Tumor patient hospitalization full-period management system

Abstract

The invention relates to the technical field of tumor patient management, in particular to a tumor patient hospitalization full-period management system, which acquires and manages personal information, physiological data and treatment data of tumor patients through a data management module and ensures the real-time performance and integrity of the data. The data processing module improves the accuracy of the data by supplementing missing values and screening out abnormal values, optimizes the calculation efficiency through dimension reduction processing, retains main information features, and provides high-quality data input for intelligent prediction. The intelligent prediction module is based on a neural network algorithm, the future physiological parameter change trend of the patient is accurately predicted, the physiological anomaly occurrence probability is calculated, the prediction accuracy and timeliness are remarkably improved, and medical staff are helped to recognize potential risks in advance and provide personalized treatment. And the full-cycle visualization module displays patient data through a timeline and generates a risk object attention set, so that key monitoring of high-risk patients is realized, and the safety management level during hospitalization is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of tumor patient management, and in particular to a full-cycle management system for hospitalized tumor patients. Background Art

[0002] With the continuous advancement of medical technology, the treatment methods and efficacy of cancer patients are improving, and many patients are able to prolong their lives and improve their quality of life. However, the management of cancer patients during hospitalization remains a major challenge facing the medical field. Cancer patients usually have complex conditions and long courses of illness, and need to receive multiple treatments during hospitalization. The conditions of these patients fluctuate greatly and are easily affected by many factors, including treatment response, the occurrence of complications, and changes in psychological state. Therefore, it is particularly important to monitor and manage the physiological data and treatment data of cancer patients during the hospitalization period in real time.

[0003] Traditional hospitalization management systems focus on recording and storing patient data, relying mainly on manual input and maintenance. This approach results in untimely and incomplete data recording, and may even result in errors. At the same time, traditional systems usually lack the ability to deeply analyze and predict data, and are unable to identify potential risks and abnormalities in a timely manner. For example, for complications and rapid changes in the condition of cancer patients, existing systems often rely on the experience and judgment of medical staff and manual monitoring. This approach is not only prone to misjudgment, but may also lead to delayed processing of emergencies, delay the best treatment time, and thus affect the patient's treatment effect. For changes in the physiological parameters of cancer patients, it is difficult for existing systems to accurately and real-time monitor them. Fluctuations in these physiological parameters may indicate potential risks, such as abnormal changes in heart rate and blood pressure, which may indicate the occurrence of complications. If these abnormalities cannot be detected in time and handled accordingly, the safety of the patient may be threatened. The current hospitalization management system is usually limited to simple recording and storage of data, lacks the ability to perform real-time analysis, abnormality detection, and future trend prediction on these data, and is difficult to meet the comprehensive and full-cycle monitoring needs of cancer patients. Summary of the invention

[0004] To solve the above problems, the present invention provides a full-cycle management system for hospitalized cancer patients, which helps medical staff pay attention to the monitoring risks of cancer patients by combining data dimensionality reduction and screening processing and introducing neural networks for predictive calculations.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A full-cycle management system for hospitalized tumor patients, comprising: a data management module, a data processing module, an intelligent prediction module and a full-cycle visualization module, wherein the data management module, the data processing module, the intelligent prediction module and the full-cycle visualization module are connected in sequence;

[0007] The data management module is used to obtain personal information, physiological data and treatment data of tumor patients, wherein the personal information includes the identity information and hospitalization information of tumor patients, the physiological data includes physiological parameters, nutritional intake data and exercise status data, the physiological parameters include heart rate, blood oxygen value, respiratory rate and blood pressure value, and the treatment data includes tumor type, test report and treatment behavior data;

[0008] The data processing module is used to supplement the missing values ​​of the physiological parameters of tumor patients during the hospitalization period, obtain the abnormal values ​​of the physiological parameters and the corresponding timestamps, screen out the abnormal values ​​of the physiological parameters through the movement type and movement occurrence time in the movement state data, and after the screening is completed, perform data dimension reduction through principal component analysis and generate a physiological parameter vector set;

[0009] The intelligent prediction module is used to output the physiological parameters of the future time series of the tumor patient based on the physiological parameter vector set and treatment data of the tumor patient through a neural network, and calculate the probability of occurrence of physiological abnormality of the patient according to the physiological parameters of the future time series;

[0010] The full-cycle visualization module is used to visualize the patient's hospitalized physiological data and treatment data through a timeline, generate a risk object focus set based on the tumor patient's hospitalization information and the probability of physiological abnormalities, and perform real-time visualization of physiological data.

[0011] Furthermore, the method of supplementing missing values ​​of physiological parameters of tumor patients during the hospitalization period includes the following steps:

[0012] Perform regression modeling on historical physiological parameter data through time series to generate a time-based physiological parameter prediction model;

[0013] Based on the physiological parameter prediction model, the missing values ​​are calculated using the consecutive physiological parameters before and after the time series as reference points;

[0014] The error calculation is performed on the data after missing values ​​are supplemented, and the value with the smallest relative error is used as the supplementation result.

[0015] Furthermore, the abnormal value of the physiological parameter is obtained by comparing the physiological parameters of the tumor patient during the hospitalization period with the preset physiological parameter threshold value.

[0016] Furthermore, the step of filtering out abnormal values ​​of physiological parameters by using the exercise type and exercise occurrence time in the exercise state data includes the following steps:

[0017] Obtain the patient's exercise type and time of occurrence during hospitalization;

[0018] Determine the sports impact level according to the sports type, and determine the corresponding physiological parameter threshold adjustment range through the sports impact level;

[0019] Abnormal values ​​of physiological parameters are screened out based on the physiological parameter threshold adjustment range.

[0020] Furthermore, the data dimension reduction by principal component analysis and the generation of a physiological parameter vector set include the following steps:

[0021] Standardization of physiological parameters to complete screening;

[0022] Calculate the covariance matrix of the standardized physiological parameter data matrix;

[0023] Perform eigenvalue decomposition on the covariance matrix and extract the eigenvalues ​​and their corresponding eigenvectors;

[0024] Sort the eigenvectors according to the size of the eigenvalues, and select the first several eigenvectors whose cumulative variance contribution rate reaches the preset threshold as the principal components;

[0025] The original physiological parameter data are mapped to the selected principal component direction to generate a reduced-dimensional physiological parameter vector set.

[0026] Furthermore, the formula of the principal component analysis is as follows:

[0027]

[0028] Among them, Z i represents the dimension reduction result of the i-th principal component, X j is the physiological parameter data before the jth standardization; μ j and σ j are the mean and standard deviation of the jth physiological parameter respectively; ω ji is the component of the i-th eigenvector corresponding to the j-th physiological parameter; m is the dimension of the original physiological parameter.

[0029] Furthermore, the step of outputting the future time series of physiological parameters of the tumor patient through a neural network based on the physiological parameter vector set and treatment data of the tumor patient includes the following steps:

[0030] The treatment data is used as static features to fuse with the physiological parameter data of each time step to generate a fused input vector;

[0031] The current fusion input vector and the hidden state at the previous moment are calculated through the forget gate to determine the forgotten information;

[0032] Evaluate the current fused input vector and the previous hidden state through the input gate to determine the input information;

[0033] Combine the forgotten information and the input information to update the cell state;

[0034] The updated cell state and the previous hidden state are combined through the output gate to generate the current hidden state output;

[0035] Generate physiological parameters for future time series based on the last hidden state.

[0036] Furthermore, the calculation of the probability of occurrence of physiological abnormality of the patient according to the physiological parameters of the future time series includes:

[0037] Compare the physiological parameters of the future time series with the preset safety thresholds to identify the physiological parameters that exceed or fall below the safety threshold range;

[0038] Based on the number and magnitude of threshold violations, the probability of the patient experiencing physiological abnormalities within a preset time period is calculated.

[0039] A method for managing the entire hospitalization cycle of cancer patients, applied to any of the above-mentioned management systems for the entire hospitalization cycle of cancer patients, comprises the following steps:

[0040] Obtaining personal information, physiological data and treatment data of tumor patients, wherein the personal information includes the identity information and hospitalization information of the tumor patients, the physiological data includes physiological parameters, nutritional intake data and exercise status data, the physiological parameters include heart rate, blood oxygen value, respiratory rate and blood pressure value, and the treatment data includes tumor type, test report and treatment behavior data;

[0041] The missing values ​​of physiological parameters of tumor patients during the hospitalization period are supplemented, and the abnormal values ​​of physiological parameters and corresponding timestamps are obtained. The abnormal values ​​of physiological parameters are screened out by the exercise type and exercise occurrence time in the exercise status data. After the screening is completed, the data dimension is reduced through principal component analysis and a physiological parameter vector set is generated;

[0042] Based on the physiological parameter vector set and treatment data of tumor patients, the physiological parameters of the future time series of tumor patients are output through a neural network, and the probability of occurrence of physiological abnormalities of patients is calculated according to the physiological parameters of the future time series;

[0043] The patient's hospitalized physiological data and treatment data are visualized through the timeline, and a risk object focus set is generated based on the hospitalization information of tumor patients and the probability of physiological abnormalities, and the physiological data is visualized in real time.

[0044] The beneficial effects of the present invention are as follows: the present invention collects and manages multi-dimensional data of tumor patients through a data management module, including personal information, physiological data and treatment data of patients, thereby ensuring the real-time and integrity of the data; the missing values ​​of physiological data are supplemented through a data processing module to ensure the integrity of the data; abnormal data caused by external factors such as patient movement are excluded through automatic screening of abnormal values, thereby improving the accuracy of the data. In particular, the principal component analysis (PCA) technology is used to reduce the dimension of physiological parameters to generate a representative set of physiological parameter vectors, which greatly reduces the data dimension, optimizes the calculation efficiency, and retains the main information characteristics of the data, providing high-quality data input for subsequent intelligent prediction. The introduction of the intelligent prediction module enables this scheme to accurately predict the future physiological parameters of tumor patients based on a neural network algorithm. The module uses a deep learning method to integrate the physiological parameter vector set and treatment data, and can generate the trend of changes in the physiological parameters of patients in the future period of time, and calculate the probability of physiological abnormalities in patients. Compared with the traditional method that relies on experience judgment, the intelligent prediction module greatly improves the accuracy and timeliness of prediction, enables medical staff to identify potential health risks in advance, and provides a scientific basis for personalized treatment. The realization of this function not only improves the early warning capability for emergencies, but also provides strong support for preventive care for patients. Finally, the full-cycle visualization module displays the patient's physiological data and treatment data through an intuitive timeline, and generates a risk object focus set based on the results of the intelligent prediction module, realizing key monitoring of high-risk patients. Medical staff can use this module to understand the changes in the patient's health status in real time and take corresponding measures quickly when abnormalities are found. This combination of real-time data visualization and risk management makes up for the shortcomings of traditional systems in risk identification and decision support, and greatly improves the level of safety management of patients during hospitalization. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a structural schematic diagram of a full-cycle management system for hospitalized tumor patients in the present invention.

[0046] Figure 2 It is a flow chart of the steps of performing data dimension reduction and generating a physiological parameter vector set through principal component analysis in the present invention. DETAILED DESCRIPTION

[0047] See also Figure 1-2 As shown, the present invention relates to a full-cycle management system for hospitalized tumor patients, comprising: a data management module, a data processing module, an intelligent prediction module and a full-cycle visualization module, wherein the data management module, the data processing module, the intelligent prediction module and the full-cycle visualization module are connected in sequence;

[0048] The data management module is used to obtain personal information, physiological data and treatment data of tumor patients, wherein the personal information includes the identity information and hospitalization information of tumor patients, the physiological data includes physiological parameters, nutritional intake data and exercise status data, the physiological parameters include heart rate, blood oxygen value, respiratory rate and blood pressure value, and the treatment data includes tumor type, test report and treatment behavior data;

[0049] The data processing module is used to supplement the missing values ​​of the physiological parameters of tumor patients during the hospitalization period, obtain the abnormal values ​​of the physiological parameters and the corresponding timestamps, screen out the abnormal values ​​of the physiological parameters through the movement type and movement occurrence time in the movement state data, and after the screening is completed, perform data dimension reduction through principal component analysis and generate a physiological parameter vector set;

[0050] The intelligent prediction module is used to output the physiological parameters of the future time series of the tumor patient based on the physiological parameter vector set and treatment data of the tumor patient through a neural network, and calculate the probability of occurrence of physiological abnormality of the patient according to the physiological parameters of the future time series;

[0051] The full-cycle visualization module is used to visualize the patient's hospitalized physiological data and treatment data through a timeline, generate a risk object focus set based on the tumor patient's hospitalization information and the probability of physiological abnormalities, and perform real-time visualization of physiological data.

[0052] It should be noted that the data management module is mainly responsible for the collection and management of multi-dimensional data of patients. The system obtains the key physiological data of patients in real time by integrating with a variety of physiological parameter monitoring devices. Specifically, the system realizes the real-time collection and upload of patients' physiological parameters by accessing wearable devices such as smart watches and smart bracelets, as well as other medical devices containing sensors. Smart watches or bracelets are one of the key data acquisition devices in this system. These devices are usually equipped with a variety of sensors, including heart rate sensors, blood oxygen sensors, accelerometers, gyroscopes, etc. Through these sensors, smart watches can continuously monitor patients' heart rate, blood oxygen saturation, respiratory rate and other physiological parameters, and transmit these data to the system's data management module in real time through wireless communication methods such as Bluetooth or Wi-Fi. In addition, the system can also access the hospital's existing bedside monitors or mobile monitoring devices. These devices are usually used to monitor important physiological indicators such as blood pressure, body temperature and respiratory rate of hospitalized patients in real time. By integrating standardized interface protocols such as HL7 or FHIR protocols, the system can seamlessly obtain patients' physiological data from these devices and integrate them into a unified database for management. To ensure the accuracy and real-time nature of the data, the data management module uses edge computing technology, so that some data preprocessing can be completed on the acquisition device side. For example, after collecting heart rate data, the smartwatch can first perform noise filtering and signal processing to reduce data errors transmitted to the system. The processed data is uploaded to the hospital's central database via wireless communication for further processing and analysis.

[0053] The full-cycle visualization module adopts a timeline-based display method. The timeline interface displays all key data points of the patient from admission to discharge, including physiological data (such as heart rate, blood oxygen value, respiratory rate, blood pressure value, etc.) and treatment data (such as surgical records, drug use, test report results, etc.). These data are arranged in chronological order, providing a comprehensive perspective, so that medical staff can clearly see the changing trends of the patient's physiological state and treatment process during hospitalization. In actual operation, the design of the timeline is highly flexible and supports multi-level and multi-dimensional data display. For example, users can choose to zoom in on a specific time period to view detailed physiological data changes within a day or an hour. At the same time, the timeline can also integrate the changing trends of multiple physiological parameters and display them in the same chart to facilitate comprehensive analysis by medical staff. In addition, the display of treatment data also supports association with physiological data. When a treatment behavior (such as drug injection or surgery) occurs, the changes in physiological data at the corresponding time point will be highlighted to help medical staff evaluate the treatment effect. In order to further improve management efficiency, the full-cycle visualization module introduces the concept of risk object attention set. The system automatically includes patients at higher risk into the focus set based on the probability of physiological abnormalities provided by the intelligent prediction module. Risk assessment is based on a comprehensive analysis of multiple factors, including the patient's historical physiological data, current physiological status, treatment response, etc. The risk object focus set will be emphasized on the timeline with different colors or marks, so that medical staff can identify patients who need to be monitored at a glance. In addition, the full-cycle visualization module supports real-time data updates and dynamic visualization. The system is seamlessly connected with various monitoring devices and data processing modules. When new physiological data or treatment data is generated, the timeline will automatically refresh and reflect these data changes in real time. This real-time visualization feature ensures that medical staff always have the latest health status of the patient, can respond quickly to emergencies, and adjust treatment plans in a timely manner.

[0054] Furthermore, the method of supplementing missing values ​​of physiological parameters of tumor patients during the hospitalization period includes the following steps:

[0055] Perform regression modeling on historical physiological parameter data through time series to generate a time-based physiological parameter prediction model;

[0056] Based on the physiological parameter prediction model, the missing values ​​are calculated using the consecutive physiological parameters before and after the time series as reference points;

[0057] The error calculation is performed on the data after missing values ​​are supplemented, and the value with the smallest relative error is used as the supplementation result.

[0058] In some embodiments, the system performs time series analysis on the patient's historical physiological parameter data to establish a time-based physiological parameter prediction model. Specifically, the time series analysis uses regression modeling technology, takes the patient's historical data of physiological parameters as input, and constructs a mathematical model that can capture the trend of physiological parameter changes. This model identifies potential patterns and laws in the data by analyzing a series of time point data. For example, the system can use physiological parameter data such as heart rate and blood pressure in the past period of time to predict future heart rate change trends. During the modeling process, the regression model will select a suitable regression method, such as a linear regression model, to adapt to the time dynamic characteristics of different physiological parameters. After generating a physiological parameter prediction model based on a time series, the system will use the model to supplement the missing physiological parameters. The specific method is to calculate the missing physiological parameter values ​​based on the predicted values ​​provided by the model, combined with the continuous physiological parameters before and after the time series as reference points. The system first extracts the values ​​before and after the time point where the missing value is located based on the context information of the time series, and derives the possible physiological parameter values ​​through the prediction model. For example, if the blood pressure data at a certain moment is missing, the system will infer the most likely blood pressure value based on the blood pressure values ​​before and after the moment and the trend predicted by the time series model. This step ensures that the supplemented missing values ​​are consistent with other data in the time series, avoiding abnormal fluctuations or misinterpretations caused by missing data. Finally, in order to ensure that the supplemented missing values ​​are accurate enough, the system will perform error calculations on the supplemented data and select the value with the smallest relative error as the final supplement result. In this step, the system calculates the relative error between the predicted value and the reference point, and uses optimization algorithms, such as the least squares method, to find the supplemented value that gives the highest fit to the time series data. Relative error is a key indicator for measuring the consistency between the supplemented value and the real data. The system will give priority to the supplemented value with the smallest relative error to ensure the accuracy of the data. This process can significantly reduce the errors that may be caused by data supplementation and improve the reliability of subsequent analysis.

[0059] Furthermore, the abnormal value of the physiological parameter is obtained by comparing the physiological parameters of the tumor patient during the hospitalization period with the preset physiological parameter threshold value.

[0060] It should be noted that the system sets a set of predefined physiological parameter thresholds based on the physiological characteristics and clinical treatment needs of tumor patients. These thresholds are usually based on the advice of medical experts and large-scale statistical analysis of clinical data, including but not limited to key physiological indicators such as heart rate, blood oxygen saturation, blood pressure, and respiratory rate. Each physiological parameter has one or more threshold ranges to represent a normal physiological state. For example, the normal range of heart rate may be set to 60 to 100 beats per minute, and the normal range of blood oxygen saturation may be set to 95% to 100%. These thresholds can be dynamically adjusted according to individual differences, disease status, and treatment plans of patients to better adapt to the actual situation of patients. After the physiological parameter thresholds are set, the system will continuously monitor and compare the physiological data collected in real time by tumor patients during hospitalization. When the patient's physiological parameters exceed the preset threshold range, the system will immediately identify the parameter as an abnormal value. For example, if the patient's heart rate suddenly rises to 120 beats per minute, and the preset upper limit of the heart rate is 100 beats, the system will automatically identify the heart rate data as an abnormal value. Similarly, if the blood oxygen saturation drops to 90%, which is lower than the preset lower limit of 95%, the system will also identify the data as an abnormal value.

[0061] Furthermore, the step of filtering out abnormal values ​​of physiological parameters by using the exercise type and exercise occurrence time in the exercise state data includes the following steps:

[0062] Obtain the patient's exercise type and time of occurrence during hospitalization;

[0063] Determine the sports impact level according to the sports type, and determine the corresponding physiological parameter threshold adjustment range through the sports impact level;

[0064] Abnormal values ​​of physiological parameters are screened out based on the physiological parameter threshold adjustment range.

[0065] In some embodiments, first, the system obtains the patient's exercise type and exercise occurrence time during hospitalization in real time through an integrated wearable device (such as a smart watch or smart bracelet) or a monitoring system in the hospital. These devices are usually equipped with sensors such as accelerometers and gyroscopes, which can accurately detect the patient's movement pattern, including different types of movement such as walking, running, standing, and bed rest. In addition, the system also records the specific time of each movement and synchronizes these data with the physiological parameter data to ensure the timeliness and accuracy of the analysis. For example, the system can record that the patient starts walking at 10:15 am, lasts for 10 minutes, and then returns to the bed to rest at 10:25. After obtaining the exercise data, the system will evaluate the impact of different exercise types on physiological parameters and determine the corresponding exercise impact level. For example, running has a greater impact on heart rate and respiratory rate, while standing or slow walking has a relatively small impact. The system assigns an impact level to each exercise according to the type of exercise, and appropriately adjusts the preset physiological parameter thresholds according to these levels. For example, for heart rate, the system may increase the upper threshold limit by 20% when running, while only increasing the upper threshold limit by 10% when walking. This adjustment takes into account the transient impact of exercise on physiological parameters, avoiding mistaking natural physiological fluctuations caused by exercise for abnormalities. Next, the system uses the adjusted physiological parameter threshold range to screen out physiological data that has been identified as abnormal. When the system detects that the value of a physiological parameter exceeds the standard threshold, it first checks whether the abnormal value occurs during the patient's exercise. If so, the system adjusts the threshold according to the impact level of the exercise type and re-evaluates whether the data is still abnormal. For example, if the patient's heart rate reaches 110 beats per minute during running, and the upper limit of the adjustment threshold for running is 120 beats per minute, the system will no longer regard this heart rate value as abnormal. On the contrary, if the heart rate reaches 130 beats per minute, even if the impact of exercise is considered, the value will still be marked as abnormal. This screening process can not only significantly reduce false alarms caused by exercise, but also accurately identify true physiological abnormalities, thereby improving the accuracy of the system's monitoring of the physiological status of tumor patients. At the same time, this method ensures the continuity and credibility of physiological parameter data, providing a more reliable data basis for subsequent intelligent prediction and decision support.

[0066] Furthermore, the data dimension reduction by principal component analysis and the generation of a physiological parameter vector set include the following steps:

[0067] Standardization of physiological parameters to complete screening;

[0068] Calculate the covariance matrix of the standardized physiological parameter data matrix;

[0069] Perform eigenvalue decomposition on the covariance matrix and extract the eigenvalues ​​and their corresponding eigenvectors;

[0070] Sort the eigenvectors according to the size of the eigenvalues, and select the first several eigenvectors whose cumulative variance contribution rate reaches the preset threshold as the principal components;

[0071] The original physiological parameter data are mapped to the selected principal component direction to generate a reduced-dimensional physiological parameter vector set.

[0072] In some embodiments, first, the system performs standardization on the physiological parameter data that has been screened out of abnormal values. Since the dimensions and value ranges of different physiological parameters are different, such as the heart rate in units of beats / minute and the blood oxygen saturation in percentage, the purpose of standardization is to convert these data of different dimensions to the same order of magnitude for subsequent analysis. The specific operation of standardization is usually to subtract the mean value of each physiological parameter data and divide it by its standard deviation, so that the mean value of the data is 0 and the standard deviation is 1. This step ensures that the influence weight of each physiological parameter on the result in the subsequent analysis is the same, avoiding deviations caused by dimensional differences. After standardization, the system calculates the covariance matrix of the standardized physiological parameter data matrix. The covariance matrix is ​​a square matrix used to describe the linear relationship between each physiological parameter, and its elements represent the covariance between two physiological parameters. For example, the covariance of heart rate and blood pressure may represent the common fluctuation trend of these two parameters over a period of time. The calculation result of the covariance matrix will directly reflect the correlation of each physiological parameter, which provides a basis for principal component analysis. Next, the system performs eigenvalue decomposition on the covariance matrix to extract eigenvalues ​​and their corresponding eigenvectors. The eigenvalue reflects the size of the variance of the data in the direction of the eigenvector, while the eigenvector represents this direction. In actual operation, the system uses linear algebra methods to decompose the covariance matrix and obtain multiple eigenvalues ​​and their corresponding eigenvectors. These eigenvalues ​​and eigenvectors together constitute the main direction of the data, revealing the inherent structure and variation characteristics of the data. After obtaining the eigenvalues ​​and eigenvectors, the system sorts the eigenvectors according to the size of the eigenvalues, and selects the first several eigenvectors whose cumulative variance contribution rate reaches the preset threshold as the principal components. The cumulative variance contribution rate refers to the proportion of the total variation that can be explained by these principal components, which is usually set to 85% or 90% to ensure that the selected principal components can fully represent the main information of the original data. For example, if the cumulative variance contribution rate of the first three eigenvectors has reached 90%, the system will use these three eigenvectors as the final principal components and ignore other eigenvectors. This selection process greatly simplifies the dimension of the data after dimensionality reduction while minimizing information loss. Finally, the system maps the original physiological parameter data to the selected principal component direction to generate a set of physiological parameter vectors after dimensionality reduction. This means that each original data point is projected into a new coordinate system consisting of the selected principal component directions. The generated set of physiological parameter vectors contains the main variation characteristics of the original data, but the dimension is significantly reduced, making subsequent data processing and analysis more efficient. For example, a data set containing 10 different physiological parameters may be reduced to a vector set containing only 3 principal components, greatly simplifying the complexity of the data.

[0073] Furthermore, the formula of the principal component analysis is as follows:

[0074]

[0075] Among them, Z i represents the dimension reduction result of the i-th principal component, X j is the physiological parameter data before the jth standardization; μ j and σ j are the mean and standard deviation of the jth physiological parameter respectively; ω ji is the component of the i-th eigenvector corresponding to the j-th physiological parameter; m is the dimension of the original physiological parameter.

[0076] Specifically, principal component analysis (PCA) first transforms each physiological parameter X j Standardization is performed by subtracting the mean μ of the parameter j and divided by the standard deviation σ j To eliminate the dimensional differences between different physiological parameters. The standardized data ensures that each physiological parameter has equal weight in principal component analysis and makes it easier to discover the potential structure in the data. Next, PCA extracts the principal components of each physiological parameter by eigenvalue decomposition of the covariance matrix, which is ω in the above formula. ji These principal components represent the direction of the data in the new coordinate system, and each direction corresponds to a eigenvector. The weight ω in the formula ji It determines the contribution of each physiological parameter to each principal component. Finally, the system projects the original high-dimensional physiological parameter data onto the principal component directions through the above formula to generate the reduced-dimensional data Z i These principal components retain the most significant variation characteristics in the original data, simplify the data dimension, and provide refined and representative input for subsequent data analysis and intelligent prediction.

[0077] Furthermore, the step of outputting the future time series of physiological parameters of the tumor patient through a neural network based on the physiological parameter vector set and treatment data of the tumor patient includes the following steps:

[0078] The treatment data is used as static features to fuse with the physiological parameter data of each time step to generate a fused input vector;

[0079] The current fusion input vector and the hidden state at the previous moment are calculated through the forget gate to determine the forgotten information;

[0080] Evaluate the current fused input vector and the previous hidden state through the input gate to determine the input information;

[0081] Combine the forgotten information and the input information to update the cell state;

[0082] The updated cell state and the previous hidden state are combined through the output gate to generate the current hidden state output;

[0083] Generate physiological parameters for future time series based on the last hidden state.

[0084] In some embodiments, the treatment data is used as a static feature and fused with the physiological parameter data at each time step to generate a fused input vector. The treatment data includes the patient's specific treatment plan, drug use, surgical records, etc., which are relatively stable in time but have an important impact on the dynamic changes of physiological parameters. The system combines these static features with the gradually collected physiological parameter vectors to ensure that the model can fully consider the impact of treatment on physiological state. For example, in a time step, the fused input vector may include the heart rate, blood pressure, and the type of chemotherapy the patient is receiving at that time. Next, the system calculates the current fused input vector and the hidden state of the previous moment through the forget gate in the LSTM unit to determine which information needs to be forgotten. The role of the forget gate is to help the model decide which historical information is no longer important for the prediction of the current time step, thereby removing this information from the cell state. The forget gate is controlled by a sigmoid function, and its output value is between 0 and 1, indicating the degree of forgetting. For example, in some cases, the system may choose to forget some information related to early treatment to better focus on the current physiological state. After that, the system evaluates the current fused input vector and the hidden state of the previous moment through the input gate to determine what new information needs to be added to the cell state. The input gate is also controlled by the sigmoid function, which determines which new physiological parameters and treatment data will be introduced into the current cell state. This step ensures that the model can flexibly adjust the cell state to adapt to the newly generated physiological parameter changes. For example, when a sharp fluctuation in the patient's heart rate is detected, the input gate may significantly introduce this information into the current state update. Combining the calculation results of the forget gate and the input gate, the system updates the cell state of the LSTM unit. The updated cell state contains the historical information filtered by the forget gate and the new information added by the input gate. This update process is the key to the LSTM model's ability to effectively capture long-term dependencies, allowing the model to maintain a memory of the patient's physiological state for a long time while flexibly adapting to new physiological changes. Finally, the system combines the updated cell state with the hidden state of the previous moment through the output gate to generate the current hidden state output. The output gate further processes the cell state to determine what information should be reflected in the prediction result of the current time step. The hidden state generated in this step not only contains the information of the current time step, but also passes it to the next time step as input to ensure the consistency between time series data. For example, when predicting the patient's future heart rate trend, the current hidden state will comprehensively consider recent heart rate changes, drug effects, and other relevant physiological parameters. Finally, the system generates physiological parameter prediction results for future time series based on the last hidden state. These prediction results provide an expectation of the patient's future physiological state and provide a scientific basis for clinical medical staff's treatment decisions.For example, the system may predict that the patient's blood pressure will gradually increase in the next few hours, which can remind medical staff to take appropriate intervention measures in advance. Through the above implementation process, the system takes advantage of the structural advantages of neural networks, especially LSTM, to accurately predict the physiological state of cancer patients in time series. This method effectively combines the dynamic changes of physiological parameters with the static characteristics of treatment data, providing a powerful tool to support personalized treatment and risk management of cancer patients during hospitalization.

[0085] Furthermore, the calculation of the probability of occurrence of physiological abnormality of the patient according to the physiological parameters of the future time series includes:

[0086] Compare the physiological parameters of the future time series with the preset safety thresholds to identify the physiological parameters that exceed or fall below the safety threshold range;

[0087] Based on the number and magnitude of threshold violations, the probability of the patient experiencing physiological abnormalities within a preset time period is calculated.

[0088] In some embodiments, first, the system compares the predicted physiological parameters of the future time series with the preset safety threshold. The safety threshold is set based on clinical experience and a large amount of statistical data, indicating the reasonable range of the patient's physiological parameters under normal circumstances. For example, the safety threshold of heart rate may be set to 60 to 100 beats per minute, and the safety range of blood pressure may be set to 120 / 80 mmHg. During the comparison process, the system will gradually check the physiological parameter values ​​at each time point and identify those physiological parameters that exceed or fall below the preset safety threshold. This comparison process is automated to ensure that potential physiological abnormalities can be detected in real time and accurately. Once the system identifies physiological parameters that exceed the safety threshold, it will further analyze the frequency and amplitude of these abnormal values. Specifically, the system will count the number of times the threshold is exceeded within a preset time period (for example, the next 24 hours) and calculate the amplitude of each threshold exceeded. For example, if the heart rate is higher than 100 beats / minute at multiple time points, and the heart rate at some time points is as high as 120 beats / minute, this indicates that the patient's heart rate abnormality may be more serious. The system quantifies these situations that exceed the threshold to provide basic data for subsequent probability calculations. Next, the system calculates the probability of the patient experiencing physiological abnormalities within a preset time period based on the number and magnitude of threshold-exceeding events. This probability calculation can be done using a statistical model or a rule-based algorithm that takes into account the combined effects of multiple factors. For example, if the heart rate exceeds the safety threshold many times and by a large magnitude over the next 24 hours, the system will determine that the probability of physiological abnormalities is high. The calculation results will be expressed in the form of probability values, such as 50%, 70%, etc., which directly reflect the possibility of physiological abnormalities in the patient within a specific time period in the future. The results of this probability calculation are crucial for clinical applications. By predicting the risk of physiological abnormalities in patients in advance, medical staff can take preventive measures in a timely manner, such as adjusting treatment plans, increasing monitoring frequency, or initiating emergency intervention procedures. This early warning mechanism significantly improves the initiative and predictability of hospitalization management, and helps reduce the incidence of health emergencies.

[0089] The present invention also includes a method for managing the entire hospitalization period of tumor patients, comprising the following steps:

[0090] Obtaining personal information, physiological data and treatment data of tumor patients, wherein the personal information includes the identity information and hospitalization information of the tumor patients, the physiological data includes physiological parameters, nutritional intake data and exercise status data, the physiological parameters include heart rate, blood oxygen value, respiratory rate and blood pressure value, and the treatment data includes tumor type, test report and treatment behavior data;

[0091] The missing values ​​of physiological parameters of tumor patients during the hospitalization period are supplemented, and the abnormal values ​​of physiological parameters and corresponding timestamps are obtained. The abnormal values ​​of physiological parameters are screened out by the exercise type and exercise occurrence time in the exercise status data. After the screening is completed, the data dimension is reduced through principal component analysis and a physiological parameter vector set is generated;

[0092] Based on the physiological parameter vector set and treatment data of tumor patients, the physiological parameters of the future time series of tumor patients are output through a neural network, and the probability of occurrence of physiological abnormalities of patients is calculated according to the physiological parameters of the future time series;

[0093] The patient's hospitalized physiological data and treatment data are visualized through the timeline, and a risk object focus set is generated based on the hospitalization information of tumor patients and the probability of physiological abnormalities, and the physiological data is visualized in real time.

[0094] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A full-cycle management system for hospitalized cancer patients, characterized in that: include: A data management module, a data processing module, an intelligent prediction module and a full-cycle visualization module, wherein the data management module, the data processing module, the intelligent prediction module and the full-cycle visualization module are connected in sequence; The data management module is used to obtain personal information, physiological data and treatment data of tumor patients, wherein the personal information includes the identity information and hospitalization information of tumor patients, the physiological data includes physiological parameters, nutritional intake data and exercise status data, the physiological parameters include heart rate, blood oxygen value, respiratory rate and blood pressure value, and the treatment data includes tumor type, test report and treatment behavior data; The data processing module is used to supplement the missing values ​​of the physiological parameters of tumor patients during the hospitalization period, obtain the abnormal values ​​of the physiological parameters and the corresponding timestamps, screen out the abnormal values ​​of the physiological parameters through the movement type and movement occurrence time in the movement state data, and after the screening is completed, perform data dimension reduction through principal component analysis and generate a physiological parameter vector set; The intelligent prediction module is used to output the physiological parameters of the future time series of the tumor patient based on the physiological parameter vector set and treatment data of the tumor patient through a neural network, and calculate the probability of occurrence of physiological abnormality of the patient according to the physiological parameters of the future time series; The full-cycle visualization module is used to visualize the patient's hospitalized physiological data and treatment data through a timeline, generate a risk object focus set based on the tumor patient's hospitalization information and the probability of physiological abnormalities, and perform real-time visualization of physiological data.

2. A full-cycle management system for hospitalized cancer patients according to claim 1, characterized in that: The method of supplementing missing values ​​of physiological parameters of tumor patients during the hospitalization period comprises the following steps: Perform regression modeling on historical physiological parameter data through time series to generate a time-based physiological parameter prediction model; Based on the physiological parameter prediction model, the missing values ​​are calculated using the consecutive physiological parameters before and after the time series as reference points; The error calculation is performed on the data after missing values ​​are supplemented, and the value with the smallest relative error is used as the supplementation result.

3. A full-cycle management system for hospitalized cancer patients according to claim 1, characterized in that: The abnormal value of the physiological parameter is obtained by comparing the physiological parameters of the tumor patient during the hospitalization period with the preset physiological parameter threshold value.

4. A full-cycle management system for hospitalized cancer patients according to claim 3, characterized in that: The method of filtering out abnormal values ​​of physiological parameters by using the exercise type and exercise occurrence time in the exercise state data comprises the following steps: Obtain the patient's exercise type and time of occurrence during hospitalization; Determine the sports impact level according to the sports type, and determine the corresponding physiological parameter threshold adjustment range through the sports impact level; Abnormal values ​​of physiological parameters are screened out based on the physiological parameter threshold adjustment range.

5. A full-cycle management system for hospitalized cancer patients according to claim 1, characterized in that: The method of performing data dimension reduction by principal component analysis and generating a physiological parameter vector set comprises the following steps: Standardization of physiological parameters to complete screening; Calculate the covariance matrix of the standardized physiological parameter data matrix; Perform eigenvalue decomposition on the covariance matrix and extract the eigenvalues ​​and their corresponding eigenvectors; Sort the eigenvectors according to the size of the eigenvalues, and select the first several eigenvectors whose cumulative variance contribution rate reaches the preset threshold as the principal components; The original physiological parameter data are mapped to the selected principal component direction to generate a reduced-dimensional physiological parameter vector set.

6. A full-cycle management system for hospitalized cancer patients according to claim 5, characterized in that: The formula for the principal component analysis is as follows: Among them, Z i represents the dimension reduction result of the i-th principal component, X j is the physiological parameter data before the jth standardization; μ j and σ j are the mean and standard deviation of the jth physiological parameter; w ji is the component of the i-th eigenvector corresponding to the j-th physiological parameter; m is the dimension of the original physiological parameter.

7. A full-cycle management system for hospitalized cancer patients according to claim 1, characterized in that: The method of outputting the future time series of physiological parameters of the tumor patient through a neural network based on the physiological parameter vector set and treatment data of the tumor patient comprises the following steps: The treatment data is used as static features to fuse with the physiological parameter data of each time step to generate a fused input vector; The current fusion input vector and the hidden state at the previous moment are calculated through the forget gate to determine the forgotten information; Evaluate the current fused input vector and the previous hidden state through the input gate to determine the input information; Combine the forgotten information and the input information to update the cell state; The updated cell state and the previous hidden state are combined through the output gate to generate the current hidden state output; Generate physiological parameters for future time series based on the last hidden state.

8. A full-cycle management system for hospitalized cancer patients according to claim 7, characterized in that: Calculating the probability of occurrence of physiological abnormality of the patient according to the physiological parameters of the future time series includes: Compare the physiological parameters of the future time series with the preset safety thresholds to identify the physiological parameters that exceed or fall below the safety threshold range; Based on the number and magnitude of threshold violations, the probability of the patient experiencing physiological abnormalities within a preset time period is calculated.

9. A method for managing the entire hospitalization cycle of cancer patients, applied to a management system for the entire hospitalization cycle of cancer patients as claimed in any one of claims 1 to 8, characterized in that: The following steps are involved: Obtaining personal information, physiological data and treatment data of tumor patients, wherein the personal information includes the identity information and hospitalization information of the tumor patients, the physiological data includes physiological parameters, nutritional intake data and exercise status data, the physiological parameters include heart rate, blood oxygen value, respiratory rate and blood pressure value, and the treatment data includes tumor type, test report and treatment behavior data; The missing values ​​of physiological parameters of tumor patients during the hospitalization period are supplemented, and the abnormal values ​​of physiological parameters and corresponding timestamps are obtained. The abnormal values ​​of physiological parameters are screened out by the exercise type and exercise occurrence time in the exercise status data. After the screening is completed, the data dimension is reduced through principal component analysis and a physiological parameter vector set is generated; Based on the physiological parameter vector set and treatment data of tumor patients, the physiological parameters of the future time series of tumor patients are output through a neural network, and the probability of occurrence of physiological abnormalities of patients is calculated according to the physiological parameters of the future time series; The patient's hospitalized physiological data and treatment data are visualized through the timeline, and a risk object focus set is generated based on the hospitalization information of tumor patients and the probability of physiological abnormalities, and the physiological data is visualized in real time.

Images