A critical patient weight prediction method and system based on artificial intelligence

Abstract

This invention relates to an artificial intelligence-based method and system for predicting the weight of critically ill patients in the field of clinical medical and nursing technology. The method includes the following steps: S1: Clinical data acquisition; S2: Data preprocessing; S3: Data grouping; dividing the preprocessed clinical data into training and testing sets; S4: Model training; constructing a weight prediction model by fusing n regression algorithms using an ensemble learning algorithm; S5: Patient weight prediction; inputting the clinical data into the weight prediction model to predict the patient's weight. This invention utilizes bioelectrical impedance analysis (BIA) technology to acquire sampled data and effectively extracts and analyzes the data, enabling weight measurement in a supine position, simplifying the measurement process, reducing the cost of weight measurement for critically ill patients, while ensuring high accuracy and robustness.

Description

Technical Field

[0001] This invention relates to the field of clinical medicine, and in particular to a method and system for predicting the weight of critically ill patients based on artificial intelligence. Background Technology

[0002] Weight is one of the most important indicators reflecting and measuring a person's health status. Changes in weight can reveal many health problems, such as obesity, emaciation, and edema, which directly affect an individual's health and quality of life. Many diseases are closely related to weight, such as obesity, diabetes, and cardiovascular disease. Clinicians can use a patient's weight to aid in the diagnosis and monitoring of the progression of these diseases.

[0003] In clinical practice, healthcare professionals typically develop personalized treatment plans based on a patient's weight, including dietary adjustments, exercise programs, and medication. Weight is a crucial monitoring indicator, especially when treating critically ill patients. By observing changes in weight, doctors can promptly detect and assess disease progression, allowing for appropriate treatment interventions. Furthermore, regular weight measurements help evaluate treatment effectiveness; stable or gradually increasing weight may indicate that the treatment plan is effective.

[0004] Currently, the most common method for measuring weight is using a scale. While this method is simple and direct, it presents significant challenges for patients with limited mobility or critical conditions. Especially for those who require prolonged bed rest, using a scale not only demands considerable manpower but also carries a potential risk of falls.

[0005] To overcome these problems, some specialized weight measurement devices, such as weight measuring beds, have emerged on the market. However, these devices are often expensive, with a single unit costing hundreds of thousands of yuan. They also require specialized personnel for operation and regular maintenance, and their operation is quite complex, increasing the operating costs of hospitals and limiting their widespread application.

[0006] In addition, the weighing bed method is inefficient: it is time-consuming, especially for bedridden critically ill patients, often requiring multiple nurses to assist in the measurement, resulting in low efficiency and poor repeatability. During the care of critically ill patients, each weight measurement requires the assistance of 2-3 nurses, taking 10-15 minutes, which not only affects nursing efficiency but also increases the workload of nursing staff.

[0007] For patients with minimal weight change, weighing beds may not provide sufficiently accurate data, thus affecting doctors' assessment of the patient's condition and the development of treatment plans. In some long-term bedridden patients, due to minimal weight change, traditional weighing scales cannot provide accurate measurements, leading to difficulties for doctors in formulating treatment plans.

[0008] Traditional weight measurement methods (such as scales) are not suitable for patients with limited mobility or who are completely bedridden, as these patients face operational difficulties and safety hazards when using scales. For example, using a traditional scale to measure the weight of a patient with a fracture or paralysis poses a significant risk, potentially leading to falls or worsening of their condition. Summary of the Invention

[0009] To address the aforementioned problems, this invention discloses an artificial intelligence-based method and system for predicting the weight of critically ill patients. It utilizes bioelectrical impedance analysis (BIA) technology to acquire sampling data, and effectively extracts and analyzes the data to achieve weight measurement in a supine position, simplifying the measurement process while ensuring high accuracy and robustness. The technical solution of this invention is implemented as follows:

[0010] The first aspect of this invention discloses an artificial intelligence-based method for predicting the weight of critically ill patients, the method comprising the following steps:

[0011] S1: Patient clinical data collection;

[0012] S2: Preprocessing of clinical data;

[0013] S3: Grouping of clinical data; dividing the preprocessed clinical data into training and testing sets;

[0014] S4: Model training;

[0015] An ensemble learning algorithm is used to fuse n regression algorithms to construct a weight prediction model;

[0016] S5: Patient weight prediction;

[0017] The clinical data is input into the weight prediction model to predict the patient's weight.

[0018] Preferably, it also includes S6: treatment guidance; based on the patient's weight prediction results and the differences in changes over time, calculate the patient's drug dosage, nutritional components, fluid management, exercise plan, etc., for clinicians' reference.

[0019] Preferably, the clinical data includes bioelectrical impedance data, gender, height, age, and weight information.

[0020] Preferably, the specific steps of the data preprocessing are as follows:

[0021] Outlier removal, missing value imputation, and normalization were performed on the patients' clinical data to obtain initial clinical data.

[0022] Preferably, the specific process of data preprocessing in step S2 is as follows:

[0023] S2.1 Outlier removal: Using statistical methods or machine learning algorithms to identify and remove outliers in clinical data;

[0024] S2.2, Missing value imputation: Use multiple imputation, mean imputation or regression imputation to handle missing data;

[0025] S2.3, Normalization: Use standardization or normalization to scale data to a uniform range.

[0026] Preferably, in step S4, the weight prediction model includes sampling, multi-model prediction, and ensemble learning;

[0027] Sampling: Select k subsets from the training set of the initial clinical data;

[0028] Multi-model prediction: A base model is established for each subset, and n regression algorithms are used to predict weight for each base model;

[0029] Ensemble learning: The weight prediction results of k base models are combined and the average is taken as the final prediction result.

[0030] Preferably, step S4 further includes model evaluation: extracting data from the test set and inputting it into the weight prediction model, calculating the mean absolute error (MAE) and the coefficient of determination (R²). 2 To measure the model's accuracy and generalization ability.

[0031] Preferably, the regression algorithm includes Bayesian Ridge regression algorithm, Ridge Regression algorithm, and Linear Regression algorithm.

[0032] Preferably, the ensemble learning method includes blending and stacking.

[0033] A second aspect of the present invention discloses an artificial intelligence-based weight prediction system for critically ill patients, the system comprising a data acquisition module, a data preprocessing module, a data grouping module, a weight prediction model, and a treatment guidance module;

[0034] The data acquisition module is used to collect and store the patient's clinical data.

[0035] The data preprocessing module is used to preprocess the patient's clinical data;

[0036] The data grouping module is used to divide the preprocessed clinical data into a training group and a test group;

[0037] The weight prediction module is used to receive clinical data output from the data grouping module and perform weight prediction.

[0038] The treatment guidance module is used to generate treatment guidance information based on the weight prediction results output by the weight prediction module.

[0039] This invention utilizes bioelectrical impedance analysis (BIA) technology to acquire sampling data and effectively extracts and analyzes it, enabling weight measurement in a supine position. This simplifies the measurement process while ensuring high accuracy and robustness. By acquiring daily clinical data from patients and using a constructed predictive model to predict weight, this invention plots a daily weight change curve for patients during their hospital stay. If the weight change exceeds 2 kg within 24 hours, an alert is issued to draw the attention of clinicians and prompt appropriate treatment measures. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only one embodiment of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Identical parts are indicated by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "up," and "down" used in the following description refer to directions in the accompanying drawings, while the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific part, respectively.

[0042] Figure 1 This is a flowchart illustrating the model construction process in an embodiment of the present invention.

[0043] Figure 2 This is a structural diagram of the Blending ensemble learning method in this invention;

[0044] Figure 3 This is a structural diagram of the Stacking ensemble learning method in this invention. Detailed Implementation

[0045] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0046] Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used in the detailed description is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0047] In the description of specific embodiments of the present invention, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present invention, "multiple" means two or more, unless otherwise explicitly defined.

[0048] In this invention, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this invention can be combined with other embodiments.

[0049] In the description of the embodiments of this invention, the term "and / or" is merely a description of the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, in this invention, the character " / " generally indicates that the preceding and following associated objects have an "or" relationship.

[0050] Throughout this invention, numerical values ​​represent approximate measures or limits of a range to cover minute deviations from a given value, as well as embodiments with the mentioned value and embodiments with the exact mentioned value. Except for the working examples provided at the end of the detailed description, all numerical values ​​of parameters (e.g., quantities or conditions) in this specification (including the appended claims) should be understood to be modified in all cases by the term “about,” regardless of whether “about” actually appears before the numerical value. “About” indicates that the stated numerical value allows for some minor inaccuracy (approaching the exact value in some way; approximately or reasonably close to the value; almost). If the inaccuracy provided by “about” is not otherwise understood in this general sense in the art, then “about” as used in this invention at least indicates a variation that can be produced by common methods of measuring and using such a parameter. For example, “about” may include a variation of less than or equal to 5%, optionally less than or equal to 4%, optionally less than or equal to 3%, optionally less than or equal to 2%, optionally less than or equal to 1%, optionally less than or equal to 0.5%, and in some respects, optionally less than or equal to 0.1%.

[0051] Additionally, the disclosure of the range includes the disclosure of all values ​​across the entire range and the disclosure of further subdivided ranges, including the endpoints and subranges given for these ranges.

[0052] Traditional weight measurement methods (such as scales) are not suitable for patients with limited mobility or who are completely bedridden, as these patients face operational difficulties and safety hazards when using scales. For example, using a traditional scale to measure the weight of a patient with a fracture or paralysis poses a significant risk, potentially leading to falls or worsening of their condition.

[0053] To address the aforementioned issues, this invention discloses an artificial intelligence-based method and system for predicting the weight of critically ill patients. It utilizes bioelectrical impedance analysis technology to acquire sampling data and effectively extracts and analyzes the data to achieve weight measurement in a supine position, simplifying the measurement process, reducing measurement costs for patients, and ensuring high accuracy and robustness.

[0054] The first aspect of this invention discloses an artificial intelligence-based method for predicting the weight of critically ill patients, the method comprising the following steps:

[0055] S1: Patient clinical data collection;

[0056] S2: Preprocessing of clinical data;

[0057] S3, Grouping of clinical data; dividing the preprocessed clinical data into training and testing sets;

[0058] S4: Model training;

[0059] An ensemble learning algorithm is used to fuse n regression algorithms to construct a weight prediction model;

[0060] S5: Patient weight prediction;

[0061] The clinical data is input into the weight prediction model to predict the patient's weight.

[0062] This invention uses multiple regression algorithms to predict the clinical data of preprocessed patients, and integrates the prediction results of multiple models through ensemble learning, thereby significantly improving the predictive performance and stability of the weight prediction model. By integrating multiple models, the error of a single model is reduced, and ultimately more robust and accurate prediction results are obtained.

[0063] In some implementations, S6 is also included: treatment guidance; based on the patient's predicted weight and the differences in weight change over time, calculate the patient's medication dosage, nutritional components, fluid management, exercise plan, etc., for clinicians' reference.

[0064] In some implementations, the clinical data includes bioelectrical impedance data, gender, height, age, and weight information.

[0065] In some implementations, the specific steps of the data preprocessing are as follows:

[0066] Outlier removal, missing value imputation, and normalization were performed on the patients' clinical data to obtain initial clinical data.

[0067] The specific process of data preprocessing in step S2 is as follows:

[0068] S2.1 Outlier removal: Using statistical methods or machine learning algorithms to identify and remove outliers in clinical data;

[0069] S2.2, Missing value imputation: Use multiple imputation, mean imputation or regression imputation to handle missing data;

[0070] S2.3, Normalization: Use standardization or normalization to scale data to a uniform range.

[0071] In some implementations, the weight prediction model in step S4 includes sampling, multi-model prediction, and ensemble learning.

[0072] Sampling: k subsets are randomly selected from the training set of the initial clinical data; the number of samples in each subset is the same as the original dataset, and replacement is used in the sampling process, which means that a sample may appear in multiple subsets.

[0073] Multi-model prediction: A base model is established for each subset, and the weight prediction is performed on each base model using the Bayesian Ridge Regression algorithm, the Ridge Regression algorithm, and the Linear Regression algorithm respectively;

[0074] Suppose we have a dataset where the training set D = {(x1,y1),(x2,y2),...,(x...} n ,y n )},x i y is the feature vector of the i-th sample. i Here, represents the target value, and n is the size of the training set. The following sections describe the model construction for the three regression algorithms.

[0075] Bayesian Ridge Regression:

[0076] Bayesian Ridge Regression is a linear regression method based on Bayesian statistics. It utilizes the ideas of ridge regression and Bayesian inference, strengthening the constraint on the regression coefficients by introducing a prior distribution. The Bayesian Ridge Regression model assumes that the target variable follows a Gaussian distribution y. i ~N(W) T x i ,σ 2 ), where W is the regression coefficient, σ 2 This represents the variance of the Gaussian noise. A prior distribution p(W) is introduced to constrain the values ​​of the regression coefficients. Bayesian inference is used to obtain the posterior distribution p(W|D), which is the distribution of the regression coefficients given a dataset D. According to Bayes' theorem, the posterior distribution can be obtained by multiplying the prior distribution and the likelihood function:

[0077]

[0078] Use maximum a posteriori (MAP) estimation to obtain point estimates of the regression coefficients:

[0079]

[0080] Bayesian Ridge regression estimates the regression coefficients by minimizing the loss function.

[0081]

[0082] Where, λ br It is a regularization parameter.

[0083] Ridge Regression:

[0084] Ridge regression controls model complexity by adding an L2 regularization term to the loss function. The L2 regularization term penalizes large coefficient values ​​based on the sum of squares of the model's coefficient vector, effectively reducing the risk of overfitting. This regularization term can be written as:

[0085]

[0086] Where, λ r These are the defined regularization parameters, where m is the number of features and β is the regularization parameter. j is the coefficient of the j-th feature. The goal of ridge regression is to minimize the following loss function:

[0087]

[0088] Here These are predicted values. By adding a regularization term, ridge regression improves generalization ability by balancing the model's fit and stability.

[0089] Linear Regression:

[0090] The goal of linear regression is to find a straight line that minimizes the error between the predicted and observed values. Mathematically, a linear regression model can be expressed as:

[0091]

[0092] in, The target variable is α0, which is the intercept, and x is the target variable for prediction. i It is the feature vector of the i-th sample, α i These are the weights of the feature variables.

[0093] To find the best-fit line, we used the least squares method to minimize the squared error between the actual observations and the predicted values, i.e., the following cost function:

[0094]

[0095] Ensemble learning: The weight prediction results of k base models are combined and the average is taken as the final prediction result.

[0096] To overcome the limitations of single models in generalization ability, this invention utilizes ensemble learning to construct and combine multiple base learners, thereby improving the model's predictive performance and stability. The core idea of ​​ensemble learning is to reduce the error of individual models by integrating multiple models, ultimately achieving more robust and accurate prediction results. This paper employs blending and stacking as the ensemble learning model.

[0097] Blending is a model fusion method. The main idea of ​​typical blending is to divide the original training set into two parts, for example, 60% of the data as a new training set and the remaining 40% as a test set. In the first layer, we train multiple models on the 60% of data and then predict the labels of the remaining 40% of the data. In the second layer, we directly use the prediction results from the first layer on the remaining 40% of data as new features to continue training.

[0098] Stacking combines two ensemble methods: Boosting and Bagging. It uses multiple base learners to learn the original data, and then feeds the data learned by these base learners into the second-layer model for fitting.

[0099] Although these two methods differ in implementation, they both effectively improve generalization ability and prediction accuracy through model ensemble.

[0100] In some implementations, step S4 further includes model evaluation: extracting data from the test set and inputting it into the weight prediction model, calculating the mean absolute error (MAE) and the coefficient of determination (R²). 2 To measure the model's accuracy and generalization ability.

[0101] A second aspect of the present invention discloses an artificial intelligence-based weight prediction system for critically ill patients, the system comprising a data acquisition module, a data preprocessing module, a data grouping module, a weight prediction model, and a treatment guidance module;

[0102] The data acquisition module is used to collect and store the patient's clinical data.

[0103] The data preprocessing module is used to preprocess the patient's clinical data;

[0104] The data grouping module is used to divide the preprocessed clinical data into a training group and a test group;

[0105] The weight prediction module is used to receive clinical data output from the data grouping module and perform weight prediction.

[0106] The treatment guidance module is used to generate treatment guidance information based on the weight prediction results output by the weight prediction module.

[0107] The embodiments of the present invention will be described in more detail below through examples and comparative examples. It should be noted that the embodiments of the present invention are not limited to these examples.

[0108] Example

[0109] In one specific embodiment, such as Figure 1As shown, an artificial intelligence-based weight prediction model for critically ill patients is constructed. The model classifies preprocessed clinical data into training and testing sets, and trains the model on the training set using Bayesian Ridge Regression, Ridge Regression, and Linear Regression algorithms, respectively. The trained model is then tested using the testing set.

[0110] The specific testing method is as follows:

[0111] First, using 34 features, regression models were constructed based on Bayesian ridge, RidgeRegression, and LinearRegression algorithms for 1000 test subjects. The results are shown in the table below:

[0112]

[0113] Finally, the three regression models were integrated using both blending and stacking ensemble learning methods. The flow structure of the blending ensemble learning method is as follows: Figure 2 As shown, the process structure of the Stacking ensemble learning method is as follows: Figure 3 As shown.

[0114] The results of the two ensemble learning methods are shown in the table below. In this embodiment, the blending ensemble learning model performed best, with MAE = 1.72 and R0 in the training set. 2 =0.97; in the test set: MAE=1.85, R 2 =0.96.

[0115]

[0116] In this embodiment, since the blending ensemble learning model performs best, its prediction result is presented as the final result in weight prediction. Of course, in other embodiments, if the stacking ensemble learning model performs better, its prediction result is used as the final prediction result.

[0117] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the weight of critically ill patients based on artificial intelligence, characterized in that, The steps include the following: S1: Patient clinical data collection; S2: Preprocessing of clinical data; S3: Grouping of clinical data; The preprocessed clinical data were divided into a training set and a test set; S4: Model training; An ensemble learning algorithm is used to fuse n regression algorithms to construct a weight prediction model; S5: Patient weight prediction; The clinical data is input into the weight prediction model to predict the patient's weight.

2. The method according to claim 1, characterized in that, It also includes S6: Treatment guidance; based on the patient's predicted weight and the differences in weight change over time, calculate the patient's medication dosage, nutritional components, fluid management, and exercise plan for clinicians' reference.

3. The method according to claim 1, characterized in that, In step S1, the clinical data includes bioelectrical impedance data, gender, height, age, and weight information.

4. The method according to claim 1, characterized in that, In step S2, the specific steps of data preprocessing are as follows: Outlier removal, missing value imputation, and normalization were performed on the patients' clinical data to obtain initial clinical data.

5. The method according to claim 4, characterized in that, The specific process of data preprocessing in step S2 is as follows: S2.1 Outlier removal: Using statistical methods or machine learning algorithms to identify and remove outliers in clinical data; S2.2, Missing value imputation: Use multiple imputation, mean imputation or regression imputation to handle missing data; S2.3, Normalization: Use standardization or normalization to scale data to a uniform range.

6. The method according to claim 5, characterized in that, In step S4, the weight prediction model includes sampling, multi-model prediction, and ensemble learning. Sampling: Select k subsets from the training set of the initial clinical data; Multi-model prediction: A base model is established for each subset, and n regression algorithms are used to predict weight for each base model; Ensemble learning: The weight prediction results of k base models are combined and the average is taken as the final prediction result.

7. The method according to claim 6, characterized in that, Step S4 also includes model evaluation: extracting data from the test set and inputting it into the weight prediction model, calculating the mean absolute error (MAE) and the coefficient of determination (R²). 2 To measure the accuracy and generalization ability of the model.

8. The method according to claim 6, characterized in that, The regression algorithms include Bayesian Ridge Regression, Ridge Regression, and Linear Regression.

9. The method according to claim 6, characterized in that, The ensemble learning methods include Bagging and Boosting.

10. A weight prediction system for critically ill patients based on artificial intelligence, characterized in that, It includes a data acquisition module, a data preprocessing module, a data grouping module, a weight prediction model, and a treatment guidance module; The data acquisition module is used to collect and store the patient's clinical data. The data preprocessing module is used to preprocess the patient's clinical data; The data grouping module is used to divide the preprocessed clinical data into a training group and a test group; The weight prediction module is used to receive clinical data output from the data grouping module and perform weight prediction. The treatment guidance module is used to generate treatment guidance information based on the weight prediction results output by the weight prediction module.

Images