CN118760973A - A method for predicting intensive care needs of patients with cerebral hemorrhage based on multimodal fusion - Google Patents

Abstract

The present invention discloses a method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion, which relates to the technical field of predicting the intensive care needs of patients with cerebral hemorrhage, including collecting multimodal data of patients with cerebral hemorrhage, preprocessing the multimodal data; extracting features from the preprocessed multimodal data, using undersampling and oversampling techniques to handle the imbalance problem of the data set; constructing a multimodal fusion monitoring demand prediction model, predicting real-time clinical data, and identifying the intensive care needs of patients with cerebral hemorrhage. The method of the present invention improves the adaptability of the model to complex and unbalanced data through feature extraction and data balancing processing, improves the accuracy and reliability of the prediction, and improves the accuracy and real-time nature of the prediction by constructing a multimodal fusion monitoring demand prediction model, ensuring that patients can receive intensive care in a timely manner when their condition worsens, reducing the mortality and disability rates of patients, and improving the overall quality of medical services.

Description

A method for predicting intensive care needs of patients with cerebral hemorrhage based on multimodal fusion

Technical Field

The present invention relates to the technical field of intensive care demand prediction for patients with cerebral hemorrhage, and in particular to a method for predicting intensive care demand for patients with cerebral hemorrhage based on multimodal fusion.

Background Art

With the rapid development of medical imaging technology, data mining technology and artificial intelligence technology, medical diagnosis and treatment have gradually entered the data-driven era. Cerebral hemorrhage is a serious neurological disease with high mortality and disability rates. Timely and effective monitoring and treatment are crucial. Traditional cerebral hemorrhage monitoring mainly relies on the experience of clinicians and a single medical imaging method, such as CT (computed tomography) and MRI (magnetic resonance imaging). However, a single imaging method is difficult to fully reflect the patient's condition, especially for patients with complex cerebral hemorrhage. Comprehensive multimodal data analysis is particularly important.

Although multimodal data fusion and artificial intelligence technology have made progress in the medical field, the existing technology still has many shortcomings in predicting the intensive care needs of patients with cerebral hemorrhage. Most of the existing multimodal data fusion methods are limited to the fusion of static data and lack the ability to dynamically analyze time series data. The condition of patients with cerebral hemorrhage changes rapidly and requires real-time monitoring and prediction, while traditional multimodal fusion models are difficult to capture such dynamic changes. When dealing with data imbalance problems, existing multimodal fusion models often use simple undersampling or oversampling methods. Although these methods can alleviate the problem of data imbalance to a certain extent, they are prone to overfitting or underfitting of the model, affecting the accuracy and robustness of the prediction. Especially in the prediction of intensive care needs, minority class (patients requiring intensive care) samples are often more important, and simple sampling methods are difficult to effectively improve the model's recognition ability for minority classes. In addition, the existing multimodal fusion technology has shortcomings in feature extraction. Traditional feature extraction methods mostly rely on manual experience and are difficult to fully mine the deep-level features in the data. Although the automatic feature extraction method based on deep learning can extract high-dimensional features, it often lacks interpretability and is difficult to be accepted and trusted by clinicians.

Summary of the invention

In view of the above-mentioned problems, the present invention is proposed.

Therefore, the technical problem solved by the present invention is: the existing methods for predicting the intensive care needs of patients with cerebral hemorrhage have low accuracy, low efficiency, and low reliability, and how to integrate multimodal data to perform monitoring needs analysis for patients with complex cerebral hemorrhage conditions.

To solve the above technical problems, the present invention provides the following technical solutions: a method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion, comprising collecting multimodal data of patients with cerebral hemorrhage and preprocessing the multimodal data; extracting features from the preprocessed multimodal data, and using undersampling and oversampling techniques to deal with the imbalance problem of the data set; constructing a multimodal fusion monitoring demand prediction model, predicting real-time clinical data, and identifying the intensive care needs of patients with cerebral hemorrhage.

As a preferred solution of the method for predicting the intensive care demand of patients with cerebral hemorrhage based on multimodal fusion described in the present invention, the multimodal data of the patients with cerebral hemorrhage include clinical data and imaging data; the clinical data include patient demographic information, vital signs, laboratory test results, and medical history information; the imaging data include CT images, MRI images, and ultrasound images.

As a preferred solution of the method for predicting the intensive care demand of patients with cerebral hemorrhage based on multimodal fusion described in the present invention, the preprocessing includes data cleaning, data integration, and data conversion of the multimodal data of patients with cerebral hemorrhage; data cleaning includes re-examining and verifying the data, checking data consistency, deleting invalid values and filling missing values; data integration includes integrating data from different sources and formats, and unifying the data into the same data set through the patient's unique identifier; data conversion includes normalizing the data.

As a preferred solution of the method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion described in the present invention, the feature extraction includes performing nonlinear feature extraction based on the preprocessed multimodal data to adapt to the dynamic changes of the condition of patients with cerebral hemorrhage over time, which is expressed as:

Wherein, _gj (X(t)) represents the j-th nonlinear feature extraction function, which is used to extract nonlinear features from the multimodal data X(t) at time t, X'(t) is the multimodal dataset at time t, which contains the clinical data and imaging data of patients with cerebral hemorrhage at different time points, q is the total number of data points, _bjl is the nonlinear weight coefficient between the j-th feature extraction function and the l-th data point, and _Xl (t) is the l-th data point in the multimodal dataset at time t.

As a preferred solution of the method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion described in the present invention, the problem of processing the imbalanced data set includes adjusting the recognition ability of the minority class based on the oversampling adjustment model, undersampling the distribution of the balanced data set, and obtaining the balanced data set, which is expressed as:

Where D _balanced is the balanced data set, which represents the multimodal data set of patients with cerebral hemorrhage after undersampling and oversampling techniques. n is the total number of samples in the data set. _Xi represents the i-th sample in the data set. X represents the multimodal data set, including clinical data and imaging data. d( _Xi , X) represents the distance function, which calculates the distance between the i-th sample and the entire data set X. λ represents the adjustment parameter used to control the amplitude of data synthesis. NN(X(t)) represents the nearest neighbor sample function at time t, which is expressed as:

Among them, X' represents the nearest neighbor sample, and the distance between the i-th sample and the entire data set X is expressed as:

Among them, Xij _' represents the j'th eigenvalue of the i-th sample in the data set, Xj _' represents the j'th feature in the data set, and m represents the number of features contained in each sample vector.

As a preferred solution of the method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion described in the present invention, the construction of a multimodal fusion monitoring needs prediction model includes performing multimodal feature fusion based on the features of the extracted clinical data and imaging data, which is expressed as:

Among them, Φ _fusion is the fusion feature set, n _c represents the number of clinical features, α _c represents the weight coefficient of the cth clinical feature, T represents the time period, Φ _clinical,c represents the value of the cth clinical feature changing with time t, n _k represents the number of imaging features, β _k represents the weight coefficient of the kth imaging feature, and Φ _imaging,k represents the value of the kth imaging feature changing with time t. The features after multimodal feature fusion are taken as input, and the feature data are weighted by the time-varying weight matrix to construct a multimodal fusion monitoring demand prediction model, which is expressed as:

in, is the prediction result, W(t) is the weight matrix at time t, which is adjusted based on the time transformation, _Xtarget (t) represents the multimodal data of the target field at time t, including the vector or matrix of clinical features and imaging features, _Wt (t) has the same dimension as _Xtarget (t), and _ftarget represents the prediction function of the target field, including the logistic regression function.

As a preferred solution of the method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion described in the present invention, wherein: the identification of the intensive care needs of patients with cerebral hemorrhage includes predicting the real-time clinical data based on the prediction results of the multimodal fusion monitoring needs prediction model to identify the intensive care needs of patients with cerebral hemorrhage; when the prediction results When the predicted result is greater than or equal to 0 and less than or equal to 0.3, the patient with cerebral hemorrhage will not be placed in intensive care, but will be placed in a general ward for monitoring and undergo routine medical examinations and care. When the value is greater than 0.3 and less than or equal to 0.7, the patient with cerebral hemorrhage needs intensive care. The patient will be further evaluated and closely monitored, and additional diagnostic tests will be performed to determine whether he needs to be transferred to the intensive care unit. The patient has a plan for priority access to intensive care resources. When the predicted result is When it is greater than 0.7 and less than or equal to 1, patients with cerebral hemorrhage must receive intensive care and be immediately transferred to the intensive care unit for close monitoring and treatment to prevent and deal with potential serious complications.

Another object of the present invention is to provide a system for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion, which can predict real-time clinical data and identify the intensive care needs of patients with cerebral hemorrhage by constructing a multimodal fusion monitoring needs prediction model, thereby solving the problem of low reliability in the current prediction of the intensive care needs of patients with cerebral hemorrhage.

As a preferred solution of the intensive care demand prediction system for cerebral hemorrhage patients based on multimodal fusion described in the present invention, it includes: a data processing module, a feature extraction module, and a demand prediction module; the data processing module is used to collect multimodal data of cerebral hemorrhage patients and preprocess the multimodal data; the feature extraction module is used to extract features from the preprocessed multimodal data, and use undersampling and oversampling techniques to deal with the imbalance problem of data sets; the demand prediction module is used to construct a multimodal fusion monitoring demand prediction model, predict real-time clinical data, and identify the intensive care needs of cerebral hemorrhage patients.

A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a method for predicting intensive care requirements of patients with cerebral hemorrhage based on multimodal fusion.

A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for predicting intensive care requirements of patients with cerebral hemorrhage based on multimodal fusion.

Beneficial effects of the present invention: The method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion provided by the present invention reduces the prediction error caused by data problems through data preprocessing, improves the robustness and accuracy of the model, and lays a solid foundation for subsequent feature extraction and prediction model construction. By extracting features from the preprocessed multimodal data and using undersampling and oversampling techniques to deal with the imbalance problem of the data set, deep mining of potentially useful information in the data and balanced data distribution are achieved. Through feature extraction and data balancing processing, the adaptability of the model to complex and unbalanced data is improved, so that the prediction model can more accurately identify patients who need intensive care, thereby improving the accuracy and reliability of the prediction. By constructing a multimodal fusion monitoring demand prediction model, real-time prediction of the intensive care needs of patients with cerebral hemorrhage is achieved, the accuracy and real-time of the prediction are improved, and it is ensured that patients can receive intensive care in time when their condition worsens, reducing the mortality and disability rates of patients and improving the overall quality of medical services. The present invention achieves better results in terms of accuracy, reliability and real-time performance.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

FIG1 is an overall flow chart of a method for predicting intensive care needs of patients with cerebral hemorrhage based on multimodal fusion provided in a first embodiment of the present invention.

FIG2 is an overall flow chart of a system for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion according to a third embodiment of the present invention.

DETAILED DESCRIPTION

In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

Example 1

Referring to FIG. 1 , an embodiment of the present invention provides a method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion, including:

S1: Collect multimodal data of patients with cerebral hemorrhage and preprocess the multimodal data.

Furthermore, the multimodal data of patients with cerebral hemorrhage include clinical data and imaging data; clinical data include patient demographic information (such as age, gender, weight, height), vital signs (such as blood pressure, heart rate, respiratory rate), laboratory test results (such as blood sugar, blood lipids, liver and kidney function indicators), and medical history information (such as history of hypertension, diabetes, and heart disease); imaging data include CT images, MRI images, and ultrasound images.

It should be noted that preprocessing includes data cleaning, data integration, and data conversion of multimodal data of patients with cerebral hemorrhage; data cleaning includes re-examination and verification of data, checking data consistency, deleting invalid values and filling missing values; data integration includes integrating data from different sources and formats, and unifying the data into the same data set through the patient's unique identifier; data conversion includes normalizing the data.

It should also be noted that data cleaning improves the quality and reliability of data by re-examining and verifying data, checking data consistency, deleting invalid values and filling missing values. Invalid data and missing values may cause errors and deviations in the model training process. Through data cleaning, the accuracy and completeness of model training data are guaranteed, thereby improving the accuracy of the predictive model. Data integration integrates data from different sources and formats and uses the patient's unique identifier to unify the data into the same data set. This can eliminate differences between different data sources, form a comprehensive and consistent patient data view, and improve data integrity and availability. Data conversion normalizes the data so that data of different dimensions can be compared and analyzed on the same scale, reducing errors caused by dimensional differences and improving data comparability and the stability of model training.

S2: Perform feature extraction on the preprocessed multimodal data and use undersampling and oversampling techniques to deal with the imbalanced data set problem.

Furthermore, feature extraction includes nonlinear feature extraction based on preprocessed multimodal data to adapt to the dynamic changes of the condition of patients with cerebral hemorrhage over time, which can be expressed as:

Wherein, _gj (X(t)) represents the j-th nonlinear feature extraction function, which is used to extract nonlinear features from the multimodal data X(t) at time t, X'(t) is the multimodal dataset at time t, which contains the clinical data and imaging data of patients with cerebral hemorrhage at different time points, q is the total number of data points, _bjl is the nonlinear weight coefficient between the j-th feature extraction function and the l-th data point, and _Xl (t) is the l-th data point in the multimodal dataset at time t.

It should be noted that dealing with the imbalanced data set problem includes adjusting the recognition ability of the model for the minority class based on oversampling, balancing the distribution of the data set by undersampling, and obtaining the balanced data set, which is expressed as:

Where D _balanced is the balanced data set, which represents the multimodal data set of patients with cerebral hemorrhage after undersampling and oversampling techniques. n is the total number of samples in the data set. _Xi represents the i-th sample in the data set. X represents the multimodal data set, including clinical data and imaging data. d( _Xi , X) represents the distance function, which calculates the distance between the i-th sample and the entire data set X. λ represents the adjustment parameter used to control the amplitude of data synthesis. NN(X(t)) represents the nearest neighbor sample function at time t, which is expressed as:

Among them, X' represents the nearest neighbor sample, and the distance between the i-th sample and the entire data set X is expressed as:

Among them, Xij _' represents the j'th eigenvalue of the i-th sample in the data set, Xj _' represents the j'th feature in the data set, and m represents the number of features contained in each sample vector.

It should also be noted that feature extraction improves the model's ability to capture complex conditions by adapting to the dynamic changes in the condition of patients with cerebral hemorrhage over time and performing nonlinear feature extraction, thereby extracting highly informative nonlinear features from time series data and enhancing the model's prediction accuracy. Data imbalance processing uses undersampling and oversampling techniques to improve the model's ability to recognize minority samples.

S3: Build a multimodal fusion monitoring demand prediction model to predict real-time clinical data and identify the intensive care needs of patients with cerebral hemorrhage.

Furthermore, constructing a multimodal fusion monitoring demand prediction model includes performing multimodal feature fusion based on the features of the extracted clinical data and imaging data, which is expressed as:

Among them, Φ _fusion is the fusion feature set, n _c represents the number of clinical features, α _c represents the weight coefficient of the cth clinical feature, T represents the time period, Φ _clinical,c represents the value of the cth clinical feature changing with time t, n _k represents the number of imaging features, β _k represents the weight coefficient of the kth imaging feature, and Φ _imaging,k represents the value of the kth imaging feature changing with time t. The features after multimodal feature fusion are taken as input, and the feature data are weighted by the time-varying weight matrix to construct a multimodal fusion monitoring demand prediction model, which is expressed as:

in, is the prediction result, W(t) is the weight matrix at time t, which is adjusted based on the time transformation, _Xtarget (t) represents the multimodal data of the target field at time t, including the vector or matrix of clinical features and imaging features, _Wt (t) has the same dimension as _Xtarget (t), and _ftarget represents the prediction function of the target field, including the logistic regression function.

It should be noted that identifying the intensive care needs of patients with cerebral hemorrhage includes predicting the real-time clinical data based on the prediction results of the multimodal fusion monitoring needs prediction model to identify the intensive care needs of patients with cerebral hemorrhage; when the prediction results When the predicted result is greater than or equal to 0 and less than or equal to 0.3, the patient with cerebral hemorrhage will not be placed in intensive care, but will be placed in a general ward for monitoring and undergo routine medical examinations and care. When the value is greater than 0.3 and less than or equal to 0.7, the patient with cerebral hemorrhage needs intensive care. The patient will be further evaluated and closely monitored, and additional diagnostic tests will be performed to determine whether he needs to be transferred to the intensive care unit. The patient has a plan for priority access to intensive care resources. When the predicted result is When it is greater than 0.7 and less than or equal to 1, patients with cerebral hemorrhage must receive intensive care and be immediately transferred to the intensive care unit for close monitoring and treatment to prevent and deal with potential serious complications.

It should also be noted that dynamic weighted processing ensures that the model can adjust the prediction results in time when facing constantly changing disease data, provide accurate real-time decision support for medical staff, extract nonlinear features in multimodal data, and capture complex disease change patterns. It is particularly effective in predicting the condition of patients with cerebral hemorrhage, because cerebral hemorrhage often has complex dynamic change characteristics. Nonlinear feature extraction can provide higher prediction accuracy. By combining undersampling and oversampling techniques, the distribution of the data set is balanced, and the model's recognition ability for minority samples (such as patients requiring intensive care) is improved. By adjusting the combination of parameters and nearest neighbor sample functions, the flexibility and adaptability of data synthesis are enhanced, avoiding the problems of overfitting or underfitting in traditional methods.

Example 2

One embodiment of the present invention provides a method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

First, 10 patients with cerebral hemorrhage were selected, and their multimodal data, including clinical data and imaging data, were collected. Multimodal data of each patient with cerebral hemorrhage, including clinical data (demographic information, vital signs, laboratory test results, medical history information) and imaging data (CT images, MRI images, ultrasound images), were collected, the collected data were cleaned, the consistency of the data was checked, invalid values were deleted, and missing values were filled. For the blood pressure data in the vital signs, its range was checked to see if it was reasonable. For the missing test result data, the mean was filled. Data from different sources and formats were integrated together and unified into the same data set through the patient's unique identifier. CT, MRI and clinical data were integrated to form a complete data record, and the integrated data were normalized so that data of different dimensions could be analyzed on the same scale. , normalize blood sugar, blood pressure and other data to the range of [0,1], based on the preprocessed multimodal data, adapt to the dynamic changes of the condition of patients with cerebral hemorrhage over time, perform nonlinear feature extraction, use undersampling and oversampling techniques to deal with the imbalance of the data set, oversampling adjusts the model's recognition ability for the minority class (patients who need intensive care), undersampling balances the distribution of the data set, obtains the balanced data set, and based on the extracted features of the clinical data and imaging data, performs multimodal feature fusion, takes the features after multimodal feature fusion as input, and weights the feature data through the time-varying weight matrix to construct a multimodal fusion monitoring demand prediction model. According to the prediction results, identify the intensive care needs of patients with cerebral hemorrhage, and take corresponding medical measures according to different prediction values. Refer to Table 1 to record and analyze the experimental data.

Table 1 Experimental data record table

The table shows the clinical data and imaging data of different patients, including age, gender, blood pressure, blood sugar, CT abnormality score and MRI abnormality score. Through the multimodal fusion model, these multi-source data can be integrated to achieve a more accurate prediction of intensive care needs. For example, the prediction result for patient 1 is 0.72, and the prediction result for patient 2 is 0.85. These results reflect the model's accurate assessment of the conditions of different patients. Our invented model can process multimodal data that changes over time, capture the dynamic and imaging data changes of the patient's condition, and adjust the prediction results in time to provide real-time intensive care needs recommendations. Undersampling and oversampling techniques are used to deal with data imbalance problems, so that the model can still maintain a high recognition rate when facing minority samples (patients requiring intensive care). For example, in the table Several patients who required intensive care were displayed, and the prediction results were all above 0.7, indicating that the model had a high recognition ability for critically ill patients. Through the nonlinear feature extraction method, key features were extracted from complex multimodal data, which improved the prediction performance of the model. For example, the prediction result of patient 4 was 0.90, combined with his high CT and MRI abnormality scores, indicating that the model was able to accurately extract and utilize these nonlinear features for prediction. Compared with the prior art, the multimodal fusion monitoring demand prediction method of the present invention is innovative and advantageous. Traditional methods mostly rely on a single data source and cannot fully utilize the advantages of multimodal data. The present invention improves the accuracy and stability of prediction by fusing multi-source data. In addition, the application of technology to deal with data imbalance problems makes the model more robust and practical in practical applications.

Example 3

2 , which is an embodiment of the present invention, provides a system for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion, including a data processing module, a feature extraction module, and a demand prediction module.

The data processing module is used to collect multimodal data of patients with cerebral hemorrhage and preprocess the multimodal data; the feature extraction module is used to extract features from the preprocessed multimodal data and use undersampling and oversampling techniques to deal with the imbalance problem of data sets; the demand prediction module is used to build a multimodal fusion monitoring demand prediction model, predict real-time clinical data, and identify the intensive care needs of patients with cerebral hemorrhage.

If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

The logic and/or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims (10)

1. A method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion, characterized by comprising: Collect multimodal data of patients with cerebral hemorrhage and preprocess the multimodal data; Perform feature extraction on the preprocessed multimodal data and use undersampling and oversampling techniques to handle the imbalanced data set problem; A multimodal fusion monitoring demand prediction model was constructed to predict real-time clinical data and identify the intensive care needs of patients with cerebral hemorrhage. 2. The method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion as claimed in claim 1, characterized in that: the multimodal data of the patients with cerebral hemorrhage include clinical data and imaging data; Clinical data included patient demographic information, vital signs, laboratory test results, and medical history information; Image data includes CT images, MRI images, and ultrasound images. 3. The method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion as claimed in claim 2, characterized in that: the preprocessing includes data cleaning, data integration, and data conversion of the multimodal data of patients with cerebral hemorrhage; Data cleaning includes re-examining and verifying the data, checking data consistency, deleting invalid values and filling missing values; Data integration involves bringing together data from different sources and formats into a single dataset using a patient’s unique identifier; Data transformation includes normalizing the data. 4. The method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion as claimed in claim 3 is characterized in that: the feature extraction includes performing nonlinear feature extraction based on the pre-processed multimodal data to adapt to the dynamic changes of the condition of patients with cerebral hemorrhage over time, expressed as: Wherein, _gj (X(t)) represents the j-th nonlinear feature extraction function, which is used to extract nonlinear features from the multimodal data X(t) at time t, X'(t) is the multimodal dataset at time t, which contains the clinical data and imaging data of patients with cerebral hemorrhage at different time points, q is the total number of data points, _bjl is the nonlinear weight coefficient between the j-th feature extraction function and the l-th data point, and _Xl (t) is the l-th data point in the multimodal dataset at time t. 5. The method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion as claimed in claim 4 is characterized in that: the processing of the imbalanced data set problem includes adjusting the recognition ability of the minority class based on the oversampling adjustment model, undersampling the distribution of the balanced data set, and obtaining the balanced data set, which is expressed as: Where D _balanced is the balanced data set, which represents the multimodal data set of patients with cerebral hemorrhage after undersampling and oversampling techniques. n is the total number of samples in the data set. _Xi represents the i-th sample in the data set. X represents the multimodal data set, including clinical data and imaging data. d( _Xi , X) represents the distance function, which calculates the distance between the i-th sample and the entire data set X. λ represents the adjustment parameter used to control the amplitude of data synthesis. NN(X(t)) represents the nearest neighbor sample function at time t, which is expressed as: Among them, X' represents the nearest neighbor sample, and the distance between the i-th sample and the entire data set X is expressed as: Among them, Xij _' represents the j'th eigenvalue of the i-th sample in the data set, Xj _' represents the j'th feature in the data set, and m represents the number of features contained in each sample vector. 6. The method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion as claimed in claim 5 is characterized in that: the construction of the multimodal fusion monitoring needs prediction model includes performing multimodal feature fusion based on the features of the extracted clinical data and imaging data, expressed as: Wherein, Φ _fusion is the fusion feature set, n _c represents the number of clinical features, α _c represents the weight coefficient of the c-th clinical feature, T represents the time period, Φ _clinical,c represents the value of the c-th clinical feature changing with time t, _nk represents the number of imaging features, β _k represents the weight coefficient of the k-th imaging feature, and Φ _imaging,k represents the value of the k-th imaging feature changing with time t; The features after multimodal feature fusion are used as input, and the feature data is weighted by the time-varying weight matrix to construct a multimodal fusion monitoring demand prediction model, which is expressed as: in, is the prediction result, W(t) is the weight matrix at time t, which is adjusted based on the time transformation, _Xtarget (t) represents the multimodal data of the target field at time t, including the vector or matrix of clinical features and imaging features, _Wt (t) has the same dimension as _Xtarget (t), and _ftarget represents the prediction function of the target field, including the logistic regression function. 7. The method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion as claimed in claim 6, characterized in that: the identifying the intensive care needs of patients with cerebral hemorrhage comprises predicting the real-time clinical data based on the prediction results of the multimodal fusion monitoring needs prediction model to identify the intensive care needs of patients with cerebral hemorrhage; When the prediction results When it is greater than or equal to 0 and less than or equal to 0.3, patients with cerebral hemorrhage will not be placed in intensive care, but will be placed in a general ward for monitoring and receive routine medical examinations and care; When the prediction results When it is greater than 0.3 and less than or equal to 0.7, the patient with ICH has a need for intensive care. The patient will be further evaluated and closely monitored, and additional diagnostic tests will be performed to determine whether he needs to be transferred to the intensive care unit. The patient has a plan for priority access to intensive care resources. When the prediction results When it is greater than 0.7 and less than or equal to 1, patients with cerebral hemorrhage must receive intensive care and be immediately transferred to the intensive care unit for close monitoring and treatment to prevent and deal with potential serious complications. 8. A system using the method for predicting the intensive care demand of patients with cerebral hemorrhage based on multimodal fusion as claimed in any one of claims 1 to 7, characterized in that it comprises a data processing module, a feature extraction module, and a demand prediction module; The data processing module is used to collect multimodal data of patients with cerebral hemorrhage and preprocess the multimodal data; The feature extraction module is used to extract features from the preprocessed multimodal data, and uses undersampling and oversampling techniques to handle the problem of data set imbalance; The demand prediction module is used to build a multimodal fusion monitoring demand prediction model, predict real-time clinical data, and identify the intensive care needs of patients with cerebral hemorrhage. 9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the method for predicting intensive care needs of patients with cerebral hemorrhage based on multimodal fusion as described in any one of claims 1 to 7 when executing the computer program. 10. A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the method for predicting the intensive care needs of patients with cerebral hemorrhage based on multimodal fusion as described in any one of claims 1 to 7 are implemented.