CN113782183B - Device and method for predicting risk of pressure injury based on multi-algorithm fusion - Google Patents
Device and method for predicting risk of pressure injury based on multi-algorithm fusion Download PDFInfo
- Publication number
- CN113782183B CN113782183B CN202110869608.8A CN202110869608A CN113782183B CN 113782183 B CN113782183 B CN 113782183B CN 202110869608 A CN202110869608 A CN 202110869608A CN 113782183 B CN113782183 B CN 113782183B
- Authority
- CN
- China
- Prior art keywords
- risk
- type
- variables
- training set
- pressure
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Software Systems (AREA)
- Pathology (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Biology (AREA)
- Epidemiology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a device and a method for predicting risk of pressure injury based on multi-algorithm fusion, wherein the device comprises a processing unit, and the processing unit is configured to execute the following steps: obtaining the analyzable medical record data and reassigning the analyzable medical record data based on the significant risk variables, thereby generating a first classification training set for the specific target population; performing regression modeling on the first classification training set by using a random forest model, so as to generate a first pressure damage risk prediction model related to the first classification training set; classifying the first pressure damage risk prediction model to obtain a second class risk variable and a second weight which characterize the characteristics of the first pressure damage risk prediction model, and combining a plurality of first pressure damage risk prediction models based on the second class risk variable to generate a second pressure damage risk prediction model; and carrying out pressure injury risk prediction by using a second pressure injury risk prediction model.
Description
Technical Field
The invention relates to the technical field of medical data processing, in particular to a device and a method for predicting risk of pressure injury based on multi-algorithm fusion.
Background
Pressure Injury (PI) is a localized Injury that occurs to the skin or underlying subcutaneous soft tissue, typically at the apophyseal or in contact with medical instrument devices. Stress injuries can have adverse effects on the patient's mind and body, and can also increase patient hospitalization time, complication rate, and mortality.
At present, research on stress injury mainly focuses on the development mechanism of the injury, the analysis of the characteristics of the injury, the research on the characteristics of an injured patient and nursing measures, and most of researches are statistical analysis and objective description on historical medical records and lack of prediction research on the injury. Risk prediction is a primary measure for preventing pressure injury, and whether the risk prediction result is accurate or not directly influences the selection and prevention effect of the prevention measure.
At present, a pressure injury risk prediction model is constructed by utilizing multi-factor regression analysis in clinical medicine to predict the pressure injury risk. For example, document [1] Liqing, su Jiang, lin Ying, etc.. Inpatient compressive injury analysis and prediction based on machine learning [ J ]. University of same aid university journal (Nature science edition), 2020 (10). A predictive model is disclosed using 3 methods of support vector machine in which a Gaussian kernel is used to build a model and genetic algorithm is used to optimize kernel parameters. However, the solution provided by this document does not take into account the complexity of the clinical patient situation, and the risk prediction model needs to have the ability to expand to incorporate new risk variables.
Furthermore, there are differences in one aspect due to understanding to those skilled in the art; on the other hand, as the inventors studied numerous documents and patents while the present invention was made, the text is not limited to details and contents of all that are listed, but it is by no means the present invention does not have these prior art features, the present invention has all the prior art features, and the applicant remains in the background art to which the rights of the related prior art are added.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a pressure damage risk prediction device based on multi-algorithm fusion, which comprises a processing unit. The processing unit is configured to perform the steps of:
obtaining the analyzable medical record data and reassigning the analyzable medical record data based on the significant risk variables, thereby generating a first classification training set for the specific target population;
performing regression modeling on the first classification training set by using a random forest model, so as to generate a first pressure damage risk prediction model related to the first classification training set;
classifying the first pressure damage risk prediction model to obtain a second class risk variable and a second weight which characterize the characteristics of the first pressure damage risk prediction model, and combining a plurality of first pressure damage risk prediction models based on the second class risk variable to generate a second pressure damage risk prediction model;
And carrying out pressure injury risk prediction by using a second pressure injury risk prediction model. Aiming at the risk prediction of the pressure injury, the risk prediction is generally carried out through an established pressure injury risk prediction model, and aiming at the risk prediction model, the comprehensive risk variable needs to be incorporated as far as possible, so that the risk variable affecting the pressure injury can be avoided, and the accuracy of the pressure injury risk prediction is further improved. However, the target population is different, the risk variable acting on the pressure injury is different, and the effective degree of the risk variable acting on the pressure injury risk prediction is also different, so that the influence of the specific risk variable of other target populations can be overcome when the pressure injury risk prediction model is used for predicting the pressure injury risk of the specific target population. The method solves the problems by adopting a multi-algorithm fusion mode, establishes a pressure damage risk prediction model by utilizing a random forest model and a multiple logistic regression model, overcomes the influence caused by irrelevant risk variables by the first pressure risk prediction model aiming at specific target groups, and on the other hand, related medical record data of a patient can be composite, namely, two or more first pressure damage risk prediction models can be matched in the medical record data, so that the first risk prediction models need to ensure the combinability, namely, the two or more first pressure damage risk prediction models need to be combined and the expansion capability of new risk variables is brought into. Since the generated first pressure injury risk prediction model needs to have the capability of incorporating new risk variables and combining a plurality of first pressure injury risk prediction models with each other, the expanded or combined first pressure injury risk prediction model needs to ensure the stability of prediction. However, the first stress risk prediction model is constructed according to a random forest model, so if a new risk variable is included and the data size is large, the output of the first stress injury risk prediction model may be inclined to the side with more data size/data record, and thus the prediction result deflection can be avoided by averaging the data sizes of the second type risk variables in the second classification training set. In addition, if more associated risk variables exist in the second type of risk variables, the output of the first pressure damage risk prediction model is inclined to one side of the more associated risk variables, so that a plurality of third type of risk variables are obtained through association degree division, and the plurality of third type of risk variables contain the same number of second type of risk variables, so that the classification number of the risk variables is balanced, and inclination of a risk prediction result can be avoided.
According to a preferred embodiment, the processing unit is configured to:
classifying medical record data in a first classification training set by utilizing a random forest model so as to acquire first type risk variables related to the first classification training set;
regression is carried out on the first classification training set and first type risk variables corresponding to the first classification training set based on a random forest model so as to obtain first weights representing correlations among a plurality of first type risk variables;
dividing the first classification training set based on the first weight to form a plurality of second classification training sets, and modeling the plurality of second classification training sets by adopting a random forest model to generate a plurality of first pressure injury risk prediction models.
According to a preferred embodiment, the processing unit is configured to:
establishing a multiple logistic regression model by taking the first type of risk variables as independent variables and taking whether the correlation among the first type of risk variables is a dependent variable or not;
acquiring the association degree among a plurality of first-type risk variables based on a multiple logistic regression model;
the first classification training set is partitioned based on the degree of association to generate a second classification training set.
According to a preferred embodiment, the processing unit is configured to:
Constructing a correlation table based on the degree of correlation between each first type of risk variables;
acquiring a first class risk variable pair with a first weight smaller than a first threshold;
the number of first type risk variables included in the first type risk variable pair is calculated based on the correlation table.
According to a preferred embodiment, the processing unit is configured to:
if the number of the same first type risk variables exceeds a second threshold, searching a first type risk variable pair of which the next first weight is smaller than the second threshold;
if the number of the same first type risk variables is smaller than or equal to a second threshold value, selecting the first type risk variable with the least number of other first type risk variables as the isolated first type risk variable.
According to a preferred embodiment, the processing unit is configured to:
and classifying the first pressure damage risk prediction model to obtain a second class of risk variables and a second weight representing model characteristics of the first class of risk variables. The second weight represents the degree of association of the second type of risk variable with occurrence of the pressure injury in the first pressure injury risk prediction model.
According to a preferred embodiment, the processing unit is configured to:
Based on the coefficient of the kene as a splitting or competing rule of the random forest model, a second type risk variable and a second weight of the first pressure injury risk prediction model are obtained. The second weight is a coefficient of kunity.
According to a preferred embodiment, the processing unit is configured to:
averaging the data volume of the second class risk variables in the second class training set;
dividing the second type of risk variables based on the degree of association, thereby generating a plurality of third type of risk variables;
modeling based on a plurality of third class risk variables to generate a second pressure damage risk prediction model.
The invention also provides a method for predicting risk of pressure injury, which comprises the following steps:
obtaining the analyzable medical record data and reassigning the analyzable medical record data based on the significant risk variables, thereby generating a first classification training set for the specific target population;
performing regression modeling on the first classification training set by using a random forest model, so as to generate a first pressure damage risk prediction model related to the first classification training set;
classifying the first pressure damage risk prediction model to obtain a second class risk variable and a second weight which characterize the characteristics of the first pressure damage risk prediction model, and combining a plurality of first pressure damage risk prediction models based on the second class risk variable to generate a second pressure damage risk prediction model;
And carrying out pressure injury risk prediction by using a second pressure injury risk prediction model.
According to a preferred embodiment, the method further comprises:
classifying medical record data in a first classification training set by utilizing a random forest model so as to acquire first type risk variables related to the first classification training set;
regression is carried out on the first classification training set and first type risk variables corresponding to the first classification training set based on a random forest model so as to obtain first weights representing correlations among a plurality of first type risk variables;
dividing the first classification training set based on the first weight to form a plurality of second classification training sets, and modeling the plurality of second classification training sets by adopting a random forest model to generate a plurality of first pressure injury risk prediction models.
Drawings
FIG. 1 is a schematic flow chart of the steps of a preferred embodiment of the method of the present invention;
fig. 2 is a schematic block diagram of a preferred embodiment of the apparatus of the present invention.
List of reference numerals
100: processing unit 200: the storage unit 300: communication unit
Detailed Description
The following detailed description refers to the accompanying drawings.
Preferably, the invention provides a method for predicting risk of pressure injury, in particular to a method for predicting risk of pressure injury by a pressure injury risk prediction model.
Preferably, the risk prediction model is a tool that predicts the absolute probability of an individual's occurrence or likelihood of occurrence of a disease based on multiple etiologies, through multi-factor analysis. The pressure injury risk prediction model aims at accurately predicting the risk of occurrence of pressure injury, and is convenient for medical staff to take targeted measures in time. The pressure injury according to the present invention refers to a medical institution-acquired pressure injury.
Preferably, referring to fig. 1, the steps of establishing a pressure damage risk prediction model are described.
S100: and screening the medical record data to obtain analyzable medical record data. Preferably, medical record data of the external institution may be acquired through a network. The external institution may be a hospital, a disease center, or a related institution storing patient medical records. The network may be a local area network, the internet, a mobile network, etc. Because of the large number of database medical records of the external institutions and the different specific situations, the precondition that the retrospective analysis of the pressure injury is required is that the patient has not suffered pressure injury or that the patient has not suffered pressure injury in a short time. It is therefore necessary to process externally accessed medical record data to screen out data that cannot be retrospectively analyzed. The step of screening medical record data is as follows.
S101: medical record data is retrieved, excluding medical record data for pressure injuries and skin type injuries occurring at the time of admission.
S102: medical record data of pressure injury occurring within a first time threshold after admission is excluded.
Preferably, the medical record data of which pressure damage occurs at the time of admission is excluded from the medical record data, so that data of which pressure damage does not occur at the time of admission can be obtained. Preferably, the medical history of skin type lesions includes burn type medical history, psoriasis, lupus erythematosus, and the like.
Preferably, the first time threshold may be set as desired, e.g., 24 hours, 10 days, 20 days, etc. To ensure the validity of medical record data for learning training, time-dependent factors need to be considered. For example, medical records are needed that exclude patients from pressure injury within 24 hours after admission. Because the occurrence of the stress injury in a short period of time after admission is likely to be related to the relevant factors when not admitted, the risk prediction model of the stress injury is incorrect.
Preferably, the characters in the analyzable medical record data are digitized. And carrying out dimension normalization processing on the digitized analyzable medical record data. Preferably, since the characterization of patient information in medical record data may not be numerical, it is necessary to convert such information into numerical values that can be identified by the model. For example, a 2-ary, 8-ary, or other multi-ary representation may be employed. Patient information includes risk variables for pressure injury. For example, a feeding condition may be taken as 0 for poor feeding and 1 for normal feeding. Incontinence conditions may be taken to mean 1 for complete control, 2 for occasional incontinence, 3 for fecal/urinary incontinence, and 4 for fecal/urinary incontinence. Skin type may be 1 for normal, 2 for thin, 3 for dry, 4 for edema, 5 for wetness, 6 for color difference, 7 for cracking, etc.
Preferably, part of the physiological index may be processed using an international organization for the conversion factor. For example, conversion of creatinine to micromoles per liter requires multiplication by 88.4. For example, converting glucose to millimoles per liter requires multiplication by 0.0555. Preferably, the dimension normalization process includes normalizing all variables to a range of 0 to 10. The normalization process can be to subtract the minimum value of the variable in the medical record data from the current value and then divide by the difference between the maximum and minimum values of the variable, and then scale up the value by a factor of 10. Through this setting method, the beneficial effect who reaches is:
in the prior art, the data is usually normalized to be within 0-1 by adopting a multivariate classification algorithm such as a random forest model, a multivariate logistic regression model, a support vector machine algorithm and the like, but in the setting mode, more decimal numbers are generated during subsequent computer calculation, and a large amount of floating point operation is needed by a computer, so that a large amount of calculation cost is consumed. In addition, the invention normalizes the data to the range of 0-10, reduces the generated decimal, and further reduces the calculated amount of floating point operation, thereby saving the calculation cost.
S200: the analyzable medical record data is randomly partitioned into at least one training set for establishing a risk prediction model of the pressure injury and at least one validation set for validating the risk prediction model of the pressure injury. The at least one training set may be one training set, two training sets, three training sets, or more. The at least one validation set may be one validation set, two validation sets, three validation sets, or more. The manner in which the analyzable medical record data is randomly divided may be to divide the analyzable medical record data into two portions uniformly at random. Preferably, the manner of randomly dividing the analyzable medical record data may also be unevenly divided, for example, randomly dividing the analyzable medical record data into ten parts, wherein nine parts are used for establishing the risk prediction model of the pressure injury, and the remaining one part is used for respectively verifying the risk prediction model of the pressure injury established by the nine training sets.
Preferably, a plurality of risk variables is obtained based on the training set.
Preferably, the risk variable refers to a relevant factor affecting the occurrence of the pressure injury. The risk variables include at least age, sex, weight, length of stay in hospital, department of stay in hospital, type of illness, type of surgery, length of surgery, vital signs, pulse, blood oxygen saturation, hemoglobin, serum protein, blood gas analysis related indicators, type of respiration (whether mechanical ventilation is or not), presence or absence of coherent complications (such as diabetes, infection), smoking conditions, medication conditions, feeding conditions, excretion conditions, and the like. Preferably, the aforementioned exemplified risk variables may be further subdivided, for example, disease types may be classified into cardiovascular disease, peripheral vascular disease, pneumonia, influenza, etc.
Preferably, the key to constructing the stress injury risk prediction model is the screening of independent variables, i.e. risk variables. Because the invention adopts multi-factor regression analysis to construct the pressure damage risk prediction model, the essence is to quantify the relation between the independent variable and the dependent variable, and thus the independent variable which can cause the variation of the dependent variable needs to be comprehensively obtained. In this embodiment, the argument refers to a risk variable. Dependent variable refers to the occurrence of pressure injury.
Preferably, the plurality of risk variables obtained are screened.
Preferably, multiple risk variables can be screened in a single-before-multiple manner, i.e., single-factor analysis is performed first, and single-factor meaningful risk variables are then incorporated together into a multi-factor analysis model. However, in some cases, there are limitations in adopting a single-before-multiple risk variable screening method, for example, when the number of risk variables is excessive, for example, when there is a collinearity between the risk variables, for example, when the number of missing values is large and samples containing the missing values are not discarded. The screening can be performed by the method adopted correspondingly for the different situations. For example, regularization techniques may be employed to address the co-linearity problem. Regularization techniques include ridge regression methods, LASSO regression models, elastic network models, and the like. For example, the problem of more missing values can be solved by adopting a random forest model. In addition, a cluster analysis method, a principal component analysis method, a stepwise regression method, a gradient lifting method, and the like may also be employed to screen a plurality of risk variables.
S300: and establishing a pressure damage risk prediction model by using a regression algorithm, and incorporating the screened risk variable into the constructed pressure damage risk prediction model.
Preferably, the regression class algorithm model may employ a parameterized model, a semi-parameterized model, or a non-parameterized model. Preferably, the parameterized model may be a generally linear model or a generalized linear model. The general linear model may be a linear regression algorithm model. The generalized linear model may be a Logistic regression model or a poisson regression model. The semi-parameterized model may be a Cox proportional hazards model or a competitive hazards model. The non-parameterized model may be a machine learning class algorithm such as a KNN nearest neighbor algorithm, an SVM support vector machine, a classification regression tree, a random forest, a neural network, or deep learning.
Preferably, the risk variable is an argument of the pressure damage risk prediction model. The dependent variable of the risk prediction model for pressure damage is occurrence of pressure damage. Illustrated by a multiple logical (Logistic) regression model.
Preferably, the relationship between the dependent variable and the independent variable can be expressed as the following formula:
y＝1/(1+e -z ) (1)
z＝p 0 +β 1 X 1 +…+β n X n (2)
wherein y represents a dependent variable. y has a value of 0 or 1.0 indicates that no compressive damage has occurred, and 1 indicates that compressive damage has occurred. X is X n Representing the argument, i.e., the risk variable. n represents the number of independent variables (risk variables). Beta n Representing regression coefficients. The regression coefficients characterize the degree of influence of the independent variables on the dependent variables, i.e., the regression coefficients are used to characterize the degree of contribution of the corresponding risk variables to the occurrence of pressure damage. Or the regression coefficients may also be weights that characterize the degree to which the risk variable causes the dependent variable to change.
Preferably, the logarithm of formula (1) is taken to obtain:
Ln(y/1-y)＝β 0 +β 1 X 1 +…+β n X n (3)
wherein Ln (y/1-y) is a logical (Logistic) transformation. In the formula (3), y may represent a probability that y takes a value of 1. 1-y may represent the probability that y takes a value of 0.
Preferably, the probability of y having a value of 1 is p (y=1) =e z /(1+e z )。
Preferably, the probability of the value of y being 0 is p (y=0) =1/(1+e) z )。
Preferably, the values of the regression coefficients can be estimated based on medical record data within the training set according to a maximum likelihood method.
S400: and verifying the established pressure damage risk prediction model based on the verification set. Preferably, the compressive injury risk prediction model is updated according to the predicted performance and consistency. Preferably, the predictive performance of the pressure injury risk prediction model may be characterized by an index of sensitivity, specificity, area under receiver operating characteristic curve (ROC) (AUC), and the like. Sensitivity is used to characterize the ability of the risk prediction model to screen truly ill patients. Specifically characterizing the ability of the risk prediction model to exclude truly ill patients. The area under the receiver operating characteristic curve (ROC) (AUC) is generally 0.5 to 1, which is an index for evaluating the discrimination ability of the risk prediction model. The larger the AUC value, the higher the authenticity. Preferably, the AUC can be interpreted by a confusion matrix specification. Confusion matrices include Positive (Positive) and Negative (Negative). The prediction is True (True). The prediction error is False. The confusion matrix includes true positive, false positive, true negative, and false negative, as shown in table 1.
TABLE 1 confusion matrix
True yang may be represented by TP. The number of true positive samples indicates the number of people with diseases classified as truly ill patients, i.e., the actual value is 1, and the predicted value is also 1.
The false positive may be represented by FP. The number of samples of the cynomorium songaricum indicates the number of persons classified as ill by healthy patients, the actual value is 0, and the predicted value is 1.
The true yin can be expressed as TN. The number of true negative samples indicates the number of persons that healthy patients were classified as disease-free, and both the actual value and the predicted value are 0.
The pseudoyin may be represented by FN. The number of samples of the prosthesis indicates the number of patients with actual illness classified as no illness, the actual value is 1, and the predicted value is 0.
Sensitivity can be expressed in terms of true positive probability. True positive probability is used to represent the probability that a patient suffering from a disease is classified as ill, and sensitivity can be characterized by the following formula.
Specificity can be expressed in terms of true negative probabilities. True negative probabilities are used to represent the probability that a healthy patient is classified as disease-free, and specificity can be characterized by the following formula.
AUC represents the area under the receiver operating characteristic curve (ROC). The vertical axis of the ROC curve is sensitivity S E . The horizontal axis of the ROC curve is 1-S P I.e. the probability of false positive. The function of the ROC curve is characterized as S E ＝F(1-S P ). F (-) represents a function. AUC is curve S E ＝F(1-S P ) At S of E And 1-S P Area within the enclosed rectangular frame. AUC 1 indicates the most ideal case, meaning that neither truly ill patients are misclassified as ill-free nor healthy patients are misclassified as ill, i.e. AUC can be used to characterize the discriminatory power of the risk prediction model for stress injury.
Preferably, the consistency can be evaluated by Goodness Of Fit (GOF). When the P value of the risk prediction model is larger than 0.05, the risk prediction model is indicated to fully extract information in the data, and the fitting goodness is higher. The P value represents: the probability of the present situation or worse occurs when the original assumption is correct. Preferably, the accuracy of the risk prediction model is typically a calibration curve. The calibration curve is a scatter plot with the predicted occurrence probability as the abscissa and the actual occurrence probability as the ordinate. And (3) performing straight line fitting on the scatter diagram, and if the straight line is a straight line with a slope of 45 degrees passing through the origin, obtaining a better model accuracy. The farther the slope 45 deg. line from the origin is, the worse the prediction accuracy of the model is. In Logistic regression analysis, the calibration curve is actually a visualization of the outcome of the goodness-of-fit evaluation.
Example 1
The embodiment provides a method for predicting risk of pressure injury, which predicts risk of pressure injury through a pressure injury risk prediction model. The present embodiment improves on the compressive injury risk prediction model established in steps S100 to S400.
Preferably, the risk prediction of the pressure injury is performed based on a multiple logic (Logistic) regression model or other regression models, on one hand, the problem of multiple collinearity possibly existing in the risk variables needs to be considered, and on the other hand, whether the established risk prediction model of the pressure injury can be suitable for different types of people, namely, whether the risk prediction model of the pressure injury has the problem of application limitation or not needs to be considered. In particular, in one aspect, multiple collinearity means that there is some correlation between multiple risk variables, in particular a certain risk variable may be characterized by a combination of other risk variables. For example, there is a correlation between systolic and diastolic blood pressure, total cholesterol, and low density lipoprotein cholesterol, and similar risk variables. If the risk variables with multiple collinearity are included and fitted in the regression model, the regression coefficient estimated by the multiple Logistic regression equation may not be identical to the common sense, or even the sign of the regression coefficient is opposite to the actual sign, which may seriously affect the result of risk prediction of the pressure damage. On the other hand, the problem that the pressure injury risk prediction model is limited is caused by different target groups, different medical record database providing data, different screened risk variables and the like. Specifically, the comprehensive risk variables are included as much as possible, so that the risk variables affecting the pressure damage can be avoided from being omitted, and the accuracy of pressure damage risk prediction is improved. However, the target population is different, the risk variable acting on the pressure injury is different, and the effective degree of the risk variable acting on the pressure injury risk prediction is also different, so that the influence of the specific risk variable of other target populations can be overcome when the pressure injury risk prediction model is used for predicting the pressure injury risk of the specific target population.
For ease of understanding, vascular disease patients and orthopedic surgery patients are illustrated. If the vascular disease patient and the orthopedic operation patient are all included in the training set, the risk variables at least comprise the specific risk variable of the vascular disease patient and the specific risk variable of the orthopedic operation patient, such as intake of the vascular disease patient for treating vascular drugs, pressure, operation duration, operation position and the like caused by the specific operation instrument of the orthopedic operation patient, so that compared with the simple vascular disease patient, the risk variables such as the included operation instrument, operation duration, operation position and the like are irrelevant risk variables. If the medical record data of the two types are mixed to be used as a training set for establishing the pressure injury risk prediction model, the pressure injury risk prediction model also considers the contribution of the irrelevant risk variable as the dependent variable, so that regression coefficients are distributed to the irrelevant risk variable, and the irrelevant risk variable is equivalent to an interference factor relative to a vascular disease patient, so that the pressure injury risk prediction model is inaccurate in prediction. In addition, in practical application, the case that the medical record data of the vascular disease patient is more likely to occur is that the medical record data of the vascular disease patient has no data about the surgical instrument, the surgical time and the surgical position, and the case that the medical record data of the patient is missing for the pressure injury risk prediction model further affects the prediction performance of the pressure injury risk prediction model.
Preferably, the difference of the present embodiment from the aforementioned steps S100 to S400 includes the following steps.
Preferably, the at least one training set and the at least one validation set acquired in step S200 can be randomly partitioned into the analyzable medical record data, typically in a random manner. By the arrangement mode, medical record data of different target groups are mixed in the training set or the verification set, so that the training of the obvious risk variable of a specific target group is required during training.
S201: and reassigning the randomly divided training set based on the significant risk variable, thereby obtaining a first classified training set aiming at the specific target crowd. Preferably, the number of the first classification training sets of the present embodiment may be one, two, three or more. Specifically, the first classification training set according to this embodiment may be in one-to-one correspondence with a specific target crowd. For example, a vascular disease patient corresponds to a first classified training set. For example, a bone surgery patient corresponds to a first classified training set.
Preferably, chi-square test may be used to analyze each risk variable and find significant risk variables that have significant impact on pressure damage. Preferably, the significant risk variable according to the present embodiment refers to a risk variable having a significant effect on stress injuries, such as departments, BMIs, skin types, incontinence, perception limitation, etc. Preferably, medical record data in the training set having the same significant risk variable is reassigned to the same first classified training set. For example, the significant risk variable may be of the type of disease the patient suffers from, such as the patient may be classified as vascular disease patient, cardiovascular surgery patient, orthopedic surgery patient, ICU patient, etc., i.e. the training set may be classified into a first classified training set for vascular disease patient, a first classified training set for cardiovascular surgery patient, a first classified training set for orthopedic surgery patient, a first classified training set for ICU patient, etc., based on the significant risk variable.
Preferably, in step S200, risk variables of medical record data in the training set need to be screened. Aiming at the multiple co-linearity problem, the related coefficients among the risk variables can be calculated, and then the risk variable with high related coefficient is removed, but the method is only used for the situation that two repeated risk variables can be approximately considered when the related coefficient is high, and if the related coefficient is low or the situation that the related coefficient cannot be considered as two repeated risk variables, one of the risk variables is removed, so that the information contained in the risk variable can not be fully utilized.
Preferably, regularization techniques may also be employed to solve the multiple collinearity problem. Specifically, in step S300, the maximum likelihood method is used to estimate the values of the regression coefficients based on the medical record data within the training set. The regularization technique is to estimate the value of the regression coefficient by using a maximum a posteriori estimation method. Preferably, the maximum a posteriori estimation method is used to estimate the values of the regression coefficients based on the medical record data within the training set. The maximum a posteriori estimate may be regarded as a regularized maximum likelihood estimate. Maximum likelihood estimation considers the parameter to be estimated (regression coefficient) to be a constant number, but the value of the parameter is not known at present, and the parameter can be estimated by randomly generated samples. The maximum posterior estimation considers the parameter to be estimated as a variable, the variable obeys a certain random distribution model, namely the parameter is considered as an unknown random variable, the prior probability of the parameter distribution condition can be given, and then the parameter is estimated based on the Bayesian theorem. However, although the generalization capability can be increased by a priori estimation based on the maximum a posteriori estimation, the multiple logistic regression model is very sensitive to multiple collinearity, and the modeling is based on the premise that risk variables are assumed to be independent of each other, and in the process of screening the risk variables, the risk variables with lower retained correlation coefficients also have influence on the result of pressure damage risk prediction. Aiming at the problems, the method adopts a multi-algorithm fusion mode to solve the problems, and establishes a pressure damage risk prediction model by utilizing a random forest model and a multiple logistic regression model.
Preferably, the random forest model is a combination of classification trees into random forests, using twice randomization in the construction of each decision tree: firstly, training data used in the process of constructing a decision tree is randomly acquired from original data by a bootstrap method; and secondly, the interpretation variable used by each decision tree is obtained randomly on the original feature set to generate a plurality of classification trees, and then the results of the classification trees are summarized.
S202: regression modeling is performed on the first classified training set using the random forest model to generate a first stress injury risk prediction model for the first classified training set. Preferably, the step of generating the first pressure damage risk prediction model comprises:
s2021: and classifying medical record data in the first classification training set by using the random forest model so as to acquire first-class risk variables related to the first classification training set.
S2022: regression is performed on the first classification training set and first type risk variables corresponding to the first classification training set based on the random forest model so as to obtain first weights of correlations among the first type risk variables.
S2023: dividing the first classification training set based on the first weight to form a plurality of second classification training sets, and modeling the plurality of second classification training sets by adopting a random forest model to generate a plurality of first pressure injury risk prediction models. Preferably, the method adopts a random forest model to classify the first classification training set again so as to comprehensively screen out risk variables related to the pressure injury in the first classification training set, namely, first type risk variables. And modeling the first type risk variables obtained by screening based on a multiple logistic regression model so as to obtain the interrelationship or the association degree between the first type risk variables, and further, the invention can screen according to the interrelationship between the first type risk variables to obtain relatively isolated variables in the first type risk variables, and classify the first classification training set by the isolated variables to obtain the second classification training set. Through this setting method, the beneficial effect who reaches is:
The second classification training set obtained by classifying the first classification training set through the first weight is equivalent to classifying the specific noise data in the first classification training set, and the random forest model modeling is carried out after the same specific noise data are classified into the same group, so that the influence caused by noise can be remarkably reduced, the occurrence of the phenomenon of excessive fitting is avoided, and the constructed risk prediction model can be generalized (applied) to new medical record data. Noise data in this embodiment refers to irrelevant risk variables. Specifically, the chi-square detection adopted in step S201 may not accurately classify the training set to obtain the first classification training set, and the reclassification method adopting the random forest model in step S202 is used for further accurately screening the irrelevant risk variables in the first classification training set. In addition, since the sensitivity of the multiple logistic regression model to independent variables, namely the interrelation between the risk variables, can relatively accurately acquire the interrelation between the first type of risk variables, the multiple logistic regression model is used for carrying out binary regression prediction on the first type of risk variables, so that the first weight of the interrelation between the multiple first type of risk variables can be obtained, the interrelation between the multiple first type of risk variables is quantitatively evaluated through the first weight, and further, the relatively isolated first type of risk variables in the multiple first type of risk variables can be acquired, namely the isolation degree of the first type of risk variables can be evaluated according to the first weight, and the first classification training set is divided according to the isolation degree of the first type of risk variables, so that the second classification training set is obtained. At this time, the medical record data in the second classification training set are medical record data with similar association degrees of the risk variables, so that interference caused by specific first-type risk variables is reduced, and the problem that the random forest model is fitted excessively is avoided.
Preferably, step S2022 further comprises the steps of:
establishing a multiple logistic regression model by taking the first type of risk variables as independent variables and taking whether the correlation among the first type of risk variables is a dependent variable or not;
and acquiring correlations among the first type of risk variables based on the multiple logistic regression model. Preferably, the first type risk variable is randomly selected, and the correlation between the first type risk variable and other first type risk variables is calculated based on a multiple logistic regression model.
Preferably, the degree of association between the plurality of first type risk variables is obtained based on a multiple logistic regression model. The first classification training set is partitioned based on the degree of association to generate a second classification training set. Through this setting method, the beneficial effect who reaches is:
although the isolated first-class risk variables cannot be accurately obtained by calculating the association degree between the first-class risk variables, and specific noise cannot be eliminated to the greatest extent, the risk of division failure caused by less related data volume can be avoided by dividing the association degree between the first-class risk variables.
Preferably, the first type of risk variable is randomly selected. And calculating the degree of association between the first type of risk variable and other first type of risk variables based on the multiple logistic regression model. Preferably, the degree of association can be characterized by calculating regression coefficients. For example, a first type risk variable a is randomly selected, and regression coefficients with other first type risk variables are calculated based on the first type risk variable a. The regression coefficients characterize the extent of change of the other first type of risk variables as the first type of risk variable a changes. For example, when a first type risk variable a varies by one unit, an associated first type risk variable B varies by 1 unit, then the degree of association is 1. If a first type risk variable A varies by 1 unit and an associated first type risk variable B varies by 0.1 units, then the degree of association is 0.1. Preferably, the plurality of first type risk variables with the association degree greater than the third threshold value are screened based on the association degree of the first type risk variables. The third threshold may be set based on the number of actual first-type risk variables and medical record data. Preferably, the third threshold may be a median of the degree of association.
Preferably, the first weight may be used to characterize the degree of association of the first type of risk variable. The step of dividing the first classification training set based on the first weight to form a plurality of second classification training sets is as follows:
constructing a correlation table based on the degree of correlation between each first type of risk variables;
acquiring a first class risk variable pair with a first weight smaller than a first threshold;
calculating the number of the first type risk variables included in the first type risk variable pair based on the correlation table;
if the number of the same first type risk variables exceeds the second threshold, searching for a first type risk variable pair with the next first weight smaller than the second threshold. Preferably, if the number of the same first type risk variables is less than or equal to the second threshold, the first type risk variable with the least number of other first type risk variables is selected as the isolated first type risk variable. Preferably, medical record data containing the isolated first type of risk variable is selected as the second classification training set based on the first classification training set. The first threshold may select a value near zero. The first threshold may be set according to the first weight obtained in practice. Preferably, the first threshold may be a value less than 20% of the average value of the first weights. Preferably, the second threshold value may be set according to the number of risk variables of the first type involved. The second threshold may be 40% of the total number of risk variables of the first type.
S203: and classifying the first pressure damage risk prediction model to obtain a second class of risk variables and a second weight representing model characteristics of the first class of risk variables. The second weight represents a degree of correlation thereof in the first pressure injury risk prediction model with respect to occurrence of pressure injury. The medical record data of the patient can be adapted according to the second weight of the second type of risk variable in actual use.
S2031: based on the coefficient of the kene as a splitting or competing rule of the random forest model, a second type risk variable and a second weight of the first pressure injury risk prediction model are obtained. Preferably, the second weight is a coefficient of kunity. The second weight represents the degree of association of the second type of risk variable with the pressure injury. The random forest algorithm utilizes a boost-strap sampling method to extract N samples from a second classification training set, then decision tree models are respectively built for the N samples, each decision tree consists of a root node, leaf nodes and branches, each decision tree model comprises random 4 variable attributes, the nodes are split in an optimal splitting mode in 4 features, and each tree grows completely without pruning, so that a combined classifier is obtained. And classifying each test sample by using the N decision tree models to obtain N classification results, and finally voting the N classification results to determine the final classification result. Preferably, the expression of the pre-grouping kunning coefficient G (t) is as follows:
Preferably, p (j|t) represents the normalized probability of the j-th class of output variable in node t. When the output quantity of the node samples all take the same sample, the difference of the output variable values is minimum, and the coefficient of the radix is 0. When the probabilities of the values of the classes are the same, the difference of the values of the output variables is maximum, and the coefficient of the radix is also maximum.
Preferably, the classification tree measures the degree of heterogeneity reduction Δg (t) using a reduced amount of the coefficient of kunity. Preferably, a simple majority voting method may be employed to determine the final classification result. The final classification decision is as follows:
where H (x) represents the combined classification model. h is a i (x) Representing a single decision classification model. Y represents a target variable. I (·) represents an indication function. The whole process is repeated k times. Samples that have never been drawn are referred to as out-of-bag data. Preferably, the effect of the model can be measured in terms of the residual mean square of the out-of-bag data predictors.
Preferably, the patient's relevant medical record data may be complex, i.e. two or more first pressure injury risk prediction models may be adapted in the medical record data, so that the first risk prediction models need to be guaranteed in their combinability, i.e. the two or more first pressure injury risk prediction models need to be able to be combined, and to incorporate the expanded capabilities of the new risk variables. Preferably, the step S300 is:
Averaging the data volume of the second class risk variables in the second class training set;
the second type of risk variable is partitioned based on the degree of association, thereby generating a plurality of third type of risk variables. Preferably, modeling is based on a plurality of third type risk variables to generate a second pressure damage risk prediction model. Preferably, the number of the second type risk variables contained in each divided third type risk variable is the same.
Through this setting method, the beneficial effect who reaches is:
since the generated first pressure injury risk prediction model needs to have the capability of incorporating new risk variables and combining a plurality of first pressure injury risk prediction models with each other, the expanded or combined first pressure injury risk prediction model needs to ensure the stability of prediction. However, the first stress risk prediction model is constructed according to a random forest model, so if a new risk variable is included and the data size is large, the output of the first stress injury risk prediction model may be inclined to the side with more data size/data record, and thus the prediction result deflection can be avoided by averaging the data sizes of the second type risk variables in the second classification training set. In addition, if more associated risk variables exist in the second type of risk variables, the output of the first pressure damage risk prediction model is inclined to one side of the more associated risk variables, so that a plurality of third type of risk variables are obtained through association degree division, and the plurality of third type of risk variables contain the same number of second type of risk variables, so that the classification number of the risk variables is balanced, and inclination of a risk prediction result can be avoided.
Preferably, referring to fig. 2, the present embodiment further provides a device for predicting risk of pressure injury. The apparatus comprises a processing unit 100, a storage unit 200 and a communication unit 300. Preferably, the processing unit 100 may perform the above steps S100 to S400 and steps S10 to S70. In one aspect, the processing unit 100 may be configured to perform steps S100 to S400. On the other hand, the processing unit 100 may be configured to perform steps S10 to S70. Preferably, the storage unit 200 is configured to store the aforementioned medical record data, the analyzable medical record data, the training set, the verification set, the pressure damage risk prediction model, the risk variable, the regression coefficient, and the like. Preferably, the communication unit 300 is used for accessing a network and connecting devices, thereby obtaining medical record data. For example, the communication unit 300 can access the medical record database through a network such as the internet, the internet of things, a mobile network, or an ethernet network by a wired and/or wireless manner. Preferably, the communication unit 300 may also be an RJ-45 interface of an Ethernet network, a BNC interface of a thin coaxial cable, an AUI interface of a thick coaxial cable, an FDDI interface, an ATM interface, or the like. The communication unit 300 may also be a Wi-Fi module, a bluetooth module, a Zigbee module, or the like. Preferably, the communication unit 300 may also be a combination of an RJ-45 interface, a BNC interface, a thick coaxial cable AUI interface, an FDDI interface, an ATM interface, a Wi-Fi module, a bluetooth module, a Zigbee module.
Preferably, the processing unit 100 may be a central processing unit (Central Processing Unit, CPU), a general purpose processor, a digital signal processor (Digital Signal Processor, DSP), an Application-specific integrated circuit (ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA), a graphics processor (Graphics Processing Unit, GPU) or other programmable logic device, a transistor logic device, a hardware component, or any combination thereof.
Preferably, the storage unit 200 may be a magnetic disk, a hard disk, an optical disk, a mobile hard disk, a solid state disk, a flash memory, etc.
The present specification contains several inventive concepts, and applicant reserves the right to issue a divisional application according to each of the inventive concepts. The description of the invention encompasses multiple inventive concepts, such as "preferably," "according to a preferred embodiment," or "optionally," all means that the corresponding paragraph discloses a separate concept, and that the applicant reserves the right to filed a divisional application according to each inventive concept.
It should be noted that the above-described embodiments are exemplary, and that a person skilled in the art, in light of the present disclosure, may devise various solutions that fall within the scope of the present disclosure and fall within the scope of the present disclosure. It should be understood by those skilled in the art that the present description and drawings are illustrative and not limiting to the claims. The scope of the invention is defined by the claims and their equivalents.
Claims (6)
1. A device for predicting risk of pressure injury based on multi-algorithm fusion, which is used for predicting risk of pressure injury obtained by medical institutions, and is characterized by comprising a processing unit (100), wherein,
the processing unit (100) is configured to perform the steps of:
obtaining the analyzable medical record data and reassigning the analyzable medical record data based on the significant risk variables, thereby generating a first classification training set for the specific target population;
performing regression modeling on the first classification training set by using a random forest model, so as to generate a first pressure damage risk prediction model related to the first classification training set;
classifying the first pressure damage risk prediction model to obtain a second class risk variable and a second weight which characterize the characteristics of the first pressure damage risk prediction model, and combining a plurality of first pressure damage risk prediction models based on the second class risk variable to generate a second pressure damage risk prediction model;
the second weight represents the association degree of the second type of risk variable with occurrence of the pressure injury in the first pressure injury risk prediction model;
based on the coefficient of the foundation as a splitting or competing rule of the random forest model, obtaining a second type risk variable and a second weight of the first pressure damage risk prediction model, wherein the second weight is the coefficient of the foundation;
Performing pressure injury risk prediction by using a second pressure injury risk prediction model,
the processing unit (100) is configured to:
classifying medical record data in a first classification training set by utilizing a random forest model so as to acquire first type risk variables related to the first classification training set;
regression is carried out on the first classification training set and first type risk variables corresponding to the first classification training set based on a random forest model so as to obtain first weights representing correlations among a plurality of first type risk variables;
dividing the first classification training set based on the first weight to form a plurality of second classification training sets, and modeling the plurality of second classification training sets by adopting a random forest model to generate a plurality of first pressure injury risk prediction models.
2. The pressure damage risk prediction device according to claim 1, wherein the processing unit (100) is configured to:
establishing a multiple logistic regression model by taking the first type of risk variables as independent variables and taking whether the correlation among the first type of risk variables is a dependent variable or not;
acquiring the association degree among a plurality of first-type risk variables based on a multiple logistic regression model;
the first classification training set is partitioned based on the degree of association to generate a second classification training set.
3. The pressure damage risk prediction device according to claim 1, wherein the processing unit (100) is configured to:
constructing a correlation table based on the degree of correlation between each first type of risk variables;
acquiring a first class risk variable pair with a first weight smaller than a first threshold;
the number of first type risk variables included in the first type risk variable pair is calculated based on the correlation table.
4. The pressure damage risk prediction device according to claim 1, wherein the processing unit (100) is configured to:
if the number of the same first type risk variables exceeds a second threshold, searching a first type risk variable pair of which the next first weight is smaller than the second threshold;
if the number of the same first type risk variables is smaller than or equal to a second threshold value, selecting the first type risk variable with the least number of other first type risk variables as the isolated first type risk variable.
5. The pressure damage risk prediction device according to claim 1, wherein the processing unit (100) is configured to:
averaging the data volume of the second class risk variables in the second class training set;
dividing the second type of risk variables based on the degree of association, thereby generating a plurality of third type of risk variables;
Modeling based on a plurality of third class risk variables to generate a second pressure damage risk prediction model.
6. A method for predicting risk of pressure injury for predicting risk of acquired pressure injury in a medical facility, the method comprising:
obtaining the analyzable medical record data and reassigning the analyzable medical record data based on the significant risk variables, thereby generating a first classification training set for the specific target population;
performing regression modeling on the first classification training set by using a random forest model, so as to generate a first pressure damage risk prediction model related to the first classification training set;
classifying the first pressure damage risk prediction model to obtain a second class risk variable and a second weight which characterize the characteristics of the first pressure damage risk prediction model, and combining a plurality of first pressure damage risk prediction models based on the second class risk variable to generate a second pressure damage risk prediction model;
the second weight represents the association degree of the second type of risk variable with occurrence of the pressure injury in the first pressure injury risk prediction model;
based on the coefficient of the foundation as a splitting or competing rule of the random forest model, obtaining a second type risk variable and a second weight of the first pressure damage risk prediction model, wherein the second weight is the coefficient of the foundation;
Performing pressure injury risk prediction by using a second pressure injury risk prediction model,
the processing unit (100) is configured to:
classifying medical record data in a first classification training set by utilizing a random forest model so as to acquire first type risk variables related to the first classification training set;
regression is carried out on the first classification training set and first type risk variables corresponding to the first classification training set based on a random forest model so as to obtain first weights representing correlations among a plurality of first type risk variables;
dividing the first classification training set based on the first weight to form a plurality of second classification training sets, and modeling the plurality of second classification training sets by adopting a random forest model to generate a plurality of first pressure injury risk prediction models.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110869608.8A CN113782183B (en) | 2021-07-29 | 2021-07-29 | Device and method for predicting risk of pressure injury based on multi-algorithm fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110869608.8A CN113782183B (en) | 2021-07-29 | 2021-07-29 | Device and method for predicting risk of pressure injury based on multi-algorithm fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113782183A CN113782183A (en) | 2021-12-10 |
| CN113782183B true CN113782183B (en) | 2023-07-14 |
Family
ID=78836358
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110869608.8A Active CN113782183B (en) | 2021-07-29 | 2021-07-29 | Device and method for predicting risk of pressure injury based on multi-algorithm fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113782183B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116309515A (en) * | 2023-03-31 | 2023-06-23 | 广东省人民医院 | Method for constructing lung sub-centimeter nodule invasiveness prediction model, diagnosis model and diagnosis device |
| CN116452559B (en) * | 2023-04-19 | 2024-02-20 | 深圳市睿法生物科技有限公司 | Tumor focus positioning method and device based on ctDNA fragmentation mode |
| CN117727448B (en) * | 2024-02-06 | 2024-04-19 | 四川省医学科学院·四川省人民医院 | Medical conjuncted-based intelligent decision control system for pressure injury |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009062A (en) * | 2019-04-18 | 2019-07-12 | 成都四方伟业软件股份有限公司 | Disaggregated model training method and device |
| CN110827992A (en) * | 2019-11-15 | 2020-02-21 | 袁洪 | Preoperative prediction method for acute renal injury after hypertension operation |
| WO2021081445A1 (en) * | 2019-10-25 | 2021-04-29 | Xy.Health, Inc. | System and method with federated learning model for geotemporal data associated medical prediction applications |
| CN113178243A (en) * | 2021-04-26 | 2021-07-27 | 南京医科大学 | Early warning method and system for acute kidney injury after extracorporeal circulation cardiac surgery |
-
2021
- 2021-07-29 CN CN202110869608.8A patent/CN113782183B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009062A (en) * | 2019-04-18 | 2019-07-12 | 成都四方伟业软件股份有限公司 | Disaggregated model training method and device |
| WO2021081445A1 (en) * | 2019-10-25 | 2021-04-29 | Xy.Health, Inc. | System and method with federated learning model for geotemporal data associated medical prediction applications |
| CN110827992A (en) * | 2019-11-15 | 2020-02-21 | 袁洪 | Preoperative prediction method for acute renal injury after hypertension operation |
| CN113178243A (en) * | 2021-04-26 | 2021-07-27 | 南京医科大学 | Early warning method and system for acute kidney injury after extracorporeal circulation cardiac surgery |
Non-Patent Citations (1)
| Title |
|---|
| 冠心病风险因素识别及其预测模型构建;李婕;向菲;;中华医学图书情报杂志(06);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113782183A (en) | 2021-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113782183B (en) | Device and method for predicting risk of pressure injury based on multi-algorithm fusion | |
| WO2021120936A1 (en) | Chronic disease prediction system based on multi-task learning model | |
| Mounika et al. | Prediction of type-2 diabetes using machine learning algorithms | |
| CN113744865B (en) | Regression analysis-based pressure damage risk prediction model correction method | |
| CN112542242A (en) | Data transformation/symptom scoring | |
| Popkes et al. | Interpretable outcome prediction with sparse Bayesian neural networks in intensive care | |
| CN113593708A (en) | Sepsis prognosis prediction method based on integrated learning algorithm | |
| CN116864139A (en) | Disease risk assessment method, device, computer equipment and readable storage medium | |
| Chou et al. | Expert-augmented automated machine learning optimizes hemodynamic predictors of spinal cord injury outcome | |
| Chinnasamy et al. | Machine learning based cardiovascular disease prediction | |
| CN113593703B (en) | Device and method for constructing pressure injury risk prediction model | |
| Xao et al. | Fasting blood glucose change prediction model based on medical examination data and data mining techniques | |
| CN114927230B (en) | Prognosis decision support system and method for severe heart failure patient based on machine learning | |
| Sun et al. | Artificial intelligence and machine learning: Definition of terms and current concepts in critical care research | |
| Ganesh et al. | Diabetes Prediction using Logistic Regression and Feature Normalization | |
| Alaria et al. | Design Simulation and Assessment of Prediction of Mortality in Intensive Care Unit Using Intelligent Algorithms | |
| WO2023106960A1 (en) | Method for predicting the onset of a medical event in a person's health | |
| Lagunzad et al. | Predicting the Early Sign of Diabetes using ID3 as a Data Model | |
| Kavya et al. | Heart Disease Prediction Using Logistic Regression | |
| Hassan et al. | Efficient prediction of coronary artery disease using machine learning algorithms with feature selection techniques | |
| Ismail et al. | A Comparative Study of Diabetes Classification Based on Machine Learning | |
| Bose et al. | Female Diabetic Prediction in India Using Different Learning Algorithms | |
| Khalaf et al. | Predicting Acute Respiratory Failure Using Fuzzy Classifier | |
| Bhat et al. | Predicting Life Style of Early Diabetes Mellitus Using Machine Learning | |
| Maheswaravenkatesh et al. | Optimal machine learning–based data classification on Internet of medical things environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |