Cerebral hemorrhage patient discharge time prediction method, readable storage medium and equipment

Abstract

The invention discloses a method for predicting the discharge time of a cerebral hemorrhage patient, a readable storage medium and equipment, and belongs to the technical field of medical data. Detection multi-modal data of examination items of a target cerebral hemorrhage patient during hospitalization is input into a pre-trained hospitalization time prediction model; obtaining the hospitalization prediction time of the target cerebral hemorrhage patient; according to the admission time and the hospitalization prediction time of the target cerebral hemorrhage patient, the discharge time of the patient can be predicted, bed turnover arrangement of a hospital is facilitated, a doctor can conveniently control the medication cycle, and the problems that the hospitalization time of the current acute cerebral hemorrhage patient cannot be reasonably arranged, the hospitalization time is long, and the medication cycle is poor are solved. And the medication cycle is unreasonable.

Description

Technical Field

[0001] The present invention relates to a method for predicting the discharge time of a patient with cerebral hemorrhage, a readable storage medium and a device, and belongs to the technical field of medical data. Background Art

[0002] Acute intracerebral hemorrhage refers to primary non-traumatic rupture of intracerebral blood vessels, resulting in blood accumulation in the brain parenchyma. Acute intracerebral hemorrhage is a subtype of stroke, with an incidence of approximately 27.89% of all strokes and 44.04% of the global stroke mortality burden. It is the most devastating form of stroke, with a high risk of early deterioration and short-term death. Acute intracerebral hemorrhage is a neurological emergency associated with high mortality and morbidity. The benefits of treatment and care in specialized tertiary hospitals for patients with acute intracerebral hemorrhage are obvious. Treatment and care in tertiary hospitals can improve patients' functional outcomes and reduce mortality, length of hospital stay, and treatment costs, but long-term hospitalization of patients with acute intracerebral hemorrhage will lead to high resource utilization in tertiary hospitals and increased personal medical costs for patients. Summary of the invention

[0003] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method for predicting the discharge time of patients with cerebral hemorrhage, a readable storage medium and a device, which can accurately predict the discharge time of patients with acute cerebral hemorrhage, which is conducive to increasing the bed turnover rate of tertiary hospitals, saving medical expenses and resources, and solving the problem that the current acute cerebral hemorrhage patients cannot reasonably arrange their hospitalization time, resulting in long hospitalization time and unreasonable medication cycle.

[0004] In order to achieve the above purpose / solve the above technical problems, the present invention is implemented by adopting the following technical solutions: The present invention provides a method for predicting the discharge time of a patient with cerebral hemorrhage, comprising: Obtain the multimodal data of the target cerebral hemorrhage patients' admission time and the examination items performed during hospitalization; Inputting the detected multimodal data into a pre-trained hospitalization time prediction model to obtain the predicted hospitalization time of the target cerebral hemorrhage patient; According to the admission time and predicted hospitalization time of the target cerebral hemorrhage patient, the predicted discharge time of the target cerebral hemorrhage patient is obtained.

[0005] Furthermore, the training of the pre-trained hospital stay prediction model includes: Obtain historical clinical data of patients with cerebral hemorrhage and historical imaging data of examination items performed during hospitalization; Extracting features from the historical clinical data and historical imaging data to obtain historical detection multimodal data; Based on historical detection multimodal data, the target hospital stay prediction model is trained and a trained hospital stay prediction model is obtained.

[0006] Furthermore, the historical clinical structured data includes patient demographic information, vital signs, laboratory test results, and medical history information; The historical image data includes CT images and MRI images.

[0007] Furthermore, the feature extraction of the historical clinical data and the historical imaging data to obtain historical detection multimodal data includes: The historical clinical data and historical imaging data of patients with cerebral hemorrhage are associated, and the discharge time is used as a label to obtain a historical sample data set; Obtain historical sample data from a historical sample data set; Feature extraction is performed on historical sample data to obtain historical detection multimodal data.

[0008] Furthermore, the feature extraction of the historical sample data to obtain the historical detection multimodal data includes: The historical sample data is converted into a feature graph, where each feature represents a node in the feature graph, and the feature graph is represented as G (N, E, ), N represents the feature nodes in the graph, E represents the correlation of the feature nodes, which is regarded as the edge of the feature graph. Indicates the time point corresponding to the feature node; The Pearson coefficient is used to measure the intrinsic correlation between historical sample data, and the intrinsic correlation is calculated by the following formula: ; in: Represents two different feature nodes , The correlation between Represents a feature node The standard deviation of Represents a feature node The standard deviation of Represents a feature node x The expected value of Represents the expected value of feature node y; The weighted adjacency matrix is ​​calculated as follows: ; in: Represents the Pearson matrix P The absolute value of represents the identity matrix; Based on the weighted adjacency matrix, the degree matrix of the feature graph is calculated by the following formula: ; Represents a feature node i The degree, Represents a feature node i With feature nodes j The weighted adjacency matrix of ; Based on the degree matrix and weighted adjacency matrix of the feature graph, the normalized graph Laplacian operator and Chebyshev polynomial approximation are used to perform convolution operation on the feature graph to obtain historical detection multimodal data.

[0009] Furthermore, the training of the target hospital stay prediction model based on the historical detection multimodal data and obtaining the trained hospital stay prediction model includes: Build a target hospital stay prediction model based on the XGBoost decision tree algorithm; Training a target hospital stay prediction model through historical detection multimodal data, and updating model parameters of the target hospital stay prediction model until the model parameters of the target hospital stay prediction model are updated to optimal model parameters that meet the requirements; Based on the optimal model parameters, a trained hospital stay prediction model was obtained.

[0010] Furthermore, the target hospital stay prediction model is trained by using historical detection multimodal data, and the model parameters of the target hospital stay prediction model are updated until the model parameters of the target hospital stay prediction model are updated to optimal model parameters that meet the requirements, including: Based on historical detection multimodal data, a genetic algorithm is used to select hyperparameters for XGBoost regression. The steps are as follows: 1.1. Randomly generate N combinations in the hyperparameter space as the initialization population; 1.2. The loss function of XGBoost is obtained through cross-validation method to calculate the fitness of each individual in the population; 1.3. Keep the parameter combination with higher fitness; 1.4. Perform crossover operation on the retained individual parameters to generate new parameter combinations, and perform random mutation operation on the new parameter combinations; 1.5. Eliminate individuals with lower fitness in the new parameter combination; 1.6. Repeat steps 1.2 to 1.5 until the requirements are met; 1.7. Select the individual with the highest fitness among all parameter combinations as the optimal model parameters for the target hospital stay prediction model.

[0011] Furthermore, the target hospital stay prediction model is constructed based on the XGBoost decision tree algorithm, including: ; in: Indicatess The target hospital stay prediction model for round training; T Indicates the number of leaf nodes of the model; , Both represent the regularization term coefficient; Represents the loss function l ( ) s- 1 round of training leaf nodes j Training samples i Predicted value The first derivative of The sum of ; ; Loss function l ( ) s- 1 round of training leaf nodes j Training samples i Predicted value The second derivative of The sum of ; ; Indicates i training samples; Indicates i The label value of the training sample, Indicates i The predicted value of samples; , , Indicates j A sample subset of leaf nodes; Indicates j The output value of each leaf node; No. j The optimal output value of a leaf node is expressed as : ; After each round of training, the optimal split point of each leaf node of the tree is : ; , , ; , , ; represents the sample set of the left leaf node, Represents the sample set of the right leaf node.

[0012] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for predicting the discharge time of patients with cerebral hemorrhage as described above is implemented.

[0013] A computer device comprising: A memory for storing instructions; The processor is used to execute the instructions so that the device implements the above-mentioned method for predicting the discharge time of patients with cerebral hemorrhage.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention inputs the detection multimodal data of the examination items performed on the target cerebral hemorrhage patient during hospitalization into a pre-trained hospitalization time prediction model to obtain the predicted hospitalization time of the target cerebral hemorrhage patient; according to the admission time and predicted hospitalization time of the target cerebral hemorrhage patient, the discharge time of the patient can be predicted, which is beneficial to the hospital's bed turnover arrangement and also facilitates the doctor's control of the medication cycle, solving the problem that the current acute cerebral hemorrhage patients cannot reasonably arrange the hospitalization time, resulting in long hospitalization time and unreasonable medication cycle.

[0015] 2. The present invention associates the historical clinical data and historical imaging data of patients with cerebral hemorrhage, uses discharge time as a label, obtains a historical sample data set, converts the historical sample data in the sample data set into a feature graph, uses the Pearson coefficient to measure the intrinsic correlation between the historical sample data, and based on the degree matrix and weighted adjacency matrix of the feature graph, uses the normalized graph Laplace operator and Chebyshev polynomial approximation to perform convolution operations on the feature graph, thereby realizing the mining of data space features and obtaining historical detection multimodal data.

[0016] 3. The present invention uses the XGBoost decision tree algorithm to construct a target hospital stay prediction model, and uses a genetic algorithm to select hyperparameters for XGBoost regression, thereby reducing the time for model training and improving the model accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of a method for predicting discharge time of a patient with cerebral hemorrhage according to an embodiment of the present invention; Figure 2 It is a training flow chart of a pre-trained hospitalization time prediction model in a method for predicting discharge time of patients with cerebral hemorrhage in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention. Example

[0019] like Figure 1 As shown, a method for predicting the discharge time of patients with cerebral hemorrhage comprises: Obtain the multimodal data of the target cerebral hemorrhage patients' admission time and the examination items performed during hospitalization; Input the detection multimodal data into the pre-trained hospitalization time prediction model to obtain the predicted hospitalization time of the target cerebral hemorrhage patient; According to the admission time and predicted hospitalization time of the target cerebral hemorrhage patient, the predicted discharge time of the target cerebral hemorrhage patient is obtained.

[0020] like Figure 2 As shown, the training of the pre-trained hospital stay prediction model includes: Obtain historical clinical data of patients with cerebral hemorrhage and historical imaging data of examination items performed during hospitalization, where historical clinical structural data include patient demographic information, vital signs, laboratory test results, and medical history information; historical imaging data include CT images and MRI images; Feature extraction is performed on historical clinical data and historical imaging data to obtain historical detection multimodal data, including: The historical clinical data and historical imaging data of patients with cerebral hemorrhage are associated, and the discharge time is used as a label to obtain a historical sample data set; Obtain historical sample data from a historical sample data set; Perform feature extraction on historical sample data to obtain historical detection multimodal data, specifically: The historical sample data is converted into a feature graph, where each feature represents a node in the feature graph, and the feature graph is represented as G (N, E, ), N represents the feature nodes in the graph, E represents the correlation of the feature nodes, which is regarded as the edge of the feature graph. Indicates the time point corresponding to the feature node; The Pearson coefficient is used to measure the intrinsic correlation between historical sample data, and the intrinsic correlation is calculated by the following formula: ; in: Represents two different feature nodes , The correlation between Represents a feature node The standard deviation of Represents a feature node The standard deviation of Represents a feature node x The expected value of Represents a feature node Expected value of The weighted adjacency matrix is ​​calculated as follows: ; in: Represents the Pearson matrix P The absolute value of represents the identity matrix; Based on the weighted adjacency matrix, the degree matrix of the feature graph is calculated by the following formula: ; Represents a feature node i The degree, Represents a feature node i With feature nodes j The weighted adjacency matrix of ; Based on the degree matrix and weighted adjacency matrix of the feature graph, the normalized graph Laplacian operator and Chebyshev polynomial approximation are used to perform convolution operation on the feature graph to obtain historical detection multimodal data.

[0021] Based on historical detection multimodal data, the target hospital stay prediction model is trained and a trained hospital stay prediction model is obtained, including: The target hospital stay prediction model is constructed based on the XGBoost decision tree algorithm. Specifically: ; in: Indicates s The target hospital stay prediction model for round training; T Indicates the number of leaf nodes of the model; , Both represent the regularization term coefficient; Represents the loss function l ( ) s- 1 round of training leaf nodes j Training samples i Predicted value The first derivative of The sum of ; ; Loss function l ( ) s- 1 round of training leaf nodes j Training samples i Predicted value The second derivative of The sum of ; ; Indicates itraining samples; Indicates i The label value of the training sample, Indicates i The predicted value of samples; , , Indicates j A sample subset of leaf nodes; Indicates j The output value of each leaf node; No. j The optimal output value of a leaf node is expressed as : ; After each round of training, the optimal split point of each leaf node of the tree is : ; , , ; , , ; represents the sample set of the left leaf node, Represents the sample set of the right leaf node; The target hospital stay prediction model is updated by weighting the newly obtained tree model in this round to the model in the previous round: ; Represents the weighting coefficient.

[0022] The target hospital stay prediction model is trained through historical detection multimodal data, and the model parameters of the target hospital stay prediction model are updated until the model parameters of the target hospital stay prediction model are updated to the optimal model parameters that meet the requirements. Specifically: Based on historical detection multimodal data, a genetic algorithm is used to select hyperparameters for XGBoost regression. The steps are as follows: 1.1. Randomly generate N combinations in the hyperparameter space as the initialization population; 1.2. The loss function of XGBoost is obtained through cross-validation method to calculate the fitness of each individual in the population; 1.3. Keep the parameter combination with higher fitness; 1.4. Perform crossover operation on the retained individual parameters to generate new parameter combinations, and perform random mutation operation on the new parameter combinations; 1.5. Eliminate individuals with lower fitness in the new parameter combination; 1.6. Repeat steps 1.2 to 1.5 until the requirements are met; 1.7. Select the individual with the highest fitness among all parameter combinations as the optimal model parameter for the target hospital stay prediction model; Based on the optimal model parameters, a trained hospital stay prediction model was obtained.

[0023] The hyperparameters of XGBoost are shown in Table 1: Table 1

[0024] In order to verify the superiority of the method proposed in this embodiment, an existing machine learning method is used for comparison, wherein the existing machine learning method used is as follows: XGB, RF, SVM, DT, KNN, LDA and NB; The historical sample data set is shown in Table 2, where the historical clinical structural data selects patient demographic information, medical history information, vascular test results, and medication status; the historical imaging data selects the diagnostic results of CT images and MRI images; Table 2

[0025] The test results are shown in Table 3: Table 3 Example

[0026] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the following method: Obtain the multimodal data of the target cerebral hemorrhage patients' admission time and the examination items performed during hospitalization; Input the detection multimodal data into the pre-trained hospitalization time prediction model to obtain the predicted hospitalization time of the target cerebral hemorrhage patient; According to the admission time and predicted hospitalization time of the target cerebral hemorrhage patient, the predicted discharge time of the target cerebral hemorrhage patient is obtained. Example

[0027] A computer device comprising: A memory for storing instructions; The processor is used to execute instructions so that the device performs the steps of implementing the following method: Obtain the multimodal data of the target cerebral hemorrhage patients' admission time and the examination items performed during hospitalization; Input the detection multimodal data into the pre-trained hospitalization time prediction model to obtain the predicted hospitalization time of the target cerebral hemorrhage patient; According to the admission time and predicted hospitalization time of the target cerebral hemorrhage patient, the predicted discharge time of the target cerebral hemorrhage patient is obtained.

[0028] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0029] The present application is described with reference to the flowcharts of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process in the flowchart can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A process or multiple processes or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0030] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A function specified in a process or multiple processes.

[0031] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 The steps of a specified function in a process or multiple processes.

[0032] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A method for predicting discharge time of patients with cerebral hemorrhage, characterized in that: include: Obtain the multimodal data of the target cerebral hemorrhage patients' admission time and the examination items performed during hospitalization; Inputting the detected multimodal data into a pre-trained hospitalization time prediction model to obtain the predicted hospitalization time of the target cerebral hemorrhage patient; According to the admission time and predicted hospitalization time of the target cerebral hemorrhage patient, the predicted discharge time of the target cerebral hemorrhage patient is obtained.

2. The method for predicting discharge time of patients with brain hemorrhage according to claim 1, characterized in that: The training of the pre-trained hospital stay prediction model includes: Obtain historical clinical data of patients with cerebral hemorrhage and historical imaging data of examination items performed during hospitalization; Extracting features from the historical clinical data and historical imaging data to obtain historical detection multimodal data; Based on historical detection multimodal data, the target hospital stay prediction model is trained and a trained hospital stay prediction model is obtained.

3. The method for predicting discharge time of patients with brain hemorrhage according to claim 2, characterized in that: The historical clinical structured data include patient demographic information, vital signs, laboratory test results and medical history information; the historical imaging data include CT images and MRI images.

4. The method for predicting discharge time of patients with brain hemorrhage according to claim 2, characterized in that: The feature extraction of the historical clinical data and the historical imaging data to obtain historical detection multimodal data includes: The historical clinical data and historical imaging data of patients with cerebral hemorrhage are associated, and the discharge time is used as a label to obtain a historical sample data set; Obtain historical sample data from a historical sample data set; Feature extraction is performed on historical sample data to obtain historical detection multimodal data.

5. The method for predicting discharge time of patients with brain hemorrhage according to claim 4, characterized in that: The feature extraction of the historical sample data to obtain the historical detection multimodal data includes: The historical sample data is converted into a feature graph, where each feature represents a node in the feature graph, and the feature graph is represented as G (N, E, ), N represents the feature nodes in the graph, E represents the correlation of the feature nodes, which is regarded as the edge of the feature graph. Indicates the time point corresponding to the feature node; The Pearson coefficient is used to measure the intrinsic correlation between historical sample data, and the intrinsic correlation is calculated by the following formula: ； in: Represents two different feature nodes , The correlation between Represents a feature node The standard deviation of Represents a feature node The standard deviation of Represents a feature node x The expected value of Represents a feature node Expected value of The weighted adjacency matrix is ​​calculated as follows: ； in: Represents the Pearson matrix P The absolute value of represents the identity matrix; Based on the weighted adjacency matrix, the degree matrix of the feature graph is calculated by the following formula: ； Represents a feature node i The degree, Represents a feature node i With feature nodes j The weighted adjacency matrix of ; Based on the degree matrix and weighted adjacency matrix of the feature graph, the normalized graph Laplacian operator and Chebyshev polynomial approximation are used to perform convolution operation on the feature graph to obtain historical detection multimodal data.

6. The method for predicting discharge time of patients with brain hemorrhage according to claim 4, characterized in that: The method of training a target hospital stay prediction model based on historical detection multimodal data and obtaining a trained hospital stay prediction model comprises: Build a target hospital stay prediction model based on the XGBoost decision tree algorithm; Training a target hospital stay prediction model through historical detection multimodal data, and updating model parameters of the target hospital stay prediction model until the model parameters of the target hospital stay prediction model are updated to optimal model parameters that meet the requirements; Based on the optimal model parameters, a trained hospital stay prediction model was obtained.

7. The method for predicting discharge time of patients with brain hemorrhage according to claim 6, characterized in that: The target hospitalization time prediction model is trained by using historical detection multimodal data, and the model parameters of the target hospitalization time prediction model are updated until the model parameters of the target hospitalization time prediction model are updated to optimal model parameters that meet the requirements, including: Based on historical detection multimodal data, a genetic algorithm is used to select hyperparameters for XGBoost regression. The steps are as follows: 1.

1. Randomly generate N combinations in the hyperparameter space as the initialization population; 1.

2. The fitness of each individual in the population is calculated by using the loss function of XGBoost obtained by cross-validation method; 1.

3. Keep the parameter combination with higher fitness; 1.

4. Perform crossover operation on the retained individual parameters to generate new parameter combinations, and perform random mutation operation on the new parameter combinations; 1.

5. Eliminate individuals with lower fitness in the new parameter combination; 1.

6. Repeat steps 1.2 to 1.5 until the requirements are met; 1.

7. Select the individual with the highest fitness among all parameter combinations as the optimal model parameters for the target hospital stay prediction model.

8. The method for predicting discharge time of patients with brain hemorrhage according to claim 6, characterized in that: Said The target hospital stay prediction model was constructed based on the XGBoost decision tree algorithm, including: ； in: Indicates s The target hospital stay prediction model for round training; T Indicates the number of leaf nodes of the model; , Both represent the regularization term coefficient; Represents the loss function l ( ) s- 1 round of training leaf nodes j Training samples i Predicted value The first derivative of The sum of ; ; Loss function l ( ) s- 1 round of training leaf nodes j Training samples i Predicted value The second derivative of The sum of ; ; Indicates i training samples; Indicates i The label value of the training sample, Indicates i The predicted value of samples; , , Indicates j A sample subset of leaf nodes; Indicates j The output value of each leaf node; No. j The optimal output value of a leaf node is expressed as : ; After each round of training, the optimal split point of each leaf node of the tree is : ； , , ; , , ; represents the sample set of the left leaf node, Represents the sample set of the right leaf node.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting the discharge time of a patient with cerebral hemorrhage as claimed in any one of claims 1 to 8 is implemented.

10. A computer device, characterized in that: include: A memory for storing instructions; A processor is used to execute the instructions so that the device implements the method for predicting the discharge time of a patient with cerebral hemorrhage as described in any one of claims 1 to 8.

Images