Metabolic flexibility assessment method and system based on respiration quotient

Abstract

The invention provides a metabolism flexibility assessment method and system based on respiration quotients. The metabolism flexibility assessment method comprises the following steps: S1, collecting a plurality of respiration quotients, questionnaire information and body measurement data of individuals with normal, abnormal and damaged metabolism flexibility; s2, performing data preprocessing and feature extraction on the data, and constructing a metabolic flexibility data set; s3, training a plurality of machine learning models by using the metabolic flexibility data set, and adjusting and optimizing to obtain candidate models; and S4, fusing the candidate models by using a stack integration method, and constructing a metabolic flexibility evaluation model. According to the method, the respiration quotient and derivative indexes of the respiration quotient serve as core characteristics, questionnaire information and body measurement data serve as auxiliary parameters, a metabolism flexibility evaluation model is established, and non-invasive and rapid glycolipid metabolism evaluation is achieved.

Description

Technical Field

[0001] This invention belongs to the field of medical device technology, and in particular relates to a method and system for assessing metabolic flexibility based on respiratory quotient. Background Technology

[0002] Metabolic flexibility refers to an organism's ability to adaptively regulate substrate oxidation to maintain energy balance and metabolic homeostasis under different metabolic demands and environmental conditions. Metabolic flexibility provides more support to the body, while a decrease in metabolic flexibility disrupts the homeostasis of substrate oxidation, leading to a series of metabolic diseases such as obesity and diabetes. Therefore, screening the body's metabolic flexibility level is of great significance for the early prevention of metabolic diseases. Traditional methods for screening metabolic flexibility, such as the oral glucose tolerance test, require frequent blood draws and are complex to perform, resulting in high compliance among test subjects.

[0003] The human body's energy supply mainly relies on three major nutrients (carbohydrates, fats, and proteins). Based on the different respiratory quotients (RQs) for each individual substrate used as a measure of consumption—1.0 for carbohydrates, 0.7 for fats, and 0.8 for proteins—the RQ reflects the percentage of each substrate used by the body at the cellular level. A characteristic of metabolic inflexibility is a decreased capacity for fat oxidation during fasting, while the capacity for carbohydrate oxidation does not increase accordingly after eating; that is, the RQ is higher during fasting and the increase after eating is not significant.

[0004] In the prior art, an invention entitled "A Method for Assessing Human Glucose and Lipid Metabolism" (CN116115196A) proposes a method to assess human glucose and lipid metabolism. This method involves measuring fasting blood glucose levels and respiratory quotient (RQ) 1 hour and 2 hours after glucose administration, and comparing the differences in changes in blood glucose levels and RQ to assess abnormalities in glucose and lipid metabolism. However, this method still requires multiple blood samples, failing to avoid the invasive procedures of traditional methods. Furthermore, its respiratory quotient data collection frequency is low and the time span is insufficient; it only uses raw measurements for comparison without any in-depth extraction of derived indicators, thus failing to comprehensively capture and quantify the dynamic changes in respiratory quotient after glucose administration, limiting the accuracy and depth of the assessment.

[0005] There are three main existing methods for screening glucose and lipid metabolism flexibility: 1) the hyperinsulin clamp test, which is complex, costly, and invasive; 2) the incremental exercise-lactate test, which is complex, requires high-intensity exercise and frequent blood draws; and 3) the oral glucose tolerance test, which is simple but also requires frequent blood draws. All of these methods are invasive and require high compliance from participants. Therefore, there is an urgent need for a simple, non-invasive technique to assess metabolic flexibility. Summary of the Invention

[0006] To address the above technical issues, this invention proposes a metabolic flexibility assessment method and system based on respiratory quotient. Using respiratory quotient and its derived indicators as core features, and combining questionnaire information and body measurement data as auxiliary parameters, a metabolic flexibility assessment model is established to achieve non-invasive and rapid assessment of glucose and lipid metabolism.

[0007] The specific technical solution is as follows:

[0008] A method for assessing metabolic flexibility based on respiratory quotient includes the following steps:

[0009] Step S1: Collect respiratory quotient, questionnaire information and body measurement data of several individuals with normal, abnormal and impaired metabolic flexibility. This data is used to collect the resting respiratory quotient values ​​of individuals at 30 min, 60 min, 120 min and 180 min after fasting and oral glucose administration, and to obtain questionnaire information and body measurement data at the same time.

[0010] Step S2: Perform data preprocessing and feature extraction on the data to calculate metabolic recovery degree, metabolic change rate, metabolic peak, peak recovery time, metabolic persistence, metabolic stability index, metabolic response area and metabolic peak width, forming high-dimensional features and constructing a metabolic flexibility dataset.

[0011] Step S3: Train and fine-tune various machine learning models using the metabolic flexibility dataset to obtain candidate models;

[0012] Step S4: Use the stacked ensemble method to fuse candidate models and construct a metabolic flexibility assessment model.

[0013] A system for assessing metabolic flexibility based on respiratory quotient, the system comprising:

[0014] Data entry module: This module is responsible for receiving and integrating raw data from multiple sources, and ensuring that all data conforms to a unified standard format and data quality requirements when it is entered;

[0015] Data storage module: This module is responsible for securely storing the entered raw data and provides an interface for the feature extraction module to call the data;

[0016] Feature Engineering Module: This module mainly performs preprocessing and feature extraction operations on the stored data, including calculation and normalization of respiratory quotient-derived indicators, to generate metabolic flexibility data suitable for candidate model input;

[0017] Metabolic flexibility assessment module: This module integrates candidate models and metabolic flexibility assessment models, takes preprocessed metabolic flexibility data as input, performs model evaluation, and generates metabolic flexibility assessment results;

[0018] Report generation module: This module generates reports that visually display an individual's metabolic flexibility assessment results and raw data, including respiratory quotient, questionnaire information, and body measurement data;

[0019] Model feedback and update module: Receives all information and feedback entered by doctors, and semi-automatically retrains and updates the candidate model and the evaluation model.

[0020] The present invention has the following beneficial effects:

[0021] This invention proposes an evaluation system based on respiratory quotient and its derived indicators: taking respiratory quotient as the core feature, it calculates and combines its derived indicators such as metabolic recovery degree, metabolic change rate, metabolic peak, peak recovery time, metabolic persistence and metabolic stability index, metabolic response area, and metabolic peak width to construct a model to achieve metabolic flexibility assessment.

[0022] This invention is non-invasive and easy to operate: relying on respiratory quotient and routine physical data, it avoids the invasive operation of frequent blood draws in traditional methods, reduces the risk of infection and improves patient compliance.

[0023] This invention automates the evaluation process: the system achieves fully automated processing from data entry to evaluation report generation through the collaborative work of various modules, significantly reducing the need for human intervention.

[0024] This invention assesses metabolic flexibility by collecting respiratory quotient data from individuals after fasting and glucose administration using an indirect calorimetry method. It completely replaces traditional blood sample monitoring and innovatively calculates respiratory quotient-derived indicators to comprehensively describe an individual's metabolic state, improving the accuracy and versatility of the metabolic flexibility model assessment and providing an efficient and universally applicable tool for assessing metabolic flexibility in clinical practice. Attached Figure Description

[0025] Figure 1 This is a flowchart of the metabolic flexibility assessment method based on respiratory quotient of the present invention;

[0026] Figure 2 This is a flowchart of the feature engineering process of the present invention;

[0027] Figure 3 This is a flowchart illustrating the construction of the metabolic flexibility assessment model for this invention.

[0028] Figure 4 This is a block diagram of the metabolic flexibility assessment system of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above objectives, this invention adopts the following technical solution.

[0030] This invention proposes a method for assessing metabolic flexibility based on respiratory quotient, such as... Figure 1 As shown, it includes the following steps:

[0031] S1: Collect respiratory quotient, questionnaire information, and anthropometric data from several individuals with normal, abnormal, and impaired metabolic flexibility;

[0032] S2: Perform data preprocessing and feature extraction on the data to construct a metabolic flexibility dataset;

[0033] S3: Train and optimize various machine learning models using the metabolic flexibility dataset to obtain candidate models;

[0034] S4: Use a stacked ensemble method to fuse candidate models and construct a metabolic flexibility assessment model.

[0035] Furthermore, step S1 includes:

[0036] The respiratory quotient (RQ) is the resting respiratory quotient value measured by indirect calorimetry at 30 min, 60 min, 120 min, and 180 min after fasting and oral administration of glucose solution. These values ​​are denoted as follows: The questionnaire information includes age, gender, daily activity intensity, history of diabetes, family history of diabetes, and other medical conditions; body measurement data includes height, weight, BMI, waist circumference, hip circumference, waist-to-hip ratio, systolic blood pressure, diastolic blood pressure, body fat mass, body fat percentage, and blood oxygen saturation.

[0037] Furthermore, such as Figure 2 As shown, step S2 includes:

[0038] (1) Preprocess the data, including unifying data types, normalization, and handling missing values;

[0039] Normalization method:

[0040] The data is normalized using the Z-Score standardization method, and the result is calculated using the following formula:

[0041] ;

[0042] Where X is the original feature value, μ is the mean of the feature in the dataset, σ is the standard deviation of the feature in the dataset, and the standardized feature value Z is calculated.

[0043] Missing value handling methods:

[0044] For numerical features (such as BMI, systolic blood pressure, etc.), the missing values ​​are filled using the mean of this feature in the dataset, calculated using the following formula:

[0045] ;

[0046] Let n represent the missing data for a certain type of sample in the dataset, and n be the total number of samples of that type in the dataset that are not missing. This refers to data from the dataset that includes samples of this type that are not missing.

[0047] Use the mode to fill in categorical features (such as intensity of daily activities, family history of diabetes, etc.).

[0048] For the respiratory quotient time series data (including missing values ​​at 30 min, 60 min, and 120 min after glucose administration), missing values ​​were imputed using linear interpolation, calculated using the following formula:

[0049] ;

[0050] in, and These are valid time points before and after the missing time points. For the missing time points of fasting and 180 minutes after glucose administration, which cannot be filled by linear interpolation, the mean of this feature in the dataset is used to fill the missing values.

[0051] (2) The respiratory quotient index was integrated and expanded to calculate derivative indicators such as metabolic recovery degree, metabolic change rate, metabolic peak, peak recovery time, metabolic persistence and metabolic stability index, metabolic response area, and metabolic peak width.

[0052] Metabolic Recovery Degree (MRD) is defined as the difference between the respiratory quotient at each time point after glucose administration and the fasting respiratory quotient. It directly reflects the dynamic recovery ability of a subject's metabolic state after glucose load stimulation and is calculated using the following formula:

[0053] ;

[0054] in, This is the baseline value for the fasting respiratory quotient. After taking sugar minute The respiratory quotient at that time.

[0055] The metabolic recovery rate (MRR) is defined as the rate of change in metabolic recovery between adjacent time points after glucose administration. It is used to quantify the dynamic efficiency of metabolic recovery and is calculated using the following formula:

[0056] ,in, > ;

[0057] The metabolic peak (MP) is defined as the peak of metabolic recovery after glucose administration, reflecting the maximum efficiency achieved by the metabolic system, and is calculated using the following formula:

[0058] ;

[0059] Peak Recovery Time (PRT) is defined as the time when metabolic recovery reaches its peak, representing the time required for the metabolic system to reach maximum efficiency. It is calculated using the following formula:

[0060] ;

[0061] argmax is a mathematical operation that finds the input value that maximizes the function.

[0062] Metabolic Recovery Persistence (MPR) is defined as the difference between peak metabolic rate and metabolic recovery at 180°C after glucose administration. It reflects the performance of the metabolic system in maintaining maximum efficiency and is calculated using the following formula:

[0063] ;

[0064] The Metabolic Stability Index (MSI) is defined as the ratio of the standard deviation of metabolic recovery at each time point to the mean. It reflects the stability of the metabolic system and is calculated using the following formula:

[0065] ;

[0066] Calculation The mean, that is, through The calculated total indivual The mean;

[0067] The metabolic response area (MRA) is defined as the area under the respiratory quotient curve calculated using the trapezoidal integral method after a glucose load. It reflects the cumulative effect of the overall metabolic response and is calculated using the following formula:

[0068] ;

[0069] Metabolic peak width (PW) is defined as the time difference between when metabolic recovery reaches 80% of the peak metabolic threshold and when it exits the high-response region. It reflects the duration of the metabolic system's maintenance of a high metabolic state and is calculated using the following formula:

[0070] ;

[0071] in, and Determined by linear interpolation, the formula is:

[0072] ;

[0073] .

[0074] (3) Feature extraction is performed using a recursive feature elimination method based on Support Vector Machine (SVM). Let the original feature set be... The dataset to be filtered, S=X, is initialized. Then, a recursive feature elimination iteration is performed. In each iteration, an SVM model is trained using the current dataset S. SVM is an algorithm model in machine learning (Support Vector Machine), not any module described in this patent, and is based on the absolute value of the weight vector. Each feature is assigned an importance score, and the features are ranked according to their importance scores. The k features with the lowest scores are removed, and the dataset S is updated. The SVM model is retrained using the updated dataset S, and its performance is evaluated through cross-validation. The iteration terminates when the dimension of the dataset S decreases to a preset value d, or the model performance decreases by more than a threshold ε, or the model performance improves by no more than a threshold λ. Finally, the metabolic flexibility dataset is constructed.

[0075] Furthermore, step S3 includes:

[0076] Multiple machine learning algorithms were selected, including logistic regression, support vector machine, decision tree, random forest, k-nearest neighbors, ridge regression, lasso regression, gradient boosting tree, and XGBoost, and the metabolic flexibility dataset was used for model training. All models underwent k-fold cross-validation for performance evaluation. The dataset was randomly divided into k mutually exclusive subsets. Each time, k-1 subsets were selected for training, and the remaining subset was used for testing. This process was repeated k times, ensuring that each subset served as a test set at least once. The average of the k test results was then used as the model's evaluation metric. Grid search was used for hyperparameter tuning. Multiple hyperparameter combinations were preset, and all possible parameter configurations were explored. The optimal parameter combination was selected based on the cross-validation evaluation results. Finally, the top five performing models in cross-validation were selected as candidate models for subsequent model ensemble.

[0077] Furthermore, step S4 includes: as Figure 3 The diagram shown is a flowchart of the metabolic flexibility assessment model construction process of this invention, specifically as follows:

[0078] Metabolic flexibility dataset: The metabolic flexibility dataset (including respiratory quotient derived indicators, questionnaire information and body measurement data) obtained after feature engineering in step S2.

[0079] Input data: The metabolic flexibility dataset is input into various machine learning algorithms (such as logistic regression, support vector machine, decision tree, random forest, K-nearest neighbors, ridge regression, lasso regression, gradient boosting tree, XGBoost, etc.), and each algorithm is independently trained and optimized through cross-validation and grid search;

[0080] Candidate models: Based on the performance of each model in cross-validation, the top 5 best-performing models are selected as candidate models;

[0081] Constructing a secondary feature set: Integrate the prediction results of candidate models 1-5 on the metabolic flexibility dataset to generate a secondary feature set;

[0082] Secondary feature set: This dataset uses the output of the candidate model as its feature vector;

[0083] Model training and tuning: Input the secondary feature set into the logistic regression model, combine cross-validation and grid search to train and tune the model, and finally output the metabolic flexibility assessment model.

[0084] The candidate model is used to predict the data, and the prediction results are used as new features to construct a secondary feature set. The secondary feature set is used as input data to train the logistic regression model, and the hyperparameters of the model are tuned using grid search to finally obtain the metabolic flexibility assessment model.

[0085] This invention further provides a system for assessing metabolic flexibility based on respiratory quotient, such as... Figure 4 As shown, the system includes:

[0086] Data entry module M1: This module is responsible for receiving and integrating raw data from multiple sources, and ensuring that all data conforms to a unified standard format and data quality requirements when input;

[0087] Data storage module M2: This module is responsible for securely storing the entered raw data and provides an interface for the feature extraction module to call the data;

[0088] Feature Engineering Module M3: This module mainly performs preprocessing and feature extraction operations on the stored data, including calculation and normalization of respiratory quotient-derived indicators, to generate metabolic flexibility data suitable for candidate model input;

[0089] Metabolic flexibility assessment module M4: This module integrates candidate models and metabolic flexibility assessment models, takes preprocessed metabolic flexibility data as input, performs model evaluation, and generates metabolic flexibility assessment results;

[0090] Report Generation Module M5: This module generates reports that visually display an individual's metabolic flexibility assessment results and raw data (respiratory quotient, questionnaire information, and body measurement data).

[0091] Model Feedback Update Module M6: Receives all information and feedback entered by doctors, and semi-automatically retrains and updates the candidate model and the evaluation model.

Claims

1. A method for assessing metabolic flexibility based on respiratory quotient, characterized in that, Includes the following steps: Step S1: Collect respiratory quotient, questionnaire information and body measurement data of several individuals with normal, abnormal and impaired metabolic flexibility. This data is used to collect the resting respiratory quotient values ​​of individuals at 30 min, 60 min, 120 min and 180 min after fasting and oral glucose administration, and to obtain questionnaire information and body measurement data at the same time. Step S2: Perform data preprocessing and feature extraction on the data to calculate metabolic recovery degree, metabolic change rate, metabolic peak, peak recovery time, metabolic persistence, metabolic stability index, metabolic response area and metabolic peak width, forming high-dimensional features and constructing a metabolic flexibility dataset. Step S3: Train and fine-tune various machine learning models using the metabolic flexibility dataset to obtain candidate models; Step S4: Use the stacked ensemble method to fuse candidate models and construct a metabolic flexibility assessment model.

2. The method according to claim 1, characterized in that, In step S1, the resting respiratory quotient values ​​at 30 min, 60 min, 120 min, and 180 min after fasting and oral glucose solution administration are recorded as follows: The questionnaire information should include at least age, gender, daily activity intensity, history of diabetes, and family history of diabetes; the body measurement data should include at least height, weight, BMI, waist circumference, hip circumference, waist-to-hip ratio, systolic blood pressure, diastolic blood pressure, body fat mass, body fat percentage, and blood oxygen saturation.

3. The method according to claim 1, characterized in that, Step S2 includes the following steps: Data preprocessing includes standardizing data types, normalizing data, and handling missing values. By integrating and expanding the respiratory quotient index, we can calculate various derived indicators such as metabolic recovery degree, metabolic rate of change, peak metabolic value, peak recovery time, metabolic persistence and metabolic stability index, metabolic response area, and metabolic peak width. Feature extraction was performed using a recursive feature elimination method based on support vector machines, and finally, a metabolic flexibility dataset was constructed.

4. The method according to claim 3, characterized in that, The normalization method is as follows: Data is normalized using the Z-Score standardization method, and the result is calculated using the following formula: ； Where X is the original feature value, μ is the mean of the feature in the dataset, σ is the standard deviation of the feature in the dataset, and the standardized feature value Z is calculated.

5. The method according to claim 3, characterized in that, For numerical features, the missing values ​​are filled using the mean of the feature in the dataset, calculated using the following formula: ； in, Let n represent the missing data for a certain type of sample in the dataset, and n be the total number of samples of that type in the dataset that are not missing. This refers to data from the dataset that includes samples of this type that are not missing.

6. The method according to claim 3, characterized in that, For the respiratory quotient time series data, including missing values ​​at 30 min, 60 min, and 120 min after glucose administration, linear interpolation was used to impute the missing values, and the calculation was performed using the following formula: ； in, and These are the valid time points before and after the missing time point, respectively.

7. The method according to claim 3, characterized in that, Metabolic recovery is defined as the difference between the respiratory quotient at each time point after glucose administration and the fasting respiratory quotient, calculated using the following formula: ； in, This is the baseline value for the fasting respiratory quotient. After taking sugar minute Respiratory quotient at that time; The rate of metabolic change is defined as the rate of change in metabolic recovery between adjacent time points after glucose administration, and is calculated using the following formula: ,in, > ; Peak metabolic rate is defined as the peak of metabolic recovery after glucose administration, and is calculated using the following formula: ； Peak recovery time is defined as the time point at which metabolic recovery reaches its peak, and is calculated using the following formula: ； Where argmax represents finding the input value that makes the function take the maximum value; Metabolic durability is defined as the difference between peak metabolic rate and metabolic recovery at 180°C after glucose administration, calculated using the following formula: ； The metabolic stability index is defined as the ratio of the standard deviation of metabolic recovery at each time point to the mean, and is calculated using the following formula: ； in, For calculation The mean; The metabolic response area is defined as the area under the curve used to calculate the respiratory quotient after a glucose load using the trapezoidal integral method, and is calculated using the following formula: ； The metabolic peak width is defined as the time difference between when the metabolic recovery reaches 80% of the peak metabolic value and when it exits the high-response region, and is calculated using the following formula: ； in, and Determined by linear interpolation, the formula is: ； 。 8. The method according to claim 3, characterized in that, Feature extraction is performed using a recursive feature elimination method based on Support Vector Machine (SVM). Let the original feature set be... The dataset to be filtered, S=X, is initialized. Then, a recursive feature elimination iteration is performed. In each iteration, an SVM model is trained using the current dataset S, and the model is optimized based on the absolute values ​​of the weight vectors. Each feature is assigned an importance score, and the features are sorted according to their importance scores. The k features with the lowest scores are removed, the dataset S is updated, and the SVM model is retrained using the updated dataset S. The model performance is evaluated through cross-validation. The iteration is terminated when the dimension of the dataset S drops to a preset value d, or the model performance decreases by more than a threshold ε, or the model performance improves by more than a threshold λ. Finally, the metabolic flexibility dataset is constructed.

9. The method according to claim 1, characterized in that, We selected various machine learning algorithms, including logistic regression, support vector machine, decision tree, random forest, K-nearest neighbors, ridge regression, lasso regression, gradient boosting tree, and XGBoost, and trained the models using the metabolic flexibility dataset to obtain candidate models.

10. The method according to claim 9, characterized in that, Using a stacked ensemble method, the prediction results of the candidate models are used as a secondary feature set to train the logistic regression model, resulting in a metabolic flexibility assessment model.

11. A metabolic flexibility assessment system based on respiratory quotient, characterized in that, The system includes: Data entry module: This module is responsible for receiving and integrating raw data from multiple sources, and ensuring that all data conforms to a unified standard format and data quality requirements when it is entered; Data storage module: This module is responsible for securely storing the entered raw data and provides an interface for the feature extraction module to call the data; Feature Engineering Module: This module mainly performs preprocessing and feature extraction operations on the stored data, including calculation and normalization of respiratory quotient-derived indicators, to generate metabolic flexibility data suitable for candidate model input; Metabolic flexibility assessment module: This module integrates candidate models and metabolic flexibility assessment models, takes preprocessed metabolic flexibility data as input, performs model evaluation, and generates metabolic flexibility assessment results; Report generation module: This module generates reports that visually display an individual's metabolic flexibility assessment results and raw data, including respiratory quotient, questionnaire information, and body measurement data; Model feedback and update module: Receives all information and feedback entered by doctors, and semi-automatically retrains and updates the candidate model and the evaluation model.

Images