Rapid dehydration detection method and device based on urine biomarkers

Abstract

The invention provides a rapid dehydration detection method based on urine biomarkers, which comprises the following steps: collecting a urine sample, and detecting the concentrations of three markers of electrolyte, metabolite and physical property in the urine sample; modeling the concentration of the electrolyte marker by using a polynomial regression model to generate an electrolyte score; modeling the concentration of the metabolite marker by using a support vector regression model based on a kernel method to generate a metabolite score; carrying out modeling on the physical characteristic type markers by utilizing a gradient boosting tree model to generate a physical characteristic score; environment conditions are introduced into all the sub-models for dynamic calculation of environment factor correction, and the dynamic score of each marker is obtained; and weighting according to the dynamic score to obtain a final dehydration score, and outputting a dehydration degree classification and a corresponding suggestion.

Description

Technical Field

[0001] The invention belongs to the field of medical detection, and in particular relates to a rapid dehydration detection method and device based on urine biomarkers. Background Art

[0002] In high-intensity field work environments, workers often face extreme work intensity and harsh environmental conditions, such as high temperature, low humidity, and long hours of physical labor, which can cause dehydration in workers. Dehydration can significantly affect workers' physical strength, endurance, and cognitive function, and in severe cases can even endanger their lives. Therefore, timely detection of dehydration and taking effective countermeasures are the key to ensuring work efficiency and personnel safety.

[0003] Symptoms of dehydration include thirst, fatigue, dizziness, and lack of concentration. If not detected and intervened in time, it may lead to serious health problems such as heat stroke, dehydration shock, and even life-threatening conditions. In a field work environment, it is difficult for workers to obtain timely medical diagnosis and professional equipment support, so traditional dehydration detection methods (such as blood tests, urine tests, etc.) are often not applicable. These traditional methods require not only specialized equipment, but also professional operators. The detection process is cumbersome, and obtaining test results often takes a long time.

[0004] Existing dehydration detection methods, such as traditional detection methods based on blood tests and urine analysis, usually rely on laboratory equipment and the support of laboratory technicians. Indicators such as electrolytes and metabolites in blood and urine can help doctors determine the degree of dehydration, but these methods are not suitable for on-site detection in field operations. This is because these methods usually require specialized detection equipment, a stable power supply, and professional operating skills, which are often not met in high-intensity field work environments.

[0005] Currently, researchers have tried to combine the information of multiple markers through machine learning and multivariate regression analysis to achieve more accurate dehydration assessment. Although these methods have achieved certain results in laboratory environments, in the field, how to quickly obtain marker data and how to process and analyze data in real time to obtain dehydration assessment results is still a difficult problem that needs to be solved.

[0006] In addition, the impact of environmental factors (such as temperature and humidity) on dehydration cannot be ignored. Traditional dehydration detection methods often ignore the interference of environmental factors. Therefore, it is necessary to develop a comprehensive model that can consider the impact of environmental factors to more accurately assess the risk of dehydration.

[0007] Therefore, there is an urgent need for a new dehydration detection technology suitable for high-intensity field work environments. This technology should be able to integrate multiple biomarkers and environmental factors to accurately assess the degree of dehydration in real time. Through portable equipment, workers can quickly and non-invasively detect dehydration status on site, and then take timely intervention measures to ensure work efficiency and personnel health, and reduce safety hazards caused by dehydration. Summary of the invention

[0008] In view of this, the present invention provides a rapid dehydration detection method based on urine biomarkers, comprising: Step 1, collecting urine samples and testing the concentrations of three types of markers: electrolytes, metabolites, and physical properties; Step 2: Use a polynomial regression model to model the concentration of electrolyte markers and generate an electrolyte score; use a support vector regression model based on a kernel method to model the concentration of metabolite markers and generate a metabolite score; use a gradient boosting tree model to model physical property markers and generate a physical property score; Step 3, introducing environmental conditions into each sub-model to perform dynamic calculation of environmental factor correction to obtain a dynamic score for each marker; Step 4, calculating a final dehydration score based on the weighted average of the dynamic scores of each of the markers, and outputting a dehydration degree classification and corresponding suggestions based on the final dehydration score.

[0009] In particular, the nonlinear relationship between the electrolyte markers can be modeled by polynomial regression: , in: stands for electrolyte score, which indicates the contribution of electrolytes to dehydration status; Representative input characteristics include: electrolyte concentrations of sodium, potassium, chloride, calcium, or magnesium; Represents the number of Gaussian distributions, indicating the complexity of the model; Indicates Gaussian distribution weights satisfy ; Indicates Gaussian distribution with mean The covariance is .

[0010] , represents the prediction score of metabolite markers, The characteristic vector of the input metabolites, including the concentrations of urea, uric acid, creatinine, lactate, and amino acids, Indicates the identity of the metabolite; represents the Lagrange multipliers of the support vectors, which represent the parameters of the model; Indicates the index training samples, Represents the kernel function, which is used to calculate the input features and training samples The similarity between them maps the input features to a high-dimensional feature space; Indicates that the training data set has samples.

[0011] In particular, the function expression of the gradient boosting tree model is: , in, It is a marker of the physical characteristics of the input and model parameters The calculated training prediction value, is the loss function, It is The true labels of samples, is the total number of training samples; the physical property markers include: urine specific gravity, osmotic pressure, urine color; is the regularization term, and its expression is: , in, is the number of leaf nodes in a single tree, It is The weight of a leaf node, and is a hyperparameter that controls the complexity of the tree, Represents the basis function in the model; the gradient boosting tree model obtains the physical property score of the physical property marker by combining multiple decision trees.

[0012] In particular, the physical property scores are It means that its expression is:

[0013] in represents the total number of trees; Indicates Trees for input features The predicted output; the model parameters obtained by training the gradient boosting tree model For predicting the dehydration score of the physical property marker, It is The weight of a tree indicates the contribution of the tree's output to the final prediction.

[0014] In particular, the dynamic calculation of introducing environmental conditions into each sub-model for environmental factor correction specifically includes: using the following expression to correct the environmental factors for each marker score , , indicating that at time , marker type , ambient temperature , pH of urine Under the influence of is a constant coefficient used for the basic impact weights of marker concentration and score; represents the marker concentration function; represents the spatial distribution or variation of the physical properties of urine, i represents the row index of stress, and j represents the column index of stress; represents the stress constant, which is used as the reference value for normalizing the stress tensor; represents the nonlinear index, R represents the ideal gas constant, is the stress gradient coefficient, which reflects the correction effect of stress change on the dehydration score; the correction term constant is used to adjust the benchmark of the entire score. is a constant related to the marker type, which represents the natural attenuation of the marker concentration change or the influence of external conditions; Represents the external correction term that affects the concentration of the marker; the dynamic score of each marker is calculated according to the dynamic scoring function of each marker.

[0015] In particular, calculating the final dehydration score according to the weighted average of the dynamic scores of each of the markers specifically includes: iteratively optimizing the weight coefficients of each dynamic score by a gradient descent method.

[0016] In particular, the dehydration degree classification is determined according to the relationship between the final dehydration score and a predetermined threshold.

[0017] In particular, the dehydration degree classification includes: mild dehydration, moderate dehydration or severe dehydration.

[0018] The present invention also discloses a rapid dehydration detection device based on urine biomarkers, comprising: The sample collection and testing module is used to collect urine samples and detect the concentrations of three types of markers: electrolytes, metabolites, and physical properties; The marker score modeling module is used to model the concentration of electrolyte markers using a polynomial regression model to generate an electrolyte score, to model the concentration of metabolite markers using a support vector regression model based on a kernel method to generate a metabolite score; and to model physical property markers using a gradient boosting tree model to generate a physical property score. A dynamic scoring module, used to introduce environmental conditions into each sub-model to perform dynamic calculation of environmental factor correction to obtain a dynamic score for each marker; The dehydration degree classification module is used to calculate a final dehydration score according to the weighted average of the dynamic scores of each of the markers, and output a dehydration degree classification and corresponding suggestions according to the final dehydration score.

[0019] Beneficial effects: 1. High-precision dehydration detection: The present invention combines the concentration data of electrolyte, metabolite and physical property urine biomarkers, and uses advanced machine learning models for modeling and prediction, which can achieve accurate assessment of the degree of dehydration, thereby providing clinicians with more accurate dehydration diagnosis results.

[0020] 2. Strong real-time performance and wide applicability: The rapid detection method adopted by the present invention does not rely on complex laboratory equipment, can complete the detection and analysis of urine samples in a short time, and has high real-time performance. This technology is not only suitable for clinical diagnosis, but also can be widely used in scenarios such as athlete health monitoring and elderly health management, meeting the health management needs of various groups.

[0021] 3. Strong environmental adaptability: The present invention introduces an environmental correction term. By adjusting the model parameters to adapt to changes in different environmental conditions (such as temperature, humidity, etc.), the adaptability of the model can be improved, the interference of the external environment on the test results can be avoided, and the reliability and stability of the dehydration degree assessment can be ensured.

[0022] 4. Personalized prediction: By introducing environmental correction items and modeling methods targeting different individual characteristics, the present invention can provide personalized dehydration predictions and diagnostic suggestions for each sample, and accurately assess the degree of dehydration based on different physiological states, living habits and environmental conditions.

[0023] 5. Optimized machine learning algorithm: The present invention adopts advanced machine learning algorithms such as support vector regression (SVR) based on kernel method and gradient boosting tree (XGBoost) for modeling, which can effectively process high-dimensional data and complex nonlinear relationships, thereby providing high-accuracy dehydration score prediction.

[0024] 6. Low cost and easy to promote and apply: The present invention provides a low-cost, high-efficiency urine dehydration detection method. The required equipment is simple and easy to operate, and can be quickly deployed and applied in various environments. This technology has good promotion prospects, especially suitable for rapid detection in the field, battlefield, sports field and other conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a schematic diagram of the process of the rapid dehydration detection method based on urine biomarkers proposed in the present invention; Figure 2 This is a schematic diagram of the rapid dehydration detection device based on urine biomarkers proposed in the present invention. DETAILED DESCRIPTION

[0026] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0027] A rapid dehydration detection method based on urine biomarkers, such as Figure 1 As shown, including: Step 1, collect urine samples and test the concentrations of three types of markers: electrolytes, metabolites, and physical properties; sample collection specifically includes: Sample source: first urination sample in the morning or random urine sample. First urine in the morning: usually highly concentrated and can more accurately reflect the body's dehydration status. Random urine: may be more convenient in practical applications and used for rapid screening. Collection method: Use a sterile urine collector to ensure that the sample volume is sufficient, usually 10-20 ml. Seal the sample immediately after collection to avoid environmental contamination. Storage conditions: If short-term storage (<2 hours) is required, place it at room temperature. When stored for a long time (>2 hours), it needs to be placed in a refrigerated environment of 2-8°C to avoid degradation of chemical components.

[0028] In this embodiment, the markers are classified into: electrolyte markers, metabolite markers and physical property markers. The target detection indicators of the electrolyte markers include: sodium ion (Na⁺), potassium ion (K⁺), chloride ion (Cl⁻), calcium ion (Ca²⁺), magnesium ion (Mg²⁺). Detection method: Ion selective electrode method (ISE): Detect ion concentration by selective electrode. Atomic absorption spectroscopy (AAS): Detect trace metal ion concentration. Detection results: Concentration range: displayed in standard units mmol / L.

[0029] Target detection indicators of the metabolite markers: urea, creatinine, uric acid, lactic acid, amino acid metabolites. Detection methods: Enzyme-linked immunosorbent assay (ELISA): used for quantitative analysis of specific metabolites. High-performance liquid chromatography (HPLC): separation and precise quantification of metabolites. Nuclear magnetic resonance spectroscopy (NMR): analysis of complex metabolite mixtures. Detection results: Concentration range is expressed in mg / dL or μmol / L.

[0030] The target detection indicators of the physical property markers include: urine specific gravity, osmotic pressure, and color.

[0031] Test method: The specific gravity of urine is measured by hydrometer (unit: g / mL), the osmotic pressure is determined by freezing point depression method (unit: mOsm / kg), and the colorimeter is used to analyze the color of urine (compared with standard color cards). Test results: The corresponding values ​​are provided according to the physical parameters.

[0032] This step also includes: marker concentration data recording and preprocessing, data recording: use standardized instruments to record data and save them as spreadsheet files (such as CSV format). Each sample data includes marker category, concentration value and unit. Data cleaning: exclude outliers, such as results that exceed the physiological range (according to medical standards). Missing data are supplemented or excluded, using interpolation or mean filling methods.

[0033] Step 2, using a polynomial regression model to model the concentration of electrolyte markers and generate an electrolyte score; The nonlinear relationship between the electrolyte markers can be modeled by polynomial regression: , in: stands for electrolyte score, which indicates the contribution of electrolytes to dehydration status; Representative input features include: electrolyte concentrations of sodium, potassium, and chloride; Represents the number of Gaussian distributions, indicating the complexity of the model; Indicates Gaussian distribution weights satisfy ; Indicates Gaussian distribution with mean The covariance is .

[0034] , represents the prediction score of metabolite markers, The input metabolite feature vector includes: urea, uric acid or creatinine concentration, Indicates the identity of the metabolite; represents the Lagrange multipliers of the support vectors, which represent the parameters of the model; Indicates the index training samples, Represents the kernel function, which is used to calculate the input features and training samples The similarity between them maps the input features to a high-dimensional feature space; Indicates that the training data set has samples.

[0035] The physical property markers were modeled using a gradient boosting tree model to generate a physical property score; the function expression of the gradient boosting tree model was: , in, It is a marker of the physical characteristics of the input and model parameters The calculated training prediction value, is the loss function, It is The true labels of samples, is the total number of training samples; the physical property markers include: urine specific gravity, osmotic pressure, urine color; is the regularization term, and its expression is: , in, is the number of leaf nodes in a single tree, It is The weight of a leaf node, and is a hyperparameter that controls the complexity of the tree, Represents the basis function in the model; the gradient boosting tree model obtains the physical property score of the physical property marker by combining multiple decision trees.

[0036] In particular, the physical property scores are It means that its expression is: , in represents the total number of trees; Indicates Trees for input features The predicted output; the model parameters obtained by training the gradient boosting tree model For predicting the dehydration score of the physical property marker, It is The weight of a tree indicates the contribution of the tree's output to the final prediction.

[0037] Step 3, introducing environmental conditions into each sub-model to perform dynamic calculation of environmental factor correction, and obtain a dynamic score of each marker; introducing environmental conditions into each sub-model to perform dynamic calculation of environmental factor correction specifically includes: using the following expression to perform environmental factor correction on each marker score , , indicating that at time , marker type , ambient temperature , pH of urine Under the influence of, the dynamic scoring function of the marker, type can be Indicates electrolyte markers, which can be , indicating metabolite markers, can be , indicating physical property markers. is a constant coefficient used for the basic impact weights of marker concentration and score; represents the marker concentration function; Represents the spatial distribution or variation of the physical properties of urine, i represents the row index of stress, j The column index representing stress; represents the stress constant, which is used as the reference value for normalizing the stress tensor; represents the nonlinear index; R is the ideal gas constant, is the stress gradient coefficient, reflecting the correction effect of stress change on the dehydration score; the correction term constant is often used to adjust the benchmark of the entire score. is a constant related to the marker type, which represents the natural attenuation of the marker concentration change or the influence of external conditions; Represents an external correction term that affects the marker concentration.

[0038] According to the scores of each marker, the environmental factor correction function is calculated to obtain the dynamic score of each marker, for example: , in, and is a constant term; The identifier representing the sample result, represents the dynamic scoring function of electrolyte markers, represents the dynamic scoring function of metabolite markers, A dynamic scoring function representing a physical property marker.

[0039] Step 4: Calculate the final dehydration score by weighted average based on the dynamic scores of electrolyte markers, metabolite markers, and physical property markers, and output the dehydration degree classification and corresponding recommendations based on the final dehydration score.

[0040] The final dehydration score is calculated by weighted average based on the dynamic scores of electrolyte markers, the dynamic scores of metabolite markers and the dynamic scores of physical property markers, which specifically includes: iteratively optimizing the weight coefficients of each score through the gradient descent method.

[0041] The final dehydration score was calculated as: is the final dehydration score, , , is the weighting coefficient,

[0042] Define the optimized loss function , used to adjust the weight , , To minimize the scoring error: , in: It is The actual dehydration degree labels of the samples, N is the total number of samples.

[0043] or is the Huber loss function , Implementing weight adjustment , , process.

[0044] Threshold parameter , determines the switching point where the loss function changes from square loss to linear loss. Usually, Adjust according to the characteristics of the data set. is the predicted output of the model, is the true value; the role of the loss function: when the model prediction error is less than δ, the square error is used for penalty; when the error is greater than δ, the linear loss is used for penalty, thereby reducing the sensitivity to outliers.

[0045] Gradient descent is achieved by iteratively optimizing weights through the gradient descent method. . Let the initial weight be

[0046] Calculate the gradient of the loss function Find the partial derivatives and update the weights using the learning rate η: , After each update, ensure that the weights satisfy the normalization constraints and obtain the value of the normalized weights. , The stopping condition is the loss function When the change is less than a certain threshold (such as 10-6), the maximum number of iterations is reached.

[0047] The dehydration degree classification is determined based on the relationship between the final dehydration score and the predetermined threshold. The dehydration degree classification includes: mild dehydration, moderate dehydration or severe dehydration. Based on the final dehydration score, it is divided into three categories: mild, moderate and severe dehydration: Mild Dehydration: Less than or equal to a lower threshold. Moderate Dehydration: Falls between mild dehydration and severe dehydration. Severe Dehydration: exceeds some upper threshold.

[0048] Assume, either analytically or empirically, that the following scoring thresholds are applicable: Mild dehydration: ≤0.4; moderate dehydration: 0.4< ≤0.70; severe dehydration: >0.7.

[0049] Example 1: Assume that a patient's final dehydration score is =0.25, according to the threshold: A score of 0.25 is less than or equal to 0.4, so the patient is diagnosed with mild dehydration.

[0050] Example 2: Assume that a patient's final dehydration score is =0.6, according to the threshold, so the patient is diagnosed with moderate dehydration.

[0051] Example 3: Assume that a patient's final dehydration score is =0.8, according to the threshold score, so the patient is diagnosed with severe dehydration. For mild dehydration, patients are advised to increase water intake and monitor symptoms. For moderate dehydration, treatment with oral rehydration or appropriate intravenous rehydration is recommended. For severe dehydration, intravenous fluids should be given immediately and further treatment should be received in the hospital.

[0052] The present invention also discloses a rapid dehydration detection device based on urine biomarkers, such as Figure 2 As shown, including: The sample collection and detection module is used to collect urine samples and detect the concentrations of three types of markers: electrolytes, metabolites and physical properties; collect urine samples and detect the concentrations of three types of markers: electrolytes, metabolites and physical properties; sample collection specifically includes: sample source: first urination sample in the morning or random urine sample. First urine in the morning: usually highly concentrated and can more accurately reflect the body's dehydration status. Random urine: may be more convenient in practical applications and used for rapid screening. Collection method: Use a sterile urine collector to ensure that the sample volume is sufficient, usually 10-20 ml. Seal the sample immediately after collection to avoid environmental contamination. Storage conditions: If it needs to be stored for a short time (<2 hours), place it at room temperature. When stored for a long time (>2 hours), it needs to be placed in a refrigerated environment of 2-8°C to avoid degradation of chemical components.

[0053] In this embodiment, the markers are classified into: electrolyte markers, metabolite markers and physical property markers. The target detection indicators of the electrolyte markers include: sodium ion (Na⁺), potassium ion (K⁺), chloride ion (Cl⁻), calcium ion (Ca²⁺), magnesium ion (Mg²⁺). Detection method: Ion selective electrode method (ISE): Detect ion concentration by selective electrode. Atomic absorption spectroscopy (AAS): Detect trace metal ion concentration. Detection results: Concentration range: displayed in standard units mmol / L.

[0054] Target detection indicators of the metabolite markers: urea, creatinine, uric acid, lactic acid, amino acid metabolites. Detection methods: Enzyme-linked immunosorbent assay (ELISA): used for quantitative analysis of specific metabolites. High-performance liquid chromatography (HPLC): separation and precise quantification of metabolites. Nuclear magnetic resonance spectroscopy (NMR): analysis of complex metabolite mixtures. Detection results: Concentration range is expressed in mg / dL or μmol / L.

[0055] The target detection indicators of the physical property markers include: urine specific gravity, osmotic pressure, and color.

[0056] Test method: The specific gravity of urine is measured by hydrometer (unit: g / mL), the osmotic pressure is determined by freezing point depression method (unit: mOsm / kg), and the colorimeter is used to analyze the color of urine (compared with standard color cards). Test results: The corresponding values ​​are provided according to the physical parameters.

[0057] This module also includes: marker concentration data recording and preprocessing, data recording: use standardized instruments to record data and save them as spreadsheet files (such as CSV format). Each sample data includes marker category, concentration value and unit. Data cleaning: exclude outliers, such as results that exceed the physiological range (according to medical standards). Missing data are supplemented or excluded, using interpolation or mean filling methods.

[0058] The marker score modeling module is used to model the concentration of electrolyte markers using a polynomial regression model to generate an electrolyte score, model the concentration of metabolite markers using a support vector regression model based on a kernel method to generate a metabolite score; model physical property markers using a gradient boosting tree model to generate a physical property score; and model the concentration of electrolyte markers using a polynomial regression model to generate an electrolyte score; The nonlinear relationship between the electrolyte markers can be modeled by polynomial regression: , in: stands for electrolyte score, which indicates the contribution of electrolytes to dehydration status; Representative input features include: electrolyte concentrations of sodium, potassium, and chloride; Represents the number of Gaussian distributions, indicating the complexity of the model; Indicates Gaussian distribution weights satisfy ; Indicates Gaussian distribution with mean The covariance is .

[0059] , represents the prediction score of metabolite markers, The input metabolite feature vector includes: urea, uric acid or creatinine concentration, Indicates the identity of the metabolite; represents the Lagrange multipliers of the support vectors, which represent the parameters of the model; Indicates the index training samples, Represents the kernel function, which is used to calculate the input features and training samples The similarity between them maps the input features to a high-dimensional feature space; Indicates that the training data set has samples.

[0060] The physical property markers were modeled using a gradient boosting tree model to generate a physical property score; the function expression of the gradient boosting tree model was: , in, It is a marker of the physical characteristics of the input and model parameters The calculated training prediction value, is the loss function, It is The true labels of samples, is the total number of training samples; the physical property markers include: urine specific gravity, osmotic pressure, urine color; is the regularization term, and its expression is: , in, is the number of leaf nodes in a single tree, It is The weight of a leaf node, and is a hyperparameter that controls the complexity of the tree, Represents the basis function in the model; the gradient boosting tree model obtains the physical property score of the physical property marker by combining multiple decision trees.

[0061] In particular, the physical property scores are It means that its expression is: , in represents the total number of trees; Indicates Trees for input features The predicted output; the model parameters obtained by training the gradient boosting tree model For predicting the dehydration score of the physical property marker, It is The weight of a tree indicates the contribution of the tree's output to the final prediction.

[0062] The dynamic scoring module is used to introduce environmental conditions into each sub-model to perform dynamic calculation of environmental factor correction to obtain a dynamic score for each marker; the dynamic calculation of introducing environmental conditions into each sub-model to perform environmental factor correction specifically includes: using the following expression to perform environmental factor correction on each marker score , , indicating that at time , marker type , ambient temperature , pH of urine The dynamic scoring function of the marker under the influence; type can be Indicates electrolyte markers, which can be , indicating metabolite markers, can be , indicating physical property markers. is a constant coefficient used for the basic impact weights of marker concentration and score; represents the marker concentration function; Represents the spatial distribution or variation of the physical properties of urine, i represents the row index of stress, j The column index representing stress; represents the stress constant, which is used as the reference value for normalizing the stress tensor; represents the nonlinear index; R is the ideal gas constant, is the stress gradient coefficient, reflecting the correction effect of stress change on the dehydration score; the correction term constant is often used to adjust the benchmark of the entire score. is a constant related to the marker type, which represents the natural attenuation of the marker concentration change or the influence of external conditions; Represents an external correction term that affects the marker concentration.

[0063] According to the scores of each marker, the environmental factor correction function is calculated to obtain the dynamic score of each marker. According to the scores of each marker, the environmental factor correction function is calculated to obtain the dynamic score of each marker. For example: , in, and is a constant term; The identifier representing the sample result, represents the dynamic scoring function of electrolyte markers, represents the dynamic scoring function of metabolite markers, A dynamic scoring function representing a physical property marker.

[0064] The dehydration degree classification module is used to calculate the final dehydration score by weighted average based on the dynamic scores of electrolyte markers, the dynamic scores of metabolite markers and the dynamic scores of physical property markers, and output the dehydration degree classification and corresponding suggestions based on the final dehydration score.

[0065] The final dehydration score is calculated by weighted average based on the dynamic scores of electrolyte markers, the dynamic scores of metabolite markers and the dynamic scores of physical property markers, which specifically includes: iteratively optimizing the weight coefficients of each score through the gradient descent method.

[0066] The final dehydration score was calculated as: is the final dehydration score, , , is the weighting coefficient,

[0067] Define the optimized loss function , used to adjust the weight , , To minimize the scoring error:

[0068] in: It is The actual dehydration degree labels of the samples, N is the total number of samples.

[0069] or is the Huber loss function

[0070] Implementing weight adjustment , , process.

[0071] Threshold parameter , determines the switching point where the loss function changes from square loss to linear loss. Usually, Adjust according to the characteristics of the data set. is the predicted output of the model, is the true value; the role of the loss function: when the model prediction error is less than δ, the square error is used for penalty; when the error is greater than δ, the linear loss is used for penalty, thereby reducing the sensitivity to outliers.

[0072] Gradient descent is achieved by iteratively optimizing weights through the gradient descent method. . Let the initial weight be

[0073] Calculate the gradient of the loss function Find the partial derivatives and update the weights using the learning rate η:

[0074] After each update, ensure that the weights satisfy the normalization constraints and obtain the value of the normalized weights. , The stopping condition is the loss function When the change is less than a certain threshold (such as 10-6), the maximum number of iterations is reached.

[0075] The dehydration degree classification is determined based on the relationship between the final dehydration score and the predetermined threshold. The dehydration degree classification includes: mild dehydration, moderate dehydration or severe dehydration. Based on the final dehydration score, it is divided into three categories: mild, moderate and severe dehydration: Mild Dehydration: Less than or equal to a lower threshold. Moderate Dehydration: Falls between mild dehydration and severe dehydration. Severe Dehydration: Exceeding some higher threshold. Assume, through analysis or experience, that the following scoring thresholds are applicable: Mild dehydration: ≤0.4; moderate dehydration: 0.4< ≤0.70; severe dehydration: >0.7 Example 1: Assume that a patient's final dehydration score is =0.25, according to the threshold: A score of 0.25 is less than or equal to 0.4, so the patient is diagnosed with mild dehydration.

[0076] Example 2: Assume that a patient's final dehydration score is =0.6, according to the threshold, so the patient is diagnosed with moderate dehydration.

[0077] Example 3: Assume that a patient's final dehydration score is =0.8, according to the threshold score, the patient is diagnosed with severe dehydration. For mild dehydration, it is recommended that patients increase water intake and monitor symptoms. For moderate dehydration, treatment is recommended through oral rehydration or appropriate intravenous rehydration. For severe dehydration, intravenous infusion should be performed immediately and further treatment should be received in the hospital. In summary, the above is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0078] It is obvious to those skilled in the art that the embodiments of the present invention are not limited to the details of the above exemplary embodiments, and that the embodiments of the present invention can be implemented in other specific forms without departing from the spirit or basic features of the embodiments of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the embodiments of the present invention is limited by the attached claims rather than the above description, so it is intended to include all changes that fall within the meaning and scope of the equivalent elements of the claims in the embodiments of the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units, modules or devices stated in the system, device or terminal claims can also be implemented by the same unit, module or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

[0079] Finally, it should be noted that the above implementation modes are only used to illustrate the technical solutions of the embodiments of the present invention and are not intended to limit them. Although the embodiments of the present invention have been described in detail with reference to the above preferred implementation modes, those skilled in the art should understand that the technical solutions of the embodiments of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A rapid dehydration detection method based on urine biomarkers, characterized in that: include: Step 1, collecting urine samples and testing the concentrations of three types of markers: electrolytes, metabolites, and physical properties; Step 2: Model the concentration of electrolyte markers using a polynomial regression model to generate an electrolyte score; model the concentration of metabolite markers using a support vector regression model based on a kernel method to generate a metabolite score; The physical property markers were modeled using a gradient boosting tree model to generate physical property scores; Step 3, introducing environmental conditions into each sub-model to perform dynamic calculation of environmental factor correction to obtain a dynamic score for each marker; Step 4, calculating a final dehydration score based on the weighted average of the dynamic scores of each of the markers, and outputting a dehydration degree classification and corresponding suggestions based on the final dehydration score.

2. The rapid dehydration detection method based on urine biomarkers according to claim 1, characterized in that: The nonlinear relationship between the electrolyte markers can be modeled by polynomial regression: ,in: stands for electrolyte score, which indicates the contribution of electrolytes to dehydration status; Representative input characteristics include: electrolyte concentrations of sodium, potassium, chloride, calcium, or magnesium; Represents the number of Gaussian distributions, indicating the complexity of the model; Indicates Gaussian distribution weights satisfy ; Indicates Gaussian distribution with mean The covariance is .

3. The rapid dehydration detection method based on urine biomarkers according to claim 1, characterized in that: The method of modeling the concentration of metabolite markers using a support vector regression model based on a kernel method includes: , represents the prediction score of metabolite markers, The characteristic vector of the input metabolites, including the concentrations of urea, uric acid, creatinine, lactate, and amino acids, Indicates the identity of the metabolite; represents the Lagrange multipliers of the support vectors, which represent the parameters of the model; Indicates the index training samples, Represents the kernel function, which is used to calculate the input features and training samples The similarity between them maps the input features to a high-dimensional feature space; Indicates that the training data set has samples.

4. The rapid dehydration detection method based on urine biomarkers according to claim 1, characterized in that: The function expression of the gradient boosting tree model is: ， in, It is a marker of the physical characteristics of the input and model parameters The calculated training prediction value, is the loss function, It is The true labels of samples, is the total number of training samples; the physical property markers include: urine specific gravity, osmotic pressure, urine color; is the regularization term, and its expression is: , in, is the number of leaf nodes in a single tree, It is The weight of a leaf node, and is a hyperparameter that controls the complexity of the tree, Represents the basis function in the model; the gradient boosting tree model obtains the physical property score of the physical property marker by combining multiple decision trees.

5. The rapid dehydration detection method based on urine biomarkers according to claim 4, characterized in that: The physical properties were scored using It means that its expression is: ， in represents the total number of trees; Indicates Trees for input features The predicted output of The model parameters obtained by training the gradient boosting tree model For predicting the dehydration score of the physical property marker, It is The weight of a tree indicates the contribution of the tree's output to the final prediction.

6. The rapid dehydration detection method based on urine biomarkers according to claim 5, characterized in that: The dynamic calculation of introducing environmental conditions into each sub-model for environmental factor correction specifically includes: using the following expression to correct the environmental factors for each marker score ， , indicating that at time , marker type , ambient temperature , pH of urine Under the influence of is a constant coefficient used for the basic impact weights of marker concentration and score; represents the marker concentration function; represents the spatial distribution or variation of the physical properties of urine, i represents the row index of stress, and j represents the column index of stress; represents the stress constant, which is used as the reference value for normalizing the stress tensor; represents the nonlinear index, R represents the ideal gas constant, is the stress gradient coefficient, reflecting the corrective effect of stress change on the dehydration score; Correction constant, used to adjust the benchmark of the entire score, is a constant related to the marker type, which represents the natural attenuation of the marker concentration change or the influence of external conditions; Represents the external correction term that affects the concentration of the marker; the dynamic score of each marker is calculated according to the dynamic scoring function of each marker.

7. The rapid dehydration detection method based on urine biomarkers according to claim 6, characterized in that: Calculating the final dehydration score according to the weighted average of the dynamic scores of each marker specifically includes: iteratively optimizing the weight coefficients of each dynamic score by a gradient descent method.

8. The rapid dehydration detection method based on urine biomarkers according to claim 7, characterized in that: The dehydration degree classification is determined according to the relationship between the final dehydration score and a predetermined threshold.

9. The rapid dehydration detection method based on urine biomarkers according to claim 8, characterized in that: The dehydration degree classification includes: mild dehydration, moderate dehydration or severe dehydration.

10. A rapid dehydration detection device based on urine biomarkers, characterized in that: include: The sample collection and testing module is used to collect urine samples and detect the concentrations of three types of markers: electrolytes, metabolites, and physical properties; A marker score modeling module is used to model the concentration of electrolyte markers using a polynomial regression model to generate an electrolyte score, and to model the concentration of metabolite markers using a support vector regression model based on a kernel method to generate a metabolite score; The physical property markers were modeled using a gradient boosting tree model to generate physical property scores; A dynamic scoring module, used to introduce environmental conditions into each sub-model to perform dynamic calculation of environmental factor correction to obtain a dynamic score for each marker; The dehydration degree classification module is used to calculate a final dehydration score according to the weighted average of the dynamic scores of each of the markers, and output a dehydration degree classification and corresponding suggestions according to the final dehydration score.

Images