each feature in the standard random forest classification model, a corresponding score is assigned to each ordered interval of each feature, thereby generating a structured clinical scoring table.

[0081] Specifically, the feature importance score is determined based on the degree of improvement in classification purity brought about by the corresponding feature during the splitting process of each decision tree node in the standard random forest classification model. It is obtained by accumulating the amount of impurity reduction of the node corresponding to the feature in each decision tree, and the amount of impurity reduction is the node splitting contribution value calculated based on the change in Gini index.

[0082] In this embodiment, the specific process of generating a clinical scoring table based on the decision logic of the standard random forest classification model is as follows: traversing the decision nodes of each decision tree in the standard random forest classification model, and statistically analyzing the splitting threshold used by each decision tree for each feature within the core feature subset. For example, for blood oxygen saturation, the splitting thresholds that appear frequently in the decision nodes of each decision tree include 74, 80, and 85, etc., and based on this, blood oxygen saturation is divided into four ordered intervals: ≥85%, 75%~84%, 65%~74%, and <65%; for heart rate, heart rate is divided into four ordered intervals: <90 beats / min, 90~99 beats / min, 100~119 beats / min, and ≥120 beats / min; for lung rales, they are divided into four ordered levels: no rales, a few rales at the base of one or both lungs, moderate rales in the middle and lower lobes of one or both lungs, and a large number of rales throughout the lungs; for cough and sputum, they are divided into none and two ordered levels. Subsequently, based on the feature importance scores of each feature in the standard random forest classification model, corresponding scores are assigned to each ordered interval of the above features. The scores between adjacent intervals within each feature increase in an arithmetic progression, and the score range of each feature is positively correlated with its feature importance score.

[0083] Specifically, the core feature subset includes: lung rales, coughing up sputum, heart rate, and blood oxygen saturation.

[0084] Specifically, the severity of high-altitude pulmonary edema is classified into mild, moderate, and severe levels.

[0085] Specifically, the feature is that the clinical dataset is preprocessed, wherein,

[0086] The preprocessing includes removing variables with a missing data percentage greater than or equal to a preset percentage, and estimating variables with a missing data percentage less than a preset percentage using a random forest method.

[0087] Specifically, the preset percentage is used to measure the missing proportion of clinical variable data. It is determined based on the data quality requirements of the clinical dataset and the number of effective samples required for model training. It can be understood that the ratio of the number of missing data for each clinical variable in the clinical dataset to the total number of samples corresponding to that clinical variable is used as the missing proportion of the clinical variable. When the missing proportion is greater than or equal to the preset percentage, it indicates that the number of effective observation samples for that clinical variable is insufficient and cannot stably reflect the patient's clinical status, so the variable is removed.

[0088] Specifically, after the preprocessing, the process also includes:

[0089] The preprocessed clinical dataset is divided into a training set for model training and a test set for model performance evaluation according to a preset ratio.

[0090] Specifically, the preprocessed clinical dataset is divided according to the patient sample dimension. The patient samples of the first proportion are used as the training set and input into the LASSO regression model and the random forest classification model for model construction. The remaining proportion of patient samples are used as the test set to verify the accuracy of the constructed model in classifying the severity of unknown patient samples. The preset proportion can be determined according to the number of clinical samples and the model training requirements.

[0091] Specifically, the test set is a collection of patient samples that were independently partitioned from the preprocessed clinical dataset and were not involved in the model training process.

[0092] Please see Figure 3 As shown, it is a logical flowchart of the process for determining whether the grading result is reliable in an embodiment of the present invention.

[0093] Specifically, the process of determining the reliability of the grading results based on the voting concentration includes:

[0094] The voting concentration is compared with the preset concentration.

[0095] If the voting concentration is less than the preset concentration, the classification result is determined to be unreliable, and the preset classification decision threshold is reduced according to the difference between the first preset concentration and the voting concentration.

[0096] If the voting concentration is greater than or equal to the preset concentration, the classification result of the standard random forest classification model is directly output.

[0097] Specifically, the reduction in the preset classification decision threshold is positively correlated with the difference between the first preset concentration and the voting concentration. This positive correlation can be linear or nonlinear. The slope of the linear positive correlation is not specifically limited. It can be understood that when the voting concentration is lower, it indicates a greater difference in classification opinions among the decision trees in the standard random forest classification model. Therefore, the adjustment range of the classification decision threshold is increased. In other words, the greater the difference between the first preset concentration and the voting concentration, the greater the reduction in the preset classification decision threshold.

[0098] Specifically, the process of obtaining the voting concentration includes:

[0099] The core feature subset of the target patient is input into the standard random forest classification model;

[0100] Obtain the independent predicted category for the patient output by each decision tree in the standard random forest classification model;

[0101] The number of decision trees corresponding to each severity level in the prediction categories output by all decision trees is counted as the number of votes for each level.

[0102] The ratio of the number of votes received by the severity level with the highest number of votes to the total number of decision trees included in the standard random forest classification model is denoted as the voting concentration.

[0103] Specifically, in both the random forest and decision tree models, heart rate and blood oxygen saturation were identified as the two most important predictors, followed by lung rales and sputum production.

[0104] In this embodiment, to verify the superiority of the random forest model constructed in this scheme, it was compared and evaluated with a multinomial logistic regression model and a decision tree model constructed based on the same core feature subset. The performance evaluation results on the independent test set are shown in Table 2.

[0105] Table 2 Performance evaluation of different models

[0106] performance Logistic Regression (Feature Selection) Decision tree (feature selection) Random Forest (Feature Selection) accuracy 0.720588 0.735294 0.779412 Accuracy 0.535088 0.615 0.702655 Recall rate 0.571111 0.662222 0.682222 F1 score 0.551921 0.632641 0.689552

[0107] The data in Table 2 show that the random forest model constructed in this embodiment of the invention significantly outperforms the comparison model on all evaluation metrics, with an accuracy of 77.94%, proving that the technical solution of combining the core features selected by LASSO with the random forest model has higher prediction accuracy.

[0108] Specifically, when constructing the LASSO regression model, the severity level of high-altitude pulmonary edema of the patients is used as the response variable, and the preprocessed clinical features are used as predictor variables. The cross-validation error of the model under different regularization parameters is evaluated by ten-fold cross-validation, and the regularization parameter with the smallest cross-validation error is selected as the optimal regularization parameter. Under the optimal regularization parameter, the predictor variables with non-zero coefficients constitute the core feature subset.

[0109] In practice, this invention quantifies the degree of consensus in the random forest model by setting a voting concentration, and optimizes unreliable results by increasing the number of decision trees, thereby filtering out ambiguous results with large disagreements in the decision trees and improving the accuracy of classification.

[0110] Please see Figure 4 The diagram shown is the overall structural block diagram of the high-altitude pulmonary edema severity grading and assessment system that combines LASSO features in this embodiment. The system described in this embodiment includes:

[0111] The data acquisition module is used to acquire clinical datasets of patients diagnosed with high-altitude pulmonary edema.

[0112] The feature selection module, which is connected to the data acquisition module, is used to execute the LASSO regression model to select features from the clinical dataset and filter out a subset of core features.

[0113] The model evaluation module, which is connected to the feature selection module, is used to construct a standard random forest classification model based on the core feature subset, and to use the standard random forest classification model to classify the severity of high-altitude pulmonary edema in the target patient.

[0114] A data processing module, which is connected to the model evaluation module, is used to obtain the voting concentration of the standard random forest classification model;

[0115] The control module, which is connected to the model evaluation module and the data processing module respectively, is used to determine whether the grading result of the severity grading of high-altitude pulmonary edema of the target patient is reliable based on the voting concentration.

[0116] An evaluation output module, which is connected to the control module, is used to generate a graded evaluation report.

[0117] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO features, characterized in that, include: Obtain a clinical dataset of patients diagnosed with high-altitude pulmonary edema, wherein the clinical dataset includes several clinical features selected from a group consisting of vital signs, clinical symptoms and medical history; The clinical dataset was subjected to feature selection using the LASSO regression model to identify a core feature subset. Based on the feature data corresponding to the core feature subset, a random forest classification model was constructed and trained to obtain a standard random forest classification model. Based on the decision logic of the standard random forest classification model, a corresponding clinical scoring table is generated. The standard random forest classification model assesses the severity of high-altitude pulmonary edema based on the current clinical status of the target patient's core feature subset and a preset classification decision threshold, and obtains a classification result. During the assessment process, the voting concentration of the standard random forest classification model is calculated. When the classification result is determined to be unreliable based on the voting concentration, the preset classification decision threshold is reduced. The grading results that meet the preset standards are output, and the scores corresponding to the core characteristics of the target patient are determined by combining the clinical scoring table to generate a grading assessment report.

2. The method for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO features according to claim 1, characterized in that, The process of using the LASSO regression model to select features from the clinical dataset and filter out a subset of core features includes: Using the severity level of high-altitude pulmonary edema in the patients diagnosed with high-altitude pulmonary edema as the response variable and several of the clinical characteristics as predictors, a LASSO regression model was constructed. The optimal regularization parameters for the LASSO regression model were determined through cross-validation. The predictor variables with non-zero coefficients under the optimal regularization parameter are selected to form the core feature subset.

3. The method for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO features according to claim 2, characterized in that, The decision logic based on the standard random forest classification model generates a corresponding clinical scoring table, including: Extract the decision nodes and split thresholds of each decision tree in the standard random forest classification model for each feature within the core feature subset; Based on the distribution of the splitting thresholds of each feature, the numerical range or state of each feature is divided into several ordered intervals. Based on the feature importance score of each feature in the standard random forest classification model, a corresponding score is assigned to each ordered interval of each feature, thereby generating a structured clinical scoring table.

4. The method for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO features according to claim 3, characterized in that, The core feature subset includes lung rales, sputum production, heart rate, and blood oxygen saturation.

5. The method for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO features according to claim 4, characterized in that, The severity of high-altitude pulmonary edema is classified into mild, moderate, and severe.

6. The method for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO features according to claim 5, characterized in that, The clinical dataset is preprocessed, wherein, The preprocessing includes removing variables with a missing data percentage greater than or equal to a preset percentage, and estimating variables with a missing data percentage less than a preset percentage using a random forest method.

7. The method for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO features according to claim 6, characterized in that, Following the preprocessing, the following is also included: The preprocessed clinical dataset is divided into a training set for model training and a test set for model performance evaluation according to a preset ratio.

8. The method for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO features according to claim 7, characterized in that, The process of determining the reliability of the grading results based on the voting concentration includes: The voting concentration is compared with the preset concentration. If the voting concentration is less than the preset concentration, the classification result is determined to be unreliable, and the preset classification decision threshold is reduced according to the difference between the first preset concentration and the voting concentration. If the voting concentration is greater than or equal to the preset concentration, the classification result of the standard random forest classification model is directly output.

9. The method for grading and assessing the severity of high-altitude pulmonary edema by combining LASSO features according to claim 8, characterized in that, The process of obtaining the voting concentration includes: The core feature subset of the target patient is input into the standard random forest classification model; Obtain the independent predicted category for the patient output by each decision tree in the standard random forest classification model; The number of decision trees corresponding to each severity level in the prediction categories output by all decision trees is counted as the number of votes for each level. The ratio of the number of votes received by the severity level with the highest number of votes to the total number of decision trees included in the standard random forest classification model is denoted as the voting concentration.

10. A grading and assessment system for assessing the severity of high-altitude pulmonary edema using the LASSO feature method as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire clinical datasets of patients diagnosed with high-altitude pulmonary edema. The feature selection module, which is connected to the data acquisition module, is used to execute the LASSO regression model to select features from the clinical dataset and filter out a subset of core features. The model evaluation module, which is connected to the feature selection module, is used to construct a standard random forest classification model based on the core feature subset, and to use the standard random forest classification model to classify the severity of high-altitude pulmonary edema in the target patient. A data processing module, which is connected to the model evaluation module, is used to obtain the voting concentration of the standard random forest classification model; The control module, which is connected to the model evaluation module and the data processing module respectively, is used to determine whether the grading result of the severity grading of high-altitude pulmonary edema of the target patient is reliable based on the voting concentration. An evaluation output module, which is connected to the control module, is used to generate a graded evaluation report.

Images