A method and system for assessing the state of agitated patients based on big data analysis

Abstract

The present invention discloses a method and system for assessing the state of agitated patients based on big data analysis, and specifically relates to the field of state assessment. By collecting the original multi-source data stream of the patient and performing feature dimension decomposition and preliminary redundant marking, nonlinear redundancy analysis is used to identify feature subsets with false emotion correlation, and combined effect cross-validation technology is used to locate the feature dimension combination that produces emotional illusions. Through the feature combination with emotional illusion effects, a corrected effective feature data set is obtained. A prediction model for agitated states based on an effective feature set is constructed. By introducing false correlation pattern mining at the feature combination level, the problem of emotion recognition bias caused by redundant feature interactions in traditional high-dimensional data analysis is overcome, and the problem of counterintuitive prediction results caused by emotional illusions in traditional methods is solved, effectively improving the biological interpretability and clinical prediction accuracy of agitated state assessment.

Description

Technical Field

[0001] The present invention relates to the technical field of state assessment, and more specifically, to a method and system for assessing the state of agitated patients based on big data analysis. Background Art

[0002] In the field of monitoring and evaluating the status of patients with mental illness, existing technologies generally use multimodal data to construct big data analysis models to assess patients' agitation. However, traditional methods have obvious defects in the integration of high-dimensional heterogeneous data: on the one hand, the nonlinear redundant associations between multimodal data features are not effectively decoupled, resulting in implicit pseudo-correlations in the feature space interfering with the model learning process; on the other hand, conventional feature screening methods have difficulty identifying false correlation patterns across dimensional combinations, causing the model to capture pseudo-features of emotional fluctuations that are unrelated to the actual pathological state. These problems produce illogical prediction biases in complex scenarios, weakening the medical credibility of the agitation assessment results, and thus affecting the accuracy and timeliness of clinical intervention decisions.

[0003] In order to solve the above problems, a technical solution is now provided. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method and system for assessing the state of an agitated patient based on big data analysis to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for assessing the state of a restless patient based on big data analysis comprises the following steps:

[0007] Collecting the original multi-source data streams of the patient within a preset time period, and performing feature dimension decomposition and preliminary redundancy marking on the original data in each data stream to obtain an initial feature dimension set of the original multi-source data stream;

[0008] For the initial feature dimension set, a nonlinear redundancy analysis method is used to identify the feature subset that produces false emotional feature correlations and form a redundant feature factor set;

[0009] Performing combined effect cross-validation analysis on the redundant feature factor set to determine the feature dimension combination that produces the emotion artifact effect and generate a set of false emotion association patterns;

[0010] According to the set of false emotion association patterns, the feature combinations with emotion artifact effects are removed from the initial feature dimension set of the original multi-source data stream to obtain a corrected valid feature data set;

[0011] Based on the effective feature data set, a prediction model for the evolution of the agitation state is constructed to evaluate the patient's current agitation state and output the recognition result of the patient's agitation state.

[0012] In a preferred embodiment, multi-source data streams of patients within a preset time period are collected, and the original data in each data stream is decomposed into feature dimensions and preliminarily marked for redundancy to obtain an initial feature dimension set of the original data stream, specifically:

[0013] Collect patient heart rate data stream, skin conductance pulse data stream, patient motion frequency data stream and spatial displacement trajectory data stream;

[0014] Decompose the patient's heart rate data stream into time domain features and frequency domain features;

[0015] Decompose the skin conductance pulse data stream into the dimensions of pulse amplitude features and peak density features;

[0016] Decompose the patient's movement frequency data stream into the dimensions of movement number per unit time and movement rhythm;

[0017] Decompose the spatial displacement trajectory data stream into the dimensions of trajectory curvature and displacement frequency characteristics;

[0018] The data dimension correlation analysis method is used to perform preliminary redundancy marking on each feature dimension after disassembly to obtain the initial feature dimension set.

[0019] In a preferred embodiment, a data dimension correlation analysis method is used to perform preliminary redundancy marking on each feature dimension after decomposition to obtain an initial feature dimension set, specifically:

[0020] Calculate the Pearson correlation coefficient matrix between each feature dimension after decomposition to determine the degree of linear correlation between feature dimensions;

[0021] Compare the correlation coefficients between any two feature dimensions in the Pearson correlation coefficient matrix. When the correlation coefficient is greater than the linear correlation threshold, mark the feature dimension with low diagnostic contribution in the two feature dimensions as a preliminary redundant dimension.

[0022] Classify and label all feature dimensions marked as preliminary redundant dimensions to form a preliminary redundant feature label set;

[0023] The preliminary redundant feature tag set is mapped and compared with each feature dimension set after disassembly to obtain the initial feature dimension set.

[0024] In a preferred embodiment, for the initial feature dimension set, a nonlinear redundancy analysis method is used to identify a feature subset that produces false emotion feature correlations to form a redundant feature factor set, specifically:

[0025] The kernel function mapping method is used to project the initial feature dimension set into a high-dimensional nonlinear feature space;

[0026] The nonlinear correlation between the projected feature dimensions is evaluated based on the mutual information index;

[0027] When the mutual information index between the projected feature dimensions is greater than the nonlinear redundancy threshold, the corresponding feature dimension combination is determined as a feature subset with false emotion feature correlation;

[0028] All identified feature subsets are integrated to form a redundant feature factor set.

[0029] In a preferred embodiment, a combination effect cross-validation analysis is performed on the redundant feature factor set to determine the feature dimension combination that causes the emotion artifact effect and generate a set of false emotion association patterns, specifically:

[0030] Each feature subset in the redundant feature factor set is randomly combined to form multiple different combination patterns;

[0031] For each combination pattern, the data of the combination pattern is input into the preset emotional state classifier using the cross-validation method to calculate the degree of interference of each combination pattern on the agitation state classification result;

[0032] When the interference degree of the combination pattern on the classification of the agitated state exceeds a predetermined interference threshold, the combination pattern is determined to be a characteristic dimension combination that causes an emotional illusion effect;

[0033] All characteristic dimension combinations that appear to have false emotional effects are integrated to generate a set of false emotional association patterns.

[0034] Each feature dimension combination in the false emotion association pattern set is removed from the initial feature dimension set in turn;

[0035] Perform relevance review analysis on the remaining features of the feature dimension set after removal;

[0036] The remaining feature dimension set after the correlation review analysis is re-labeled to obtain a corrected valid feature data set.

[0037] In a preferred embodiment, a prediction model for the evolution of agitation is constructed based on a valid feature data set, the patient's current agitation state is evaluated, and the identification result of the patient's agitation state is output, specifically:

[0038] Taking the effective feature data set as input variables, a time series-based prediction model for the evolution of agitated states is constructed;

[0039] The parameters of the agitation state evolution prediction model are trained using a historical effective feature data set;

[0040] Input the patient's effective feature data at the current time node into the agitation state evolution prediction model to calculate the patient's current agitation state probability;

[0041] The patient's current agitation level is determined according to the agitation probability.

[0042] In a preferred embodiment, the patient's current agitation level is determined based on the probability of agitation, specifically:

[0043] Comparing the probability of the agitated state with a preset probability threshold interval;

[0044] When the probability of the agitation state is within the first probability threshold interval, the patient's current agitation state level is determined to be mild agitation;

[0045] When the probability of the agitation state is within the second probability threshold interval, the patient's current agitation state level is determined to be moderate agitation;

[0046] When the probability of the agitation state is within the third probability threshold interval, it is determined that the patient's current agitation state level is severe agitation.

[0047] On the other hand, the present invention provides a system for assessing the state of agitated patients based on big data analysis, comprising a data processing module, a redundancy analysis module, a cross-validation analysis module, a feature correction module, and a state assessment module;

[0048] Data processing module: collects the original multi-source data streams of patients within a preset time period, and performs feature dimension decomposition and preliminary redundancy marking on the original data in each data stream to obtain the initial feature dimension set of the original multi-source data stream;

[0049] Redundancy analysis module: For the initial feature dimension set, a nonlinear redundancy analysis method is used to identify the feature subsets that produce false emotional feature correlations and form a redundant feature factor set;

[0050] Cross-validation analysis module: performs combined effect cross-validation analysis on redundant feature factor sets, determines the feature dimension combination that produces emotion artifact effects, and generates a set of false emotion association patterns;

[0051] Feature correction module: Based on the set of false emotion association patterns, the feature combinations with emotion artifact effects are removed from the initial feature dimension set of the original multi-source data stream to obtain the corrected valid feature data set;

[0052] State assessment module: Builds an agitation state evolution prediction model based on a valid feature data set, evaluates the patient's current agitation state, and outputs the identification result of the patient's agitation state.

[0053] The technical effects and advantages of the method and system for assessing the state of agitated patients based on big data analysis of the present invention are as follows:

[0054] By constructing a dynamic screening and verification mechanism for feature dimensions, the accuracy and reliability of the emotional state assessment of agitated patients have been improved. By introducing nonlinear redundancy analysis and combined effect cross-validation methods, the problem of misjudgment of emotional features caused by feature redundancy in traditional assessments is effectively solved. It can accurately identify and remove false correlation patterns between multimodal data, avoiding the interference of emotional illusion effects on assessment results. By constructing a prediction model for the evolution of agitated states based on a corrected set of effective feature data, in-depth analysis and real-time tracking of patients' agitated states are achieved, enhancing the interpretability of clinical decision-making. It breaks through the limitations of traditional single-dimensional assessment methods, provides new ideas for the accurate identification of complex mental symptoms, has the advantage of intelligent analysis that adapts to individual differences, and provides a more reliable objective basis for clinical intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a schematic diagram of a method for assessing the state of a restless patient based on big data analysis according to the present invention;

[0056] Figure 2 This is a structural diagram of a system for assessing the state of agitated patients based on big data analysis according to the present invention. DETAILED DESCRIPTION

[0057] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0058] Example 1

[0059] Figure 1 The present invention provides a method for assessing the state of a restless patient based on big data analysis, which comprises the following steps:

[0060] Collecting the original multi-source data streams of the patient within a preset time period, and performing feature dimension decomposition and preliminary redundancy marking on the original data in each data stream to obtain an initial feature dimension set of the original multi-source data stream;

[0061] For the initial feature dimension set, a nonlinear redundancy analysis method is used to identify the feature subset that produces false emotional feature correlations and form a redundant feature factor set;

[0062] Performing combined effect cross-validation analysis on the redundant feature factor set to determine the feature dimension combination that produces the emotion artifact effect and generate a set of false emotion association patterns;

[0063] According to the set of false emotion association patterns, the feature combinations with emotion artifact effects are removed from the initial feature dimension set of the original multi-source data stream to obtain a corrected valid feature data set;

[0064] Based on the effective feature data set, a prediction model for the evolution of the agitation state is constructed to evaluate the patient's current agitation state and output the recognition result of the patient's agitation state.

[0065] Specifically, the multi-source data streams of patients within a preset time period are collected, and the original data in each data stream is decomposed into feature dimensions and preliminarily marked for redundancy to obtain the initial feature dimension set of the original data stream, including:

[0066] Collect patient heart rate data stream, skin conductance pulse data stream, patient motion frequency data stream and spatial displacement trajectory data stream;

[0067] Specifically, within a preset time period, the patient's heart rate data is collected through biosensors, skin conductance pulse data is collected using skin electrical sensors, and the patient's movement frequency data and spatial displacement trajectory data are collected through cameras and motion sensors.

[0068] Each data stream reflects information from different dimensions. For example, the heart rate data stream describes the patient's autonomic nervous system regulation; the skin conductance data stream reflects the patient's physiological stress response; and the movement frequency and spatial displacement data streams reflect the patient's behavioral activity.

[0069] Decompose the patient's heart rate data stream into time domain features and frequency domain features;

[0070] Specifically, the time domain features include calculating the mean, standard deviation, maximum, minimum, etc. of the heart rate data; the frequency domain features include calculating the low-frequency power, high-frequency power, and the ratio of low frequency to high frequency using fast Fourier transform.

[0071] Decompose the skin conductance pulse data stream into the dimensions of pulse amplitude features and peak density features;

[0072] Specifically, the pulse amplitude feature refers to the amplitude of each pulse in the skin conductance signal, and the peak density feature refers to the frequency of peaks occurring per unit time.

[0073] For example, by detecting the amplitude difference between the peak and trough of the signal, the amplitude of each pulse is calculated, and the number of pulses per minute is counted as the peak density.

[0074] Decompose the patient's movement frequency data stream into the dimensions of movement number per unit time and movement rhythm;

[0075] Specifically, the number of movements per unit time can be obtained by counting the number of movements made by the patient within a specified time period; the movement rhythm reflects the distribution pattern of the time intervals between movements.

[0076] For example, if 30 movements are recorded in one minute, the number of movements per unit time is 30; if the average interval between movements is 2 seconds, it can be used as a reference indicator of movement rhythm.

[0077] Decompose the spatial displacement trajectory data stream into the dimensions of trajectory curvature and displacement frequency characteristics;

[0078] Specifically, the trajectory curvature reflects the degree of curvature of the patient's walking trajectory, which can be obtained by calculating the curvature of the curve formed by continuous position points; the displacement frequency characteristic is the number of position changes per unit time.

[0079] The data dimension correlation analysis method is used to perform preliminary redundancy marking on each feature dimension after disassembly to obtain the initial feature dimension set.

[0080] Specifically, the data dimension correlation analysis method is used to perform preliminary redundancy marking on each feature dimension after decomposition to obtain an initial feature dimension set, including:

[0081] Calculate the Pearson correlation coefficient matrix between each feature dimension after decomposition to determine the degree of linear correlation between feature dimensions;

[0082] Specifically, all the decomposed feature dimension data are arranged into a matrix form, and the Pearson correlation coefficient is calculated for any two columns in the matrix (representing two feature dimensions) to form a symmetric matrix.

[0083] Compare the correlation coefficients between any two feature dimensions in the Pearson correlation coefficient matrix. When the correlation coefficient is greater than the linear correlation threshold, mark the feature dimension with low diagnostic contribution in the two feature dimensions as a preliminary redundant dimension.

[0084] Specifically, when the Pearson correlation coefficient between two feature dimensions is greater than a predetermined linear correlation threshold (eg, 0.8), it is considered that there is linear redundancy between the two dimensions.

[0085] For example, if the correlation coefficient between feature A and feature B is 0.85, and the predetermined threshold is 0.8, it is determined that redundancy exists between feature A and feature B.

[0086] By comparing the diagnostic contribution of each feature in disease status assessment (which can be verified by historical data), feature dimensions with low diagnostic contribution are selected for marking.

[0087] For example, if both feature A and feature B are correlated, and statistical analysis shows that feature B contributes less to the prediction of agitation, feature B is marked as a preliminary redundant dimension.

[0088] Classify and label all feature dimensions marked as preliminary redundant dimensions to form a preliminary redundant feature label set;

[0089] Map and compare the preliminary redundant feature tag set with each feature dimension set after disassembly to obtain the initial feature dimension set;

[0090] Specifically, the classified redundant feature tag set is compared with the decomposed set of all feature dimensions, and the feature dimensions marked as preliminary redundant dimensions are removed. The remaining feature dimensions constitute the initial feature dimension set.

[0091] For example, if there are 10 feature dimensions after decomposition, and 3 redundant ones are removed after comparison, the initial feature dimension set contains 7 non-redundant feature dimensions.

[0092] Specifically, for the initial feature dimension set, a nonlinear redundancy analysis method is used to identify feature subsets that produce false emotional feature correlations, forming a redundant feature factor set, including:

[0093] The kernel function mapping method is used to project the initial feature dimension set into a high-dimensional nonlinear feature space;

[0094] Specifically, a kernel function (such as a Gaussian kernel function) is used to map the initial feature dimensions so that the nonlinear relationship that was originally difficult to distinguish in the low-dimensional space can be reflected in the high-dimensional space.

[0095] The nonlinear correlation between the projected feature dimensions is evaluated based on the mutual information index;

[0096] Specifically, for the random variables X and Y, first consider all possible combinations of values; for each pair of values ​​(x, y), calculate the probability of simultaneous occurrence, that is, the joint probability; then calculate the product of the probabilities when they appear independently, that is, multiply their respective marginal probabilities; then, take the logarithm of the ratio of the joint probability to the product of the marginal probabilities, which represents the difference between the amount of information when the pair of values ​​actually appear together and the amount of information when they appear independently; finally, multiply the logarithm by the corresponding joint probability and sum it over all possible combinations of values. The resulting sum is the mutual information index.

[0097] When the mutual information index between the projected feature dimensions is greater than the nonlinear redundancy threshold, the corresponding feature dimension combination is determined as a feature subset with false emotion feature correlation;

[0098] Specifically, a nonlinear redundancy threshold is set. When the mutual information index of any pair or multi-dimensional features is greater than the nonlinear redundancy threshold, it is judged that these features have nonlinear redundancy in the high-dimensional space, that is, there is false emotional feature correlation.

[0099] For example, if the mutual information index of the heart rate frequency domain feature and the skin conductance peak density feature after mapping is greater than the nonlinear redundancy threshold, the two features are classified as a group of false emotion feature correlation combinations.

[0100] All identified feature subsets are integrated to form a redundant feature factor set.

[0101] Specifically, a combination effect cross-validation analysis is performed on the redundant feature factor set to determine the feature dimension combination that produces the emotion illusion effect and generate a set of false emotion association patterns, including:

[0102] Each feature subset in the redundant feature factor set is randomly combined to form multiple different combination patterns;

[0103] Specifically, for each feature subset in the redundant feature factor set, several combination patterns are generated by random permutation and combination. When generating the combination pattern, the minimum and maximum number of combinations can be set to ensure that the combination pattern is representative.

[0104] For each combination pattern, the data of the combination pattern is input into the preset emotional state classifier using the cross-validation method to calculate the degree of interference of each combination pattern on the agitation state classification result;

[0105] Specifically, the cross-validation method may adopt K-fold validation, in which the data is divided into K parts, and K-1 parts are alternately used as training data and 1 part as test data.

[0106] For each combination pattern, calculate the classification accuracy or error rate after the combination pattern is input, and define the interference degree as:

[0107] Interference degree = max(0, standard classification accuracy - current combination mode classification accuracy). This means that the interference effect is calculated only when the classification accuracy under the current combination mode is lower than the standard classification accuracy. If the classification accuracy under the current combination mode is higher than the standard classification accuracy, no interference effect is considered.

[0108] The standard classification accuracy is the classification accuracy when no redundant features are added.

[0109] When the interference degree of the combination pattern on the classification of the agitated state exceeds a predetermined interference threshold, the combination pattern is determined to be a characteristic dimension combination that causes an emotional illusion effect;

[0110] All characteristic dimension combinations that appear to have false emotional effects are integrated to generate a set of false emotional association patterns.

[0111] Specifically, based on the false emotion association pattern set, feature combinations with emotion artifact effects are removed from the initial feature dimension set of the original multi-source data stream to obtain a corrected valid feature data set, including:

[0112] Each feature dimension combination in the false emotion association pattern set is removed from the initial feature dimension set in turn;

[0113] Specifically, for each feature dimension combination included in the false emotion association pattern set, the corresponding feature dimensions are located in the initial feature dimension set in turn and deleted to remove redundant features that have a false emotion impact on the agitation state assessment.

[0114] Perform relevance review analysis on the remaining features of the feature dimension set after removal;

[0115] Specifically, for the remaining feature dimension set after the removal process, the correlation analysis method (such as the Pearson correlation coefficient) is used again to verify the correlation between the features to ensure that the removal process does not introduce new redundancy.

[0116] Re-label the remaining feature dimension set after correlation review and analysis to obtain a corrected valid feature data set;

[0117] Specifically, after correlation review, the remaining feature dimensions are re-labeled and reorganized according to their contribution to the assessment of the patient's agitation state, thereby obtaining a corrected valid feature data set.

[0118] Specifically, a prediction model for the evolution of agitation is constructed based on a set of valid feature data, the patient's current agitation state is evaluated, and the identification result of the patient's agitation state is output, including:

[0119] Taking the effective feature data set as input variables, a time series-based prediction model for the evolution of agitated states is constructed;

[0120] Specifically, using the corrected set of valid feature data, the data is organized into time series, and the feature vectors at each time node are used as input to the agitation state evolution prediction model. When constructing the agitation state evolution prediction model, a prediction model suitable for time series data is used to ensure that the model can capture the evolution of patient characteristics over time.

[0121] The parameters of the agitation state evolution prediction model are trained using a historical effective feature data set;

[0122] Specifically, the parameters of the agitation state evolution prediction model are trained using historically collected valid feature data and correspondingly labeled agitation states, so that the model can accurately reflect the historical changes in the patient's agitation state. During the training process, the minimum mean square error (MSE) loss function can be used.

[0123] Input the patient's effective feature data at the current time node into the agitation state evolution prediction model to calculate the patient's current agitation state probability;

[0124] Specifically, after the training of the agitation state evolution prediction model is completed, the effective feature data of the patient at the current time node is input into the agitation state evolution prediction model, and the agitation state evolution prediction model outputs the patient's current agitation state probability according to the training parameters.

[0125] The patient's current agitation level is determined according to the agitation probability.

[0126] Specifically, the patient's current agitation level is determined based on the probability of the agitation state, including:

[0127] Comparing the probability of the agitated state with a preset probability threshold interval;

[0128] When the probability of the agitation state is within the first probability threshold interval, the patient's current agitation state level is determined to be mild agitation;

[0129] When the probability of the agitation state is within the second probability threshold interval, the patient's current agitation state level is determined to be moderate agitation;

[0130] When the probability of the agitation state is within the third probability threshold interval, the patient's current agitation state level is determined to be severe agitation;

[0131] Specifically, the preset probability threshold interval is set based on the probability distribution of historically collected valid feature data and the corresponding labeled agitation states in the output of the agitation state evolution prediction model, combined with the behavioral judgment standards of medical experts for different agitation levels (medical behavioral science usually has empirical thresholds for the grading of agitation levels. These behavioral labels are mapped to the model output probability, and the actual clinical grading standards are expressed as probability intervals to form a matching relationship between probability and medical labels). The probability values ​​output by the agitation state evolution prediction model are divided into three level intervals. Specifically, the first probability threshold interval is [0, 0.33), the second probability threshold interval is [0.33, 0.66), and the third probability threshold interval is [0.66, 1]. Ensure that the agitation level output by the agitation state evolution prediction model has clear and verifiable probability boundaries.

[0132] Example 2

[0133] The difference between Example 2 of the present invention and Example 1 is that this example introduces a system for assessing the state of agitated patients based on big data analysis.

[0134] Figure 2 A schematic structural diagram of a system for assessing the state of an agitated patient based on big data analysis is provided. The system comprises a data processing module, a redundancy analysis module, a cross-validation analysis module, a feature correction module, and a state assessment module.

[0135] Data processing module: collects the original multi-source data streams of patients within a preset time period, and performs feature dimension decomposition and preliminary redundancy marking on the original data in each data stream to obtain the initial feature dimension set of the original multi-source data stream;

[0136] Redundancy analysis module: For the initial feature dimension set, a nonlinear redundancy analysis method is used to identify the feature subsets that produce false emotional feature correlations and form a redundant feature factor set;

[0137] Cross-validation analysis module: performs combined effect cross-validation analysis on redundant feature factor sets, determines the feature dimension combination that produces emotion artifact effects, and generates a set of false emotion association patterns;

[0138] Feature correction module: Based on the set of false emotion association patterns, the feature combinations with emotion artifact effects are removed from the initial feature dimension set of the original multi-source data stream to obtain the corrected valid feature data set;

[0139] State assessment module: Builds an agitation state evolution prediction model based on a valid feature data set, evaluates the patient's current agitation state, and outputs the identification result of the patient's agitation state.

[0140] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0141] The above embodiments can be implemented in whole or in part via software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in the embodiments of this application are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0142] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0143] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0144] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0145] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0146] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0147] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, 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 and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0148] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0149] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A system for assessing the state of agitated patients based on big data analysis, characterized in that: include: Data processing module: collects the original multi-source data streams of patients within a preset time period, and performs feature dimension decomposition and preliminary redundancy marking on the original data in each data stream to obtain the initial feature dimension set of the original multi-source data stream; Redundancy analysis module: For the initial feature dimension set, a nonlinear redundancy analysis method is used to identify the feature subsets that produce false emotional feature correlations and form a redundant feature factor set; Cross-validation analysis module: performs combined effect cross-validation analysis on redundant feature factor sets, determines the feature dimension combination that produces emotion artifact effects, and generates a set of false emotion association patterns; Each feature subset in the redundant feature factor set is randomly combined to form multiple different combination patterns; For each combination pattern, the data of the combination pattern is input into the preset emotional state classifier using the cross-validation method to calculate the degree of interference of each combination pattern on the agitation state classification result; When the interference degree of the combination pattern on the classification of the agitated state exceeds a predetermined interference threshold, the combination pattern is determined to be a characteristic dimension combination that causes an emotional illusion effect; Integrate all the characteristic dimension combinations that appear to have false emotion effects to generate a set of false emotion association patterns; Feature correction module: Based on the set of false emotion association patterns, the feature combinations with emotion artifact effects are removed from the initial feature dimension set of the original multi-source data stream to obtain the corrected valid feature data set; Each feature dimension combination in the false emotion association pattern set is removed from the initial feature dimension set in turn; Perform relevance review analysis on the remaining features of the feature dimension set after removal; Re-label the remaining feature dimension set after correlation review and analysis to obtain a corrected valid feature data set; State assessment module: Builds an agitation state evolution prediction model based on a valid feature data set, evaluates the patient's current agitation state, and outputs the identification result of the patient's agitation state.

2. The system for assessing the state of agitated patients based on big data analysis according to claim 1, characterized in that: Collect multi-source data streams of patients within a preset time period, and perform feature dimension decomposition and preliminary redundancy marking on the original data in each data stream to obtain the initial feature dimension set of the original data stream, specifically: Collect patient heart rate data stream, skin conductance pulse data stream, patient motion frequency data stream and spatial displacement trajectory data stream; Decompose the patient's heart rate data stream into time domain features and frequency domain features; Decompose the skin conductance pulse data stream into the dimensions of pulse amplitude features and peak density features; Decompose the patient's movement frequency data stream into the dimensions of movement number per unit time and movement rhythm; Decompose the spatial displacement trajectory data stream into the dimensions of trajectory curvature and displacement frequency characteristics; The data dimension correlation analysis method is used to perform preliminary redundancy marking on each feature dimension after disassembly to obtain the initial feature dimension set.

3. The system for assessing the state of agitated patients based on big data analysis according to claim 2, characterized in that: The data dimension correlation analysis method is used to perform preliminary redundancy marking on each feature dimension after decomposition, and the initial feature dimension set is obtained, which is specifically: Calculate the Pearson correlation coefficient matrix between each feature dimension after decomposition to determine the degree of linear correlation between feature dimensions; Compare the correlation coefficients between any two feature dimensions in the Pearson correlation coefficient matrix. When the correlation coefficient is greater than the linear correlation threshold, mark the feature dimension with low diagnostic contribution in the two feature dimensions as a preliminary redundant dimension. Classify and label all feature dimensions marked as preliminary redundant dimensions to form a preliminary redundant feature label set; The preliminary redundant feature tag set is mapped and compared with each feature dimension set after disassembly to obtain the initial feature dimension set.

4. The system for assessing the state of agitated patients based on big data analysis according to claim 3, characterized in that: For the initial feature dimension set, a nonlinear redundancy analysis method is used to identify the feature subset that produces false emotional feature correlations, forming a redundant feature factor set, specifically: The kernel function mapping method is used to project the initial feature dimension set into a high-dimensional nonlinear feature space; The nonlinear correlation between the projected feature dimensions is evaluated based on the mutual information index; When the mutual information index between the projected feature dimensions is greater than the nonlinear redundancy threshold, the corresponding feature dimension combination is determined as a feature subset with false emotion feature correlation; All identified feature subsets are integrated to form a redundant feature factor set.

5. The system for assessing the state of agitated patients based on big data analysis according to claim 4, characterized in that: Based on the effective feature data set, a prediction model for the evolution of agitation state is constructed to evaluate the patient's current agitation state and output the identification results of the patient's agitation state, specifically: Taking the effective feature data set as input variables, a time series-based prediction model for the evolution of agitated states is constructed; The parameters of the agitation state evolution prediction model are trained using a historical effective feature data set; Input the patient's effective feature data at the current time node into the agitation state evolution prediction model to calculate the patient's current agitation state probability; The patient's current agitation level is determined according to the agitation probability.

6. The system for assessing the state of agitated patients based on big data analysis according to claim 5, characterized in that: The patient's current agitation level is determined based on the probability of agitation, specifically: Comparing the probability of the agitated state with a preset probability threshold interval; When the probability of the agitation state is within the first probability threshold interval, the patient's current agitation state level is determined to be mild agitation; When the probability of the agitation state is within the second probability threshold interval, the patient's current agitation state level is determined to be moderate agitation; When the probability of the agitation state is within the third probability threshold interval, it is determined that the patient's current agitation state level is severe agitation.

Images