CN113299390A - System and method for predicting in-hospital mortality of acute kidney injury patient - Google Patents
System and method for predicting in-hospital mortality of acute kidney injury patient Download PDFInfo
- Publication number
- CN113299390A CN113299390A CN202110551906.2A CN202110551906A CN113299390A CN 113299390 A CN113299390 A CN 113299390A CN 202110551906 A CN202110551906 A CN 202110551906A CN 113299390 A CN113299390 A CN 113299390A
- Authority
- CN
- China
- Prior art keywords
- node
- layer
- data
- neural network
- network model
- 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.)
- Pending
Links
- 208000009304 Acute Kidney Injury Diseases 0.000 title claims abstract description 96
- 208000033626 Renal failure acute Diseases 0.000 title claims abstract description 77
- 201000011040 acute kidney failure Diseases 0.000 title claims abstract description 77
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000003062 neural network model Methods 0.000 claims abstract description 70
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 31
- 238000013528 artificial neural network Methods 0.000 claims abstract description 15
- 239000011159 matrix material Substances 0.000 claims description 56
- 238000007667 floating Methods 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 9
- 238000004140 cleaning Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000005406 washing Methods 0.000 claims 2
- 230000006870 function Effects 0.000 description 30
- 238000004364 calculation method Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 235000019800 disodium phosphate Nutrition 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 208000028399 Critical Illness Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 230000003907 kidney function Effects 0.000 description 1
- 238000011269 treatment regimen Methods 0.000 description 1
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
-
- 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
- G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Databases & Information Systems (AREA)
- Biophysics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Treatment And Welfare Office Work (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention provides a prediction system and a method of the hospital mortality of acute kidney injury patients, wherein the system comprises: the data input unit is used for acquiring clinical medical data of the patient with acute renal injury and inputting the clinical medical data to the data processing unit; and the data processing unit is used for inputting the clinical medical data into a pre-trained neural network algorithm model and outputting a predicted value of the in-hospital mortality of the acute kidney injury patient. The system and the method provided by the invention utilize the structured neural network model, effectively avoid the noise value of clinical medical data, and improve the accuracy of the in-hospital mortality prediction of acute kidney injury patients.
Description
Technical Field
The invention relates to the technical field of medical data mining, in particular to a system and a method for predicting the in-hospital mortality of patients with acute renal injury.
Background
Acute Kidney Injury (AKI) is a common critical illness in hospitalized patients, and AKI is commonly used to define patients with sudden decline in renal function. Although medical technology has advanced over the past few years, global AKI remains a high-grade disease and has a high mortality rate. Death of AKI patients is well avoided if the corresponding treatment regimen can be discovered and made at an early stage. Currently, clinical experience methods are used to determine whether patients with AKI will suffer nosocomial death, with low accuracy and depending heavily on the experience of the physician.
Disclosure of Invention
In order to solve the technical problems, the invention provides a hospital mortality prediction system and a hospital mortality prediction method for acute kidney injury patients, which can objectively and accurately predict the hospital mortality of the acute kidney injury patients.
In a first aspect, the present invention provides a system for predicting the in-hospital mortality of a patient with acute renal injury, comprising:
the data input unit is used for acquiring clinical medical data of the patient with acute renal injury and inputting the clinical medical data to the data processing unit;
and the data processing unit is used for inputting the clinical medical data into a pre-trained neural network algorithm model and outputting a predicted value of the in-hospital mortality of the acute kidney injury patient.
Further, the data processing unit includes:
the data preprocessing module is used for cleaning noise data and missing data in the clinical medical data, and extracting features of the cleaned data to obtain feature data;
and the mortality prediction module is used for predicting the hospital mortality of the acute kidney injury patient according to the characteristic data.
Further, the performing feature extraction on the cleaned data to obtain feature data includes:
establishing a first neural network model, inputting the cleaned data into the first neural network model, and outputting a first node and a second node; wherein the first node is the probability of hospital death of the acute kidney injury patient and the second node is the probability of hospital non-death of the acute kidney injury patient; wherein the structure of the first neural network model is a four-layer structure;
initializing weights and bias values between the first node and the second node, and setting the node hiding rate between layers in the first neural network model to be 0.2;
and calculating the sample data according to the first neural network model algorithm to obtain the characteristic data.
Further, the prediction of the hospital mortality of the acute kidney injury patient according to the characteristic data comprises the following steps:
establishing a second neural network model, inputting the characteristic data into the second neural network model, and inputting a third node and a fourth node; wherein the third node is the probability of hospital death of the acute kidney injury patient, and the fourth node is the probability of hospital non-death of the acute kidney injury patient; wherein the structure of the second neural network model is a five-layer structure;
initializing the weight and the offset value between the third node and the fourth node, and setting the node hiding rate between the first layer and the second layer in the second neural network model to be 0.2, the node hiding rate between the second layer and the third layer to be 0.3, the node hiding rate between the third layer and the fourth layer to be 0.3, and the node hiding rate between the fourth layer and the fifth layer to be 0.1;
and calculating the sample data according to a second neural network model algorithm, stopping calculating when the loss function value is lower than the loss threshold value, and outputting the predicted value of the in-hospital mortality of the acute renal injury patient.
Further, the first neural network model algorithm includes:
calculating the node output of the target layer according to the node weight and the offset value of the current layer; specifically, it is calculated by the following formula:
wherein z isjIs the node output of the target layer, i is the node of the current layer, m is the total number of nodes of the current layer, omegaiIs the node weight, x, of the current level iiAn input matrix for the current layer i, biIs the bias value of the current layer i;
calculating a sigmoid function value according to an input matrix of a current layer; specifically, it is calculated by the following formula:
wherein, S (x)i) Sigmoid function value, x, of input matrixiAn input matrix of a current layer;
calculating a loss function value; specifically calculated by the following formula:
wherein loss is a loss function value, y is an actual output value, o is a floating point number between 0 and 1, i is a node of the current layer, and m is the total number of nodes of the current layer;
calculating a back propagation value; specifically, it is calculated by the following formula:
wherein, ω isi+1Node weight, ω, for level i +1iIs the node weight of the current layer, eta is the coefficient, bi+1Is the bias value of the i +1 layer, biIs the offset value of the current layer i.
In a second aspect, the present invention provides a method for predicting hospital mortality of a patient with acute renal injury, comprising:
acquiring clinical medical data of a patient with acute renal injury;
and inputting the clinical medical data into a pre-trained neural network algorithm model, and outputting a predicted value of the hospital mortality of the acute kidney injury patient.
Further, the inputting the clinical medical data into a pre-trained neural network algorithm model and outputting the predicted value of the in-hospital mortality of the acute kidney injury patient comprises the following steps:
cleaning noise data and missing data in the clinical medical data, and performing feature extraction on the cleaned data to obtain feature data;
and predicting the hospital mortality of the acute kidney injury patient according to the characteristic data.
Further, the performing feature extraction on the cleaned data to obtain feature data includes:
establishing a first neural network model, inputting the cleaned data into the first neural network model, and outputting a first node and a second node; wherein the first node is the probability of hospital death of the acute kidney injury patient and the second node is the probability of hospital non-death of the acute kidney injury patient; wherein the structure of the first neural network model is a four-layer structure;
initializing weights and bias values between the first node and the second node, and setting the node hiding rate between layers in the first neural network model to be 0.2;
and calculating the sample data according to the first neural network model algorithm to obtain the characteristic data.
Further, the prediction of the hospital mortality of the acute kidney injury patient according to the characteristic data comprises the following steps:
establishing a second neural network model, inputting the characteristic data into the second neural network model, and inputting a third node and a fourth node; wherein the third node is the probability of hospital death of the acute kidney injury patient, and the fourth node is the probability of hospital non-death of the acute kidney injury patient; wherein the structure of the second neural network model is a five-layer structure;
initializing the weight and the offset value between the third node and the fourth node, and setting the node hiding rate between the first layer and the second layer in the second neural network model to be 0.2, the node hiding rate between the second layer and the third layer to be 0.3, the node hiding rate between the third layer and the fourth layer to be 0.3, and the node hiding rate between the fourth layer and the fifth layer to be 0.1;
and calculating the sample data according to a second neural network model algorithm, stopping calculating when the loss function value is lower than the loss threshold value, and outputting the predicted value of the in-hospital mortality of the acute renal injury patient.
Further, the first neural network model algorithm includes:
calculating the node output of the target layer according to the node weight and the offset value of the current layer; specifically, it is calculated by the following formula:
wherein z isjIs the node output of the target layer, i is the node of the current layer, m is the total number of nodes of the current layer, omegaiIs the node weight, x, of the current level iiAn input matrix for the current layer i, biIs the bias value of the current layer i;
calculating a sigmoid function value according to an input matrix of a current layer; specifically, it is calculated by the following formula:
wherein, S (x)i) Sigmoid function value, x, of input matrixiAn input matrix of a current layer;
calculating a loss function value; specifically calculated by the following formula:
wherein loss is a loss function value, y is an actual output value, o is a floating point number between 0 and 1, i is a node of the current layer, and m is the total number of nodes of the current layer;
calculating a back propagation value; specifically, it is calculated by the following formula:
wherein, ω isi+1Node weight, ω, for level i +1iIs the node weight of the current layer, eta is the coefficient, bi+1Is the bias value of the i +1 layer, biIs the offset value of the current layer i.
Compared with the prior art, the embodiment of the invention has the beneficial effects that:
the invention provides a prediction system and a method of the hospital mortality of acute kidney injury patients, wherein the system comprises: the data input unit is used for acquiring clinical medical data of the patient with acute renal injury and inputting the clinical medical data to the data processing unit; and the data processing unit is used for inputting the clinical medical data into a pre-trained neural network algorithm model and outputting a predicted value of the in-hospital mortality of the acute kidney injury patient. The system and the method provided by the invention utilize the structured neural network model, effectively avoid the noise value of clinical medical data, and improve the accuracy of the in-hospital mortality prediction of acute kidney injury patients.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a diagram of an apparatus for a hospital mortality prediction system for acute kidney injury patients, according to an embodiment of the present invention;
FIG. 2 is a diagram of an apparatus of a hospital mortality prediction system for acute kidney injury patients, according to another embodiment of the present invention;
FIG. 3 is a flow chart of a method for predicting hospital mortality in a patient with acute kidney injury according to an embodiment of the present invention;
FIG. 4 is a flow chart of a method for predicting hospital mortality in a patient with acute kidney injury according to another embodiment of the present invention;
fig. 5 is a block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be understood that the step numbers used herein are for convenience of description only and are not intended as limitations on the order in which the steps are performed.
It is to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
The terms "comprises" and "comprising" indicate the presence of the described features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
The term "and/or" refers to and includes any and all possible combinations of one or more of the associated listed items.
A first aspect.
Referring to FIGS. 1-2, the present invention provides a system for predicting hospital mortality in patients with acute renal injury, comprising:
the data input unit 10 is used for acquiring clinical medical data of the patient with acute renal injury and inputting the clinical medical data into the data processing unit.
And the data processing unit 20 is used for inputting the clinical medical data into a pre-trained neural network algorithm model and outputting a predicted value of the in-hospital mortality of the acute kidney injury patient.
In a specific embodiment, the data processing unit 20 includes:
and the data preprocessing module 21 is configured to clean noise data and missing data in the clinical medical data, and perform feature extraction on the cleaned data to obtain feature data.
In a specific embodiment, the extracting features from the cleaned data to obtain feature data includes:
establishing a first neural network model, inputting the cleaned data into the first neural network model, and outputting a first node and a second node; wherein the first node is the probability of hospital death of the acute kidney injury patient and the second node is the probability of hospital non-death of the acute kidney injury patient; wherein the first neural network model has a four-layer structure.
Initializing the weight and the offset value between the first node and the second node, and setting the node hiding rate between layers in the first neural network model to be 0.2.
And calculating the sample data according to the first neural network model algorithm to obtain the characteristic data.
In one embodiment, the first neural network model algorithm includes:
calculating the node output of the target layer according to the node weight and the offset value of the current layer; specifically, it is calculated by the following formula:
wherein z isjIs the node output of the target layer, i is the node of the current layer, m is the total number of nodes of the current layer, omegaiIs the node weight, x, of the current level iiAn input matrix for the current layer i, biIs the bias value of the current layer i;
calculating a sigmoid function value according to an input matrix of a current layer; specifically, it is calculated by the following formula:
wherein, S (x)i) Sigmoid function value, x, of input matrixiAn input matrix of a current layer;
calculating a loss function value; specifically calculated by the following formula:
wherein loss is a loss function value, y is an actual output value, o is a floating point number between 0 and 1, i is a node of the current layer, and m is the total number of nodes of the current layer;
calculating a back propagation value; specifically, it is calculated by the following formula:
wherein, ω isi+1Node weight, ω, for level i +1iIs the node weight of the current layer, eta is the coefficient, bi+1Is the bias value of the i +1 layer, biIs the offset value of the current layer i.
And the mortality prediction module 22 is used for predicting the hospital mortality of the acute kidney injury patient according to the characteristic data.
In a specific embodiment, said predicting in-hospital mortality of acute kidney injury patients based on said characteristic data comprises:
establishing a second neural network model, inputting the characteristic data into the second neural network model, and inputting a third node and a fourth node; wherein the third node is the probability of hospital death of the acute kidney injury patient, and the fourth node is the probability of hospital non-death of the acute kidney injury patient; wherein the structure of the second neural network model is a five-layer structure.
Initializing the weight and the offset value between the third node and the fourth node, and setting the node hiding rate between the first layer and the second layer in the second neural network model to be 0.2, the node hiding rate between the second layer and the third layer to be 0.3, the node hiding rate between the third layer and the fourth layer to be 0.3, and the node hiding rate between the fourth layer and the fifth layer to be 0.1.
And calculating the sample data according to a second neural network model algorithm, stopping calculating when the loss function value is lower than the loss threshold value, and outputting the predicted value of the in-hospital mortality of the acute renal injury patient.
The system provided by the invention utilizes the structured neural network model, effectively avoids the noise value of clinical medical data, and improves the accuracy of the prediction of the in-hospital mortality of acute kidney injury patients.
A second aspect.
Referring to FIGS. 3-4, the present invention provides a method for predicting hospital mortality in patients with acute renal injury, comprising:
s10, obtaining clinical medical data of the acute kidney injury patient.
And S20, inputting the clinical medical data into a pre-trained neural network algorithm model, and outputting the predicted value of the in-hospital mortality of the acute kidney injury patient.
In a specific embodiment, the S20, inputting the clinical medical data into a pre-trained neural network algorithm model, and outputting a predicted value of the in-hospital mortality of the acute kidney injury patient, includes:
and S21, cleaning the noise data and the missing data in the clinical medical data, and performing feature extraction on the cleaned data to obtain feature data.
In a specific embodiment, the extracting features from the cleaned data to obtain feature data includes:
establishing a first neural network model, inputting the cleaned data into the first neural network model, and outputting a first node and a second node; wherein the first node is the probability of hospital death of the acute kidney injury patient and the second node is the probability of hospital non-death of the acute kidney injury patient; wherein the first neural network model has a four-layer structure.
Initializing the weight and the offset value between the first node and the second node, and setting the node hiding rate between layers in the first neural network model to be 0.2.
And calculating the sample data according to the first neural network model algorithm to obtain the characteristic data.
In one embodiment, the first neural network model algorithm includes:
calculating the node output of the target layer according to the node weight and the offset value of the current layer; specifically, it is calculated by the following formula:
wherein z isjIs the node output of the target layer, i is the node of the current layer, m is the total number of nodes of the current layer, omegaiIs the node weight, x, of the current level iiAn input matrix for the current layer i, biIs the bias value of the current layer i;
calculating a sigmoid function value according to an input matrix of a current layer; specifically, it is calculated by the following formula:
wherein, S (x)i) Sigmoid function value, x, of input matrixiAn input matrix of a current layer;
calculating a loss function value; specifically calculated by the following formula:
wherein loss is a loss function value, y is an actual output value, o is a floating point number between 0 and 1, i is a node of the current layer, and m is the total number of nodes of the current layer;
calculating a back propagation value; specifically, it is calculated by the following formula:
wherein, ω isi+1Node weight, ω, for level i +1iIs the node weight of the current layer, eta is the coefficient, bi+1Is the bias value of the i +1 layer, biIs the offset value of the current layer i.
And S22, predicting the hospital mortality of the acute kidney injury patient according to the characteristic data.
In a specific embodiment, said predicting in-hospital mortality of acute kidney injury patients based on said characteristic data comprises:
establishing a second neural network model, inputting the characteristic data into the second neural network model, and inputting a third node and a fourth node; wherein the third node is the probability of hospital death of the acute kidney injury patient, and the fourth node is the probability of hospital non-death of the acute kidney injury patient; wherein the structure of the second neural network model is a five-layer structure.
Initializing the weight and the offset value between the third node and the fourth node, and setting the node hiding rate between the first layer and the second layer in the second neural network model to be 0.2, the node hiding rate between the second layer and the third layer to be 0.3, the node hiding rate between the third layer and the fourth layer to be 0.3, and the node hiding rate between the fourth layer and the fifth layer to be 0.1.
And calculating the sample data according to a second neural network model algorithm, stopping calculating when the loss function value is lower than the loss threshold value, and outputting the predicted value of the in-hospital mortality of the acute renal injury patient.
The method provided by the invention utilizes the structured neural network model, effectively avoids the noise value of clinical medical data, and improves the accuracy of the prediction of the in-hospital mortality of acute kidney injury patients.
In one embodiment, the invention provides a method for predicting the in-hospital mortality of AKI patients by combining the clinical data of the AKI patients with a deep neural network algorithm.
The data that this patent adopted include: the clinical indexes are as follows:
the method for predicting the in-hospital mortality comprises the following steps:
stepl: characteristic value extraction: listing 38 items of input data, and firstly, sorting the characteristic values by using a neural network algorithm;
step1.1 establishes a neural network structure with a structure of 38-80-50-2, and inputs 38 indexes x in the table1,x2......x38The output is two nodes, namely the probability of death in the hospital and the probability of non-death in the hospital;
initializing weight omega and offset value b between nodes in a Step1.2 random number mode;
step1.3 sets the node hiding rates between layers to be 0.2 respectively;
the Step1.4 data sample has 17000 patients which are divided into 50 patients to form a matrix, and the structure of the patient and index matrix in each group is a matrix with the size of 50 x 38;
step1.5, according to the following calculation formula, if the connection weight matrix of the first layer and the second layer is 38 × 64 (the number of nodes of the second layer is 80 × the (1 — the display rate of the first node is 0.2), and 64 nodes are randomly extracted from 80 nodes), the data matrix and the weight matrix are subjected to matrix operation, a bias value specific to each node is added, and then an activation function sigmoid function value is calculated, so that the calculation result of the first layer and the second layer is a matrix of 50 × 64 ([50 × 38] × [38 × 64] ═ 50 × 64 ]);
ω is a weight matrix, x is an input matrix, and b is an offset value matrix.
S (x) is sigmoid function.
Step1.6 the result of the calculation of the second layer- > the third layer is a matrix of 50 x 40 size according to the above formula; the result of the third layer- > the fourth layer is a matrix of 50 x 2 size;
step1.7 calculate the loss function value:
y is the actual output value, which is two numbers of 0 and 1, 0 is no death in the hospital, 1 is death in the hospital, and o is a floating point number between the two calculated output values, which is between 0 and 1, and represents the probability of whether or not there is death in the hospital. And m is the number of output nodes. The output loss is a matrix of 50 x 1.
Step1.8, calculating an average value of each row of the result matrix obtained in the last step to obtain a forward propagation result, wherein the result is a floating point numerical value;
step1.9 calculation of the back-propagation:
step1.10 returns to Step1.5, inputs the next set of data for calculation, repeats Step1.5-Step1.9
Step1.11 when 17000 patients have been run, looking at the weight matrix omega 1 between the first layer and the second layer, selecting the points with the smallest weight in the input nodes, deleting the points, namely changing 38 input variables into 37, and using the remaining 37 variables to execute the steps Step1.5-Step1.10 again. Until the input variables become 25.
Step 2: in-hospital mortality prediction;
the 25 indexes x obtained by Step1 are utilized in the Step1,x2......x25And performing in-hospital mortality prediction of AKI patients.
Step2.1 establishes a neural network with a structure of 25-50-100-80-2 (different from the network of Step 1), and inputs 25 indexes x1,x2......x25The output is two nodes which are respectively the probability of death in the hospital and the probability of non-death in the hospital;
initializing weight omega and offset value b between nodes in a Step2.2 random number mode;
step2.3 sets the node hiding rates between layers as follows: 0.2, 0.3, 0.3, 0.1;
the Step2.4 data sample comprises 12000 training set patients which are divided into a group of 60 patients to form a matrix, and the structure of the patient and index matrix in each group is a matrix with the size of 60 × 25;
step2.5, according to the following calculation formula, if the connection weight matrix of the first layer and the second layer is 25 × 40 (the number of nodes of the second layer is 50 × 40 (1 — the display rate of the first node is 0.2), and 40 nodes are randomly extracted from 50 nodes), the data matrix and the weight matrix are subjected to matrix operation, a bias value specific to each node is added, and then an activation function sigmoid function value is calculated, so that the calculation result of the first layer and the second layer is a 60 × 40 matrix ([60 × 25] × [25 × 40] ═ 60 ]);
ω is a weight matrix, x is an input matrix, and b is an offset value matrix.
S (x) is sigmoid function.
Step2.6 according to the above formula, the result of the second layer- > the third layer is calculated to be a matrix with the size of 60 x 70; the result of the third layer- > the fourth layer is a matrix of 60 × 56 size; the result of the fourth layer- > fifth layer is a matrix of 60 × 2 size;
step2.7 calculate the loss function value:
y is the actual output value, which is two numbers of 0 and 1, 0 is no death in the hospital, 1 is death in the hospital, and o is a floating point number between the two calculated output values, which is between 0 and 1, and represents the probability of whether or not there is death in the hospital. And m is the number of output nodes. The output loss is a matrix of 60 x 1.
Step2.8, calculating an average value of each row of the result matrix obtained in the last step to obtain a forward propagation result, wherein the result is a floating point numerical value;
step2.9 calculation of the back-propagation:
returning to Step2.5 at Step2.10, inputting the next set of data for calculation, and repeating Step2.5-Step2.9
Step2.11 sets the cycle number epoch to 5000, i.e. how many times step2.5-step2.9 are repeated in total; the loop is stopped when the loss function value loss _ stop to be stopped is set to 0.001, that is, the value of the loss function loss.
In a third aspect.
The present invention provides an electronic device, including:
a processor, a memory, and a bus;
the bus is used for connecting the processor and the memory;
the memory is used for storing operation instructions;
the processor is used for calling the operation instructions, and the executable instructions cause the processor to execute the operation corresponding to the method for predicting the in-hospital mortality of the acute kidney injury patient as shown in the second aspect of the application.
In an alternative embodiment, an electronic device is provided, as shown in fig. 5, the electronic device 5000 shown in fig. 5 includes: a processor 5001 and a memory 5003. The processor 5001 and the memory 5003 are coupled, such as via a bus 5002. Optionally, the electronic device 5000 may also include a transceiver 5004. It should be noted that the transceiver 5004 is not limited to one in practical application, and the structure of the electronic device 5000 is not limited to the embodiment of the present application.
The processor 5001 may be a CPU, general purpose processor, DSP, ASIC, FPGA or other programmable logic device, transistor logic device, hardware component, or any combination thereof. Which may implement or perform the various illustrative logical blocks, modules, and circuits described in connection with the disclosure. The processor 5001 may also be a combination of processors implementing computing functionality, e.g., a combination comprising one or more microprocessors, a combination of DSPs and microprocessors, or the like.
The memory 5003 may be, but is not limited to, a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, an EEPROM, a CD-ROM or other optical disk storage, optical disk storage (including compact disk, laser disk, optical disk, digital versatile disk, blu-ray disk, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer.
The memory 5003 is used for storing application program codes for executing the present solution, and the execution is controlled by the processor 5001. The processor 5001 is configured to execute application program code stored in the memory 5003 to implement the teachings of any of the foregoing method embodiments.
Among them, electronic devices include but are not limited to: mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and the like, and fixed terminals such as digital TVs, desktop computers, and the like.
A fourth aspect.
The present invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method for predicting in-hospital mortality in a patient with acute renal injury as set forth in the second aspect of the present application.
Yet another embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored, which, when run on a computer, enables the computer to perform the corresponding content in the aforementioned method embodiments.
Claims (10)
1. An in-hospital mortality prediction system for patients with acute kidney injury, comprising:
the data input unit is used for acquiring clinical medical data of the patient with acute renal injury and inputting the clinical medical data to the data processing unit;
and the data processing unit is used for inputting the clinical medical data into a pre-trained neural network algorithm model and outputting a predicted value of the in-hospital mortality of the acute kidney injury patient.
2. The system of claim 1, wherein the data processing unit comprises:
the data preprocessing module is used for cleaning noise data and missing data in the clinical medical data, and extracting features of the cleaned data to obtain feature data;
and the mortality prediction module is used for predicting the hospital mortality of the acute kidney injury patient according to the characteristic data.
3. The acute kidney injury patient in-hospital mortality prediction system of claim 2, wherein the feature extraction of the data after washing to obtain feature data comprises:
establishing a first neural network model, inputting the cleaned data into the first neural network model, and outputting a first node and a second node; wherein the first node is the probability of hospital death of the acute kidney injury patient and the second node is the probability of hospital non-death of the acute kidney injury patient; wherein the structure of the first neural network model is a four-layer structure;
initializing weights and bias values between the first node and the second node, and setting the node hiding rate between layers in the first neural network model to be 0.2;
and calculating the sample data according to the first neural network model algorithm to obtain the characteristic data.
4. The acute kidney injury patient hospital mortality prediction system of claim 2, wherein the prediction of acute kidney injury patient hospital mortality based on the characteristic data comprises:
establishing a second neural network model, inputting the characteristic data into the second neural network model, and inputting a third node and a fourth node; wherein the third node is the probability of hospital death of the acute kidney injury patient, and the fourth node is the probability of hospital non-death of the acute kidney injury patient; wherein the structure of the second neural network model is a five-layer structure;
initializing the weight and the offset value between the third node and the fourth node, and setting the node hiding rate between the first layer and the second layer in the second neural network model to be 0.2, the node hiding rate between the second layer and the third layer to be 0.3, the node hiding rate between the third layer and the fourth layer to be 0.3, and the node hiding rate between the fourth layer and the fifth layer to be 0.1;
and calculating the sample data according to a second neural network model algorithm, stopping calculating when the loss function value is lower than the loss threshold value, and outputting the predicted value of the in-hospital mortality of the acute renal injury patient.
5. The acute kidney injury patient hospital mortality prediction system of claim 3 wherein said first neural network model algorithm comprises:
calculating the node output of the target layer according to the node weight and the offset value of the current layer; specifically, it is calculated by the following formula:
wherein z isjIs the node output of the target layer, i is the node of the current layer, m is the total number of nodes of the current layer, omegaiIs the node weight, x, of the current level iiAn input matrix for the current layer i, biIs the bias value of the current layer i;
calculating a sigmoid function value according to an input matrix of a current layer; specifically, it is calculated by the following formula:
wherein, S (x)i) Sigmoid function value, x, of input matrixiAn input matrix of a current layer;
calculating a loss function value; specifically calculated by the following formula:
wherein loss is a loss function value, y is an actual output value, o is a floating point number between 0 and 1, i is a node of the current layer, and m is the total number of nodes of the current layer;
calculating a back propagation value; specifically, it is calculated by the following formula:
wherein, ω isi+1Node weight, ω, for level i +1iIs the weight of the node at the current level,eta is a coefficient, bi+1Is the bias value of the i +1 layer, biIs the offset value of the current layer i.
6. A method for predicting in-hospital mortality of a patient with acute kidney injury, comprising:
acquiring clinical medical data of a patient with acute renal injury;
and inputting the clinical medical data into a pre-trained neural network algorithm model, and outputting a predicted value of the hospital mortality of the acute kidney injury patient.
7. The method of claim 6, wherein the inputting the clinical medical data into a pre-trained neural network algorithm model and outputting the predicted value of the hospital mortality of the acute kidney injury patient comprises:
cleaning noise data and missing data in the clinical medical data, and performing feature extraction on the cleaned data to obtain feature data;
and predicting the hospital mortality of the acute kidney injury patient according to the characteristic data.
8. The method of claim 7, wherein the performing feature extraction on the data after washing to obtain feature data comprises:
establishing a first neural network model, inputting the cleaned data into the first neural network model, and outputting a first node and a second node; wherein the first node is the probability of hospital death of the acute kidney injury patient and the second node is the probability of hospital non-death of the acute kidney injury patient; wherein the structure of the first neural network model is a four-layer structure;
initializing weights and bias values between the first node and the second node, and setting the node hiding rate between layers in the first neural network model to be 0.2;
and calculating the sample data according to the first neural network model algorithm to obtain the characteristic data.
9. The method of claim 7, wherein the predicting the hospital mortality of the acute kidney injury patient according to the characteristic data comprises:
establishing a second neural network model, inputting the characteristic data into the second neural network model, and inputting a third node and a fourth node; wherein the third node is the probability of hospital death of the acute kidney injury patient, and the fourth node is the probability of hospital non-death of the acute kidney injury patient; wherein the structure of the second neural network model is a five-layer structure;
initializing the weight and the offset value between the third node and the fourth node, and setting the node hiding rate between the first layer and the second layer in the second neural network model to be 0.2, the node hiding rate between the second layer and the third layer to be 0.3, the node hiding rate between the third layer and the fourth layer to be 0.3, and the node hiding rate between the fourth layer and the fifth layer to be 0.1;
and calculating the sample data according to a second neural network model algorithm, stopping calculating when the loss function value is lower than the loss threshold value, and outputting the predicted value of the in-hospital mortality of the acute renal injury patient.
10. The method of claim 8, wherein the first neural network model algorithm comprises:
calculating the node output of the target layer according to the node weight and the offset value of the current layer; specifically, it is calculated by the following formula:
wherein z isjIs the node output of the target layer, i is the node of the current layer, m is the total number of nodes of the current layer, omegaiOf current layer iNode weight, xiAn input matrix for the current layer i, biIs the bias value of the current layer i;
calculating a sigmoid function value according to an input matrix of a current layer; specifically, it is calculated by the following formula:
wherein, S (x)i) Sigmoid function value, x, of input matrixiAn input matrix of a current layer;
calculating a loss function value; specifically calculated by the following formula:
wherein loss is a loss function value, y is an actual output value, o is a floating point number between 0 and 1, i is a node of the current layer, and m is the total number of nodes of the current layer;
calculating a back propagation value; specifically, it is calculated by the following formula:
wherein, ω isi+1Node weight, ω, for level i +1iIs the node weight of the current layer, eta is the coefficient, bi+1Is the bias value of the i +1 layer, biIs the offset value of the current layer i.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110551906.2A CN113299390A (en) | 2021-05-20 | 2021-05-20 | System and method for predicting in-hospital mortality of acute kidney injury patient |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110551906.2A CN113299390A (en) | 2021-05-20 | 2021-05-20 | System and method for predicting in-hospital mortality of acute kidney injury patient |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113299390A true CN113299390A (en) | 2021-08-24 |
Family
ID=77323133
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110551906.2A Pending CN113299390A (en) | 2021-05-20 | 2021-05-20 | System and method for predicting in-hospital mortality of acute kidney injury patient |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113299390A (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150213227A1 (en) * | 2012-08-31 | 2015-07-30 | Koninklijke Philips N.V. | The modeling techniques for predicting mortality in intensive care units |
CN110051324A (en) * | 2019-03-14 | 2019-07-26 | 深圳大学 | A kind of acute respiratory distress syndrome anticipated mortality method and system |
CN111508604A (en) * | 2020-04-20 | 2020-08-07 | 深圳大学 | Acute kidney injury patient mortality prediction method, server and storage medium |
-
2021
- 2021-05-20 CN CN202110551906.2A patent/CN113299390A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150213227A1 (en) * | 2012-08-31 | 2015-07-30 | Koninklijke Philips N.V. | The modeling techniques for predicting mortality in intensive care units |
CN110051324A (en) * | 2019-03-14 | 2019-07-26 | 深圳大学 | A kind of acute respiratory distress syndrome anticipated mortality method and system |
CN111508604A (en) * | 2020-04-20 | 2020-08-07 | 深圳大学 | Acute kidney injury patient mortality prediction method, server and storage medium |
Non-Patent Citations (2)
Title |
---|
刘瑞: "深度学习技术在重症监护室患者住院死亡风险预测中的应用", 《中国医学工程》 * |
刘经纬等: "《创新创业+新工科教学质量提升研究——零基础学会培养python人工智能》", 31 August 2020 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP3446260B1 (en) | Memory-efficient backpropagation through time | |
CN110276442B (en) | Searching method and device of neural network architecture | |
KR101953814B1 (en) | Analysis of health events using recurrent neural networks | |
CN111382255B (en) | Method, apparatus, device and medium for question-answering processing | |
CN111145076B (en) | Data parallelization processing method, system, equipment and storage medium | |
CN112686382B (en) | Convolution model lightweight method and system | |
US20190138929A1 (en) | System and method for automatic building of learning machines using learning machines | |
CN111160049B (en) | Text translation method, apparatus, machine translation system, and storage medium | |
CN115910325A (en) | Modeling method for cognitive task evaluation, cognitive task evaluation method and system | |
CN109979591B (en) | Method and device for analyzing plaque progress factor based on graph neural network | |
CN113299390A (en) | System and method for predicting in-hospital mortality of acute kidney injury patient | |
US10990525B2 (en) | Caching data in artificial neural network computations | |
CN116503608A (en) | Data distillation method based on artificial intelligence and related equipment | |
CN114238611B (en) | Method, apparatus, device and storage medium for outputting information | |
CN116109449A (en) | Data processing method and related equipment | |
CN114999599A (en) | Data processing method and device, electronic equipment and computer readable storage medium | |
CN113724017A (en) | Pricing method and device based on neural network, electronic equipment and storage medium | |
CN114298329A (en) | Model training method, device, equipment and storage medium | |
Shang et al. | Non parametric Bayesian analysis of the two-sample problem with censoring | |
CN113436689B (en) | Drug molecular structure prediction method, device, equipment and storage medium | |
EP3948685A1 (en) | Accelerating neuron computations in artificial neural networks by skipping bits | |
EP3895024A1 (en) | Caching data in artificial neural network computations | |
CN113111254B (en) | Training method, fitting method and device of recommendation model and electronic equipment | |
CN114792573B (en) | Drug combination effect prediction method, model training method, device and equipment | |
CN116089809B (en) | Financial characteristic data screening method and device, electronic equipment and storage medium |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210824 |