Multi-parameter dynamic assessment of urinary-derived sepsis early warning system and method

Abstract

This invention relates to a multi-parameter dynamic assessment system and method for early warning of urosepsis, belonging to the field of data processing technology. The system includes: a data acquisition and preprocessing module for obtaining unified time-series multi-source pathological data; a feature extraction and interactive evolution modeling module for obtaining a temporal evolution feature representation of the patient's pathological state; a risk fusion assessment module for obtaining the patient's dynamic urosepsis risk score; a risk mapping and graded early warning module for generating corresponding level of visualized early warning information; a risk attribution and clinical intervention prompt module for generating risk attribution results and outputting corresponding clinical intervention prompt information; and an adaptive learning and cross-domain transfer module for online updating of the early warning model based on the doctor's records of handling the visualized early warning information and the patient's final outcome information. This system can improve the accuracy of urosepsis early warning.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a multi-parameter dynamic assessment system and method for early warning of urosepsis. Background Technology

[0002] Ureteral stones are a common emergency in urology. Some patients with stone obstruction are prone to secondary urinary tract infections, which can further develop into urosepsis in severe cases. Urosepsis has an insidious onset and rapid progression. If it is not identified and effectively intervened in time, patients may develop septic shock and multiple organ failure in a short period of time, with a high clinical mortality rate.

[0003] Currently, the clinical assessment of the risk of urosepsis mainly relies on several methods: First, using a universal sepsis scoring system, such as the SOFA score, qSOFA score, and APACHE II score, to assess the risk of patients by comprehensively considering vital signs, organ function, and laboratory indicators; second, using the monitoring of single laboratory indicators, such as changes in inflammatory markers like white blood cell count, procalcitonin (PCT), and C-reactive protein (CRP), to determine the severity of infection.

[0004] However, most existing scoring systems are general models designed for patients with common infections or severe illnesses, and have not been optimized for the specific pathophysiological characteristics of ureteral stones complicated with infection, resulting in insufficient specificity in the identification of urinary sepsis. Secondly, most scoring or detection methods rely on single or intermittent test data, lacking continuous analysis of the dynamic changes in patients' vital signs and inflammatory indicators, making it difficult to capture early disease evolution signals in a timely manner, resulting in poor early warning timeliness. Summary of the Invention

[0005] This invention addresses the technical problems existing in the prior art by providing a multi-parameter dynamic assessment system and method for early warning of urosepsis that can improve the accuracy of early warning.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: This invention provides a multi-parameter dynamic assessment early warning system for urosepsis, the system comprising: The data acquisition and preprocessing module is used to acquire data from patients with ureteral stones, and to preprocess the acquired data to obtain unified time-series multi-source pathological data. The feature extraction and interaction evolution modeling module is used to extract dynamic change features and multi-source index interaction features based on the multi-source pathological data, and encode them through a long temporal neural network to obtain a temporal evolution feature representation of the patient's pathological state. The risk fusion assessment module is used to construct an early warning model based on the temporal evolution characteristics, and to dynamically weight and fuse the output results of each analysis subnetwork in the model through the early warning model to obtain the patient's dynamic urosepsis risk score. The risk mapping and graded early warning module is used to normalize and map the dynamic urosepsis risk score, classify the risk level according to the preset risk threshold, and generate corresponding level of visual early warning information. The risk attribution and clinical intervention prompt module is used to analyze the contribution of key features involved in the calculation based on the dynamic urosepsis risk score, identify major abnormal indicators, generate risk attribution results, and output corresponding clinical intervention prompt information. The adaptive learning and cross-domain transfer module is used to update the early warning model online based on the doctor's handling records of the visualized early warning information and the patient's final outcome information.

[0007] Optionally, the data acquisition and preprocessing module is further used for: Based on the patient's vital signs data, laboratory test data, microbiological test data, stone-related parameters, and imaging characteristics data, multi-source data were collected to obtain raw data; Based on the original data, missing values ​​are filled and outliers are processed to obtain cleaned data; The purified data is standardized and time-axis aligned to obtain the multi-source pathological data.

[0008] Optionally, the feature extraction and interaction evolution modeling module is further used for: Based on the multi-source pathological data, dynamic change characteristics of continuous monitoring indicators are extracted to obtain a dynamic feature set; Based on the multi-source pathological data, multi-source index interaction features are constructed to obtain a fusion feature set; The temporal evolution feature representation is obtained by encoding the fused feature set through a long temporal neural network.

[0009] Optionally, the risk fusion assessment module is also used for: Based on the temporal evolution characteristics, multiple analysis sub-networks are constructed, and the output results of each sub-network are obtained. Calculate the corresponding dynamic weights based on the output results of each subnetwork to obtain the weighting parameters; The output results of each subnetwork are fused based on the weighting parameters to obtain the dynamic urosepsis risk score.

[0010] Optionally, the risk fusion assessment module is also used for: Based on the changing trends of key monitoring indicators, abnormal change characteristics are extracted to obtain change enhancement factors; Structural features were extracted based on the degree of ureteral obstruction and hydronephrosis to obtain anatomical risk factors; The dynamic urosepsis risk score is modified based on the change enhancement factor and the anatomical risk factor to obtain an optimized risk score.

[0011] Optionally, the risk mapping and graded early warning module is also used for: The dynamic urosepsis risk score was normalized to obtain a standardized score. The standardized score is divided into intervals based on a preset risk threshold to obtain the risk level; Based on the risk level, corresponding visual early warning information is generated.

[0012] Optionally, the risk attribution and clinical intervention prompting module is also used for: The contribution value of each feature was obtained by performing feature contribution analysis based on the dynamic urosepsis risk score. Based on the contribution values ​​of each feature, the main abnormal indicators are screened to obtain the key risk factors; Based on the key risk factors, preset clinical rules are matched to generate clinical intervention prompts.

[0013] Optionally, the adaptive learning and cross-domain transfer module is further used for: Based on the doctor's records of handling the visualized early warning information and the patient's final outcome information, feedback data is constructed to obtain training samples; The early warning model is updated online based on the training samples to obtain an updated model. The updated model is adaptively adjusted according to different data distributions to obtain an optimized model.

[0014] Optionally, the collected data includes the patient's vital signs data, laboratory test data, microbiological test data, stone-related parameters, and imaging characteristics data; The early warning model includes a laboratory indicator analysis subnetwork, a vital signs analysis subnetwork, and an image and clinical fusion analysis subnetwork.

[0015] This invention also provides a method for early warning of urosepsis based on multi-parameter dynamic assessment, the method comprising: Data was collected from patients with ureteral stones, and the collected data was preprocessed to obtain multi-source pathological data with unified time series. Dynamic change features and multi-source index interaction features are extracted from the multi-source pathological data and encoded through a long temporal neural network to obtain a temporal evolution feature representation of the patient's pathological state. An early warning model is constructed based on the temporal evolution characteristics, and the output results of each analysis subnetwork in the model are dynamically weighted and fused to obtain the dynamic urosepsis risk score of the patient. The dynamic urosepsis risk score is normalized and mapped, and risk levels are divided according to preset risk thresholds to generate corresponding visual early warning information. Based on the dynamic urosepsis risk score, the contribution of key features involved in the calculation is analyzed and major abnormal indicators are identified. Risk attribution results are generated and corresponding clinical intervention prompts are output. The early warning model is updated online based on the doctor's record of handling the visualized early warning information and the patient's final outcome information.

[0016] Furthermore, to achieve the above objectives, the present invention also proposes an electronic device, comprising: a memory for storing computer software programs; and a processor for reading and executing the computer software programs, thereby realizing a multi-parameter dynamic assessment method for early warning of urosepsis as described above.

[0017] Furthermore, to achieve the above objectives, the present invention also proposes a non-transitory computer-readable storage medium storing a computer software program, which, when executed by a processor, implements a multi-parameter dynamic assessment method for early warning of urosepsis as described above.

[0018] The beneficial effects of this invention are: (1) By integrating vital signs, laboratory indicators, microbial data and imaging and anatomical parameters related to ureteral stones, this invention avoids the one-sidedness of traditional single-indicator or small-scale indicator assessment, thereby significantly improving the comprehensiveness and accuracy of urosepsis risk assessment.

[0019] (2) By extracting dynamic change features from continuous monitoring data and combining them with long-term neural networks for modeling, this invention can capture early trend changes in the disease evolution process. Compared with static assessment methods, it can identify potential deterioration risks in advance.

[0020] (3) By establishing laboratory indicators, vital signs and image-clinical fusion subnetworks respectively and performing dynamic weight fusion, the present invention enables the system to characterize the patient's pathological state from multiple dimensions and improves the modeling ability of complex, multi-factor coupled pathological processes. Attached Figure Description

[0021] Figure 1 A scenario diagram illustrating a multi-parameter dynamic assessment method for early warning of urosepsis provided by this invention; Figure 2A schematic diagram of the structure of a multi-parameter dynamic assessment early warning system for urosepsis provided by the present invention; Figure 3 A flowchart of a multi-parameter dynamic assessment method for early warning of urosepsis provided by the present invention; Figure 4 A schematic diagram of a possible hardware structure of an electronic device provided by the present invention; Figure 5 This is a schematic diagram of the hardware structure of a possible computer-readable storage medium provided by the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0024] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0025] Please see Figure 1 , Figure 1 This is a scenario diagram illustrating a multi-parameter dynamic assessment method for early warning of urosepsis provided by the present invention. Figure 1As shown, the terminal and server are connected via a network, such as a wired or wireless network. The terminal can include, but is not limited to, portable devices such as mobile phones and tablets with various network platform applications installed, as well as fixed terminals such as computers, kiosks, and advertising machines. The server provides users with various business services, including service push servers and user recommendation servers.

[0026] It should be noted that, Figure 1 The scenario diagram of a multi-parameter dynamic assessment method for early warning of urosepsis is merely an example. The terminal, server, and application scenario described in the embodiments of the present invention are for the purpose of more clearly illustrating the technical solutions of the embodiments of the present invention, and do not constitute a limitation on the technical solutions provided by the embodiments of the present invention. As those skilled in the art will know, with the evolution of the system and the emergence of new business scenarios, the technical solutions provided by the embodiments of the present invention are also applicable to similar technical problems.

[0027] The terminal can be used for: Data was collected from patients with ureteral stones, and the collected data was preprocessed to obtain multi-source pathological data with unified time series. Dynamic change features and multi-source index interaction features are extracted from multi-source pathological data and encoded through a long temporal neural network to obtain a temporal evolution feature representation of the patient's pathological state. An early warning model is constructed based on the temporal evolution characteristics, and the output results of each analysis subnetwork in the model are dynamically weighted and fused to obtain the patient's dynamic urosepsis risk score. The dynamic urosepsis risk score is normalized and mapped, and the risk level is divided according to the preset risk threshold to generate corresponding visual early warning information. Based on the dynamic urosepsis risk score, the contribution of key features involved in the calculation is analyzed and major abnormal indicators are identified. Risk attribution results are generated and corresponding clinical intervention prompts are output. The early warning model is updated online based on doctors' records of handling the visualized early warning information and the patient's final outcome information.

[0028] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of a multi-parameter dynamic assessment early warning system for urinary sepsis provided by the present invention.

[0029] like Figure 2 As shown in the embodiment of the present invention, a multi-parameter dynamic assessment early warning system for urosepsis includes: The data acquisition and preprocessing module 201 is used to acquire data from patients with ureteral stones, and to preprocess the acquired data to obtain unified time-series multi-source pathological data.

[0030] In some embodiments, the collected data may include the patient's vital signs data, laboratory test data, microbiological test data, stone-related parameters, and imaging features. The data acquisition and preprocessing module 201 can be used to collect multi-source data from patients with ureteral stones, including but not limited to vital signs, routine tests and biochemical indicators, microbiological test results, stone-related parameters, and imaging features. The module performs missing value imputation, outlier removal, and verification on the collected data, and standardizes the units and dimensions of various data types and aligns them to a timeline based on the admission time. Through the above processing, the module obtains a unified time-series multi-source pathological dataset that can be used for subsequent calculations. In this embodiment, the time-series data can be stored at fixed sampling intervals or in an event-driven manner.

[0031] In some embodiments, the data acquisition and preprocessing module 201 can also be used for: Raw data was obtained by collecting data from multiple sources, including the patient's vital signs, laboratory test data, microbiological test data, stone-related parameters, and imaging characteristics. The original data is imputed and outliers are removed to obtain cleaned data. The purification data was standardized and time-axis aligned to obtain multi-source pathological data.

[0032] In some embodiments, firstly, raw data can be obtained by collecting data from multiple sources, including the patient's vital signs, laboratory test data, microbiological test data, stone-related parameters, and imaging features. Specifically, vital signs data can be obtained through bedside monitoring equipment, including body temperature, heart rate, respiratory rate, blood pressure, and blood oxygen saturation; laboratory test data can be obtained through a laboratory information system, including complete blood count, biochemical indicators, and inflammation-related indicators; microbiological test data includes urine and blood culture results; stone-related parameters include stone size, location, and degree of obstruction; and imaging features can be obtained from an image archiving system, including CT values, stone density, and degree of hydronephrosis. All types of data are recorded with corresponding time information and data source information during collection to form a raw data set with time stamps.

[0033] Secondly, missing value imputation and outlier handling can be performed on the original data to obtain purified data. Specifically, for missing data caused by different detection frequencies or intermittent collection, interpolation methods or imputation methods based on historical data can be used to complete the data; data that deviates significantly from the normal physiological range or is discontinuous with changes at adjacent time points are identified as outliers and are removed or corrected; at the same time, the processing process is marked and recorded to maintain the traceability of data processing, thereby obtaining purified data with high data quality.

[0034] Finally, the purified data can be standardized and time-axis aligned to obtain multi-source pathological data. Specifically, data from different sources and with different dimensions undergo unified unit conversion and numerical standardization to ensure comparability of various indicators; a unified time axis is established based on the patient's admission time or initial assessment time, and data with different sampling frequencies are resampled or interpolated to align, allowing all types of data to be expressed within a unified time dimension; after the above processing, multi-source pathological data with a unified structure and continuous time are formed, providing a foundation for subsequent feature extraction and risk assessment.

[0035] In some embodiments, after completing missing value imputation, outlier handling, and timeline alignment, the system further performs data quality scoring on various types of collected data to dynamically reflect the credibility of data from different sources. Specifically, the system assigns a quality score to each data point or each type of indicator based on data integrity, temporal consistency, degree of abnormal fluctuation, and source reliability, and uses the quality score as a reference weight when inputting subsequent models. When the quality score of a data item is lower than a preset threshold, the system reduces the weight of that data item in risk calculation or marks it as low-credibility data to avoid low-quality data having an excessive impact on subsequent risk assessment results.

[0036] In some implementations, the data quality score can be determined by a combination of missing rate, outlier rate, acquisition latency, device stability, and source type. When the same indicator fluctuates frequently or is supplemented multiple times in a short period of time, the system further reduces its quality score, thereby dynamically decreasing the model's trust in low-quality data.

[0037] The feature extraction and interaction evolution modeling module 202 is used to extract dynamic change features and multi-source index interaction features from multi-source pathological data, and encode them through a long temporal neural network to obtain a temporal evolution feature representation of the patient's pathological state.

[0038] In some embodiments, the feature extraction and interaction evolution modeling module 202 can be used to extract dynamic change features based on multi-source pathological time series data, including the rate of change, trend of change and sliding window statistical features of each continuous monitoring indicator; the module is also used to construct interaction features between different modalities to reflect the coupling relationship between indicators; the above features are input into a long-term neural network for encoding processing to obtain a temporal evolution feature representation that can characterize the evolution of the patient's pathological state over time, and the encoding result is used to reflect the joint distribution of short-term mutations and long-term trends.

[0039] In some embodiments, the feature extraction and interaction evolution modeling module 202 can also be used for: Dynamic feature sets are obtained by extracting dynamic change characteristics of continuous monitoring indicators from multi-source pathological data. Multi-source indicator interaction features were constructed based on multi-source pathological data to obtain a fusion feature set; The temporal evolution feature representation is obtained by encoding the fused feature set through a long temporal neural network.

[0040] In some embodiments, dynamic change characteristics of continuous monitoring indicators can be extracted from multi-source pathological data to obtain a dynamic feature set. Specifically, for vital signs and some continuous test indicators, each indicator is analyzed time-by-time according to a unified time axis to calculate the change amplitude and trend between adjacent time points, so as to reflect the change of indicators over time; at the same time, statistical analysis is performed on each indicator within a preset time window to extract statistical characteristics including mean, fluctuation range and change stability, thereby obtaining a dynamic feature set that can characterize the dynamic change process of the disease.

[0041] Secondly, multi-source indicator interaction features can be constructed based on multi-source pathological data to obtain a fusion feature set. Specifically, data from different sources are correlated and analyzed, and key indicators that can reflect the degree of infection and organ function are combined to form multi-dimensional interaction features. Interaction features may include the correlation between different test indicators, the coupling relationship between vital signs and laboratory indicators, and the joint expression between imaging parameters and clinical indicators, thereby forming a fusion feature set that can comprehensively reflect the patient's overall pathological state.

[0042] Finally, the fused feature set can be encoded using a long-term neural network to obtain a temporal evolution feature representation. Specifically, the fused feature set is input into the long-term neural network model in chronological order to progressively model historical information, thereby extracting the dependencies and evolution patterns of features in the time dimension. Through this encoding process, the original multidimensional time series data is transformed into a temporal evolution feature representation that can characterize the development trend of the patient's condition, which can then be used as input for the subsequent risk assessment model.

[0043] The risk fusion assessment module 203 is used to construct an early warning model based on the temporal evolution characteristics, and to dynamically weight and fuse the output results of each analysis subnetwork in the model through the early warning model to obtain the patient's dynamic urosepsis risk score.

[0044] In some embodiments, the early warning model may include a laboratory indicator analysis subnetwork, a vital signs analysis subnetwork, and an image and clinical fusion analysis subnetwork. The risk fusion assessment module 203 can be used to construct an early warning model based on temporal evolution characteristics. The early warning model includes several analysis subnetworks for different data modalities, used to calculate the local risk quantification results for each modality. The module further dynamically allocates the weights of each subnetwork based on real-time data and performs weighted fusion of the output results of each subnetwork. After weighted fusion, the module outputs a dynamic urosepticemia risk score representing the patient's current disease risk, serving as the basis for subsequent early warning and decision-making.

[0045] In some embodiments, the risk fusion assessment module 203 can also be used for: Multiple sub-networks are constructed based on the temporal evolution characteristics, and the output results of each sub-network are obtained. Calculate the corresponding dynamic weights based on the output results of each subnetwork to obtain the weighting parameters; The outputs of each subnetwork are fused based on the weighted parameters to obtain a dynamic urosepsis risk score.

[0046] In some embodiments, multiple analysis subnetworks can be constructed based on the temporal evolution feature representation to obtain the output results of each subnetwork. Specifically, corresponding analysis subnetworks are set up for different types of data features to characterize patient risk from different dimensions; for example, analysis subnetworks can be set up to process laboratory test features, analysis subnetworks to process vital sign features, and analysis subnetworks to process image and clinical fusion features; the temporal evolution feature representation is input into each analysis subnetwork for independent calculation, and the corresponding risk assessment results are output respectively, thereby obtaining multiple subnetwork output results.

[0047] Secondly, dynamic weights can be calculated based on the output results of each sub-network to obtain weighting parameters. Specifically, based on the changes in the output results of each sub-network and their contribution to the overall risk, corresponding weights are assigned to each sub-network. The weights can be adaptively adjusted according to the changing trend of the input features at the current moment or the importance of different features, so that the sub-networks with a greater impact on risk occupy higher weights in the fusion process, thereby obtaining the weighting parameters corresponding to each sub-network.

[0048] Finally, the outputs of each subnetwork can be fused according to the weighted parameters to obtain a dynamic urosepsis risk score. Specifically, the outputs of each subnetwork are combined according to the corresponding weighted parameters to obtain a unified comprehensive risk value, which is used as the patient's dynamic urosepsis risk score at the current moment for subsequent early warning and clinical decision support.

[0049] In some embodiments, the risk fusion assessment module 203 can also be used for: Based on the changing trends of key monitoring indicators, abnormal change characteristics are extracted to obtain change enhancement factors; Structural features were extracted based on the degree of ureteral obstruction and hydronephrosis to obtain anatomical risk factors; The dynamic urosepsis risk score was modified based on the change enhancement factor and anatomical risk factor to obtain an optimized risk score.

[0050] In some embodiments, abnormal change features can be extracted based on the changing trends of key monitoring indicators to obtain change enhancement factors. Specifically, key monitoring indicators that can reflect the progression of infection and the body's response status are selected from multi-source pathological data, and their changing trends in continuous time series are analyzed; when an indicator shows a significant increase, decrease, or increased fluctuation in a short period of time, it is judged as an abnormal change, and the degree of the abnormal change is quantified to obtain change enhancement factors used to characterize the rapid deterioration trend of the disease.

[0051] Secondly, structural features can be extracted based on the degree of ureteral obstruction and hydronephrosis to obtain anatomical risk factors. Specifically, based on imaging data and clinical records, the degree of obstruction caused by ureteral stones is graded and assessed, and the compression status of the urinary system is comprehensively determined in conjunction with the hydronephrosis. Based on the above assessment results, structural features reflecting the degree of anatomical abnormalities are extracted and quantified to obtain anatomical risk factors.

[0052] Finally, the dynamic urosepsis risk score can be modified based on change-enhancing factors and anatomical risk factors to obtain an optimized risk score. Specifically, change-enhancing factors are introduced into the original risk score to amplify the impact of significant abnormal changes in the short term on the risk, while anatomical risk factors are introduced to reflect the promoting effect of structural factors such as obstruction and hydronephrosis on the development of infection. By modifying the combined effects of the above two types of factors, an optimized risk score that better reflects the patient's actual pathological state is obtained.

[0053] In some embodiments, the dynamic urosepsis risk score can be expressed as: ; in, Representative moment The dynamic urosepsis susceptibility risk score is the result of a comprehensive analysis of multi-source data over a historical period, which can reflect the patient's current and short-term risk of developing urosepsis.

[0054] This indicates the number of discrete sampling points that are traced back to the current moment, which is the length of the time window used for risk assessment. The time window can be set according to the actual monitoring frequency. Represents the first in a discrete time series Each sampling time, Corresponding to the earliest time, Corresponding to the current or most recent moment. Indicates the first The local risk value obtained by the fusion of the multimodal analysis subnetwork at each sampling time point comprehensively reflects the contribution of various data to the risk at that time point. Represents the time decay weight, where This is the attenuation coefficient, used to control the degree of influence of historical data on current risk; when The closer At that time, the larger the weight, the more the model will pay attention to recent data changes. This represents the result of weighted summation of risk values ​​at each moment within the stated time window, used to characterize the basic risk level over a period of time.

[0055] This indicates the maximum change in key monitoring indicators within the stated time window. Specifically, it involves selecting the most significantly changing indicator from a preset set of indicators and quantifying its degree of change to reflect the severity of the disease fluctuations. The change sensitivity coefficient is used to adjust the amplification effect of the maximum change magnitude on the overall risk score. When the monitoring indicators show rapid abnormal changes, this coefficient is used to enhance the risk response. This represents a structural score that quantifies the degree of obstruction and hydronephrosis caused by ureteral stones, and can be assigned a value based on imaging grading or clinical assessment criteria. This is a structural correction coefficient used to adjust the degree of influence of the structural score on the overall risk outcome, so as to reflect the role of anatomical factors in risk assessment.

[0056] The risk mapping and graded early warning module 204 is used to normalize and map the dynamic urosepsis risk score, classify the risk level according to the preset risk threshold, and generate corresponding visual early warning information.

[0057] In some embodiments, the risk mapping and grading early warning module 204 can be used to normalize the dynamic risk score and map it into a standardized risk scale; the module divides the standardized risk scale into several levels (e.g., low, medium, high, and very high) according to a pre-set risk threshold range, and displays the corresponding levels in a visual manner on the terminal interface; the visual early warning information may include color codes, text descriptions, and suggested priorities, which facilitates rapid identification and handling by clinical personnel.

[0058] In some embodiments, the risk mapping and graded early warning module 204 can also be used for: The dynamic urosepsis risk score was normalized to obtain a standardized score; The standardized score is divided into intervals based on a preset risk threshold to obtain the risk level; Generate corresponding visual early warning information based on the risk level.

[0059] In some embodiments, the dynamic urosepticemia risk score can be normalized to obtain a standardized score. Specifically, the risk score output by the model is numerically mapped to a uniform preset range, such as a continuous range from 0 to 100. The normalization process can be based on historical sample distribution or preset mapping rules to ensure the comparability of score results for different patients and at different time points, thereby obtaining a standardized score.

[0060] Secondly, standardized scores can be divided into intervals based on preset risk thresholds to obtain risk levels. Specifically, multiple risk threshold intervals can be set based on clinical experience or statistical analysis results, and standardized scores can be mapped to corresponding intervals, such as low risk, medium risk, high risk, and very high risk levels. Through the above interval division, continuous scoring results are transformed into graded results with clear clinical significance.

[0061] Finally, corresponding visual early warning information can be generated based on the risk level. Specifically, different display methods can be set according to different risk levels, including but not limited to color coding, graphic prompts, and text descriptions; when the risk level reaches a preset threshold, it will be highlighted or a prompt message will be triggered on the terminal interface to provide timely early warning to high-risk patients, thereby assisting medical staff in quickly identifying and taking appropriate measures.

[0062] In some embodiments, the early warning model is updated using one or both of event-driven and periodic updates. In the event-driven update mode, an incremental update is triggered when the system receives records of a doctor's handling of an early warning message, a significant change in a patient's condition, or when the number of new cases reaches a preset threshold, allowing the model to promptly absorb the latest clinical information. In the periodic update mode, the system performs uniform training or parameter fine-tuning on the accumulated samples at preset time intervals, such as daily, weekly, or monthly updates, to ensure the model's performance remains consistently stable.

[0063] To prevent the model from forgetting existing knowledge during the update process, this embodiment employs a protection mechanism to prevent catastrophic forgetting. Specifically, during the update, the system retains a portion of representative historical samples as replay samples, which are then used in training along with the newly added samples to maintain the model's ability to recognize common patterns. Simultaneously, constraints can be set on the model's core parameters to limit excessive deviation from the original parameter distribution during the update process. In some implementations, the system can also perform stratified sampling of samples from different time periods and sources, maintaining a reasonable ratio between new and historical data, thereby preserving model stability while absorbing new knowledge.

[0064] In some embodiments, false alarm suppression processing can be performed on risk results before generating visual warning information. Specifically, the system performs stability assessment on risk scores over a series of consecutive time points. When the risk score rises briefly at a certain time point but does not continuously exceed a preset confirmation threshold, the system does not immediately trigger a high-level warning. Instead, it reviews the changes based on the trends of the preceding and following time points to reduce false alarms caused by a single abnormal fluctuation.

[0065] In some implementations, the system may also introduce a smoothing confirmation mechanism, which requires that the risk score meets the warning conditions within multiple consecutive sampling periods before outputting the corresponding level of warning information, thereby improving the stability and clinical usability of the warning results.

[0066] The risk attribution and clinical intervention prompt module 205 is used to analyze the contribution of key features involved in the calculation based on the dynamic urosepsis risk score, identify major abnormal indicators, generate risk attribution results, and output corresponding clinical intervention prompt information.

[0067] In some embodiments, the risk attribution and clinical intervention prompt module 205 can be used to perform reverse analysis on the calculation process of risk scores to determine key contributing features. The module ranks the features involved in the risk calculation by contribution and identifies major abnormal indicators. Based on the identified key risk factors, the module retrieves a pre-established clinical rule base and generates corresponding intervention suggestion text or operation prompts. Intervention prompts may include suggested examinations, adjustments to anti-infection regimens, or priorities for imaging / interventional procedures to assist physicians in developing individualized treatment plans.

[0068] In some embodiments, the risk attribution and clinical intervention prompting module 205 can also be used for: The contribution value of each feature was obtained by analyzing the feature contribution of the dynamic urosepsis risk score. Key risk factors are obtained by screening the main abnormal indicators based on the contribution values ​​of each feature; Based on key risk factors, pre-defined clinical rules are matched to generate clinical intervention prompts.

[0069] In some embodiments, feature contribution analysis can be performed based on the dynamic urosepsis risk score to obtain the contribution value of each feature. Specifically, based on the calculation process of the risk score, each input feature involved in the model calculation is back-analyzed to quantify the degree of influence of each feature on the current risk score; the degree of influence can be obtained by comparing the score changes before and after feature perturbation or by analyzing based on the internal weights of the model, thereby obtaining the contribution value corresponding to each feature, which is used to characterize the role of different indicators in the current risk formation process.

[0070] Secondly, key anomaly indicators can be screened based on the contribution values ​​of each feature to obtain key risk factors. Specifically, the contribution values ​​of each feature are sorted, and features with a contribution exceeding a preset threshold or ranking high are selected as key anomaly indicators. At the same time, combined with the actual value status of the corresponding features, key factors that have a significant impact on risk improvement are identified, thereby forming a set of key risk factors.

[0071] Finally, clinical intervention prompts can be generated by matching key risk factors with pre-established clinical rules. Specifically, key risk factors are matched with a pre-established clinical rule base, which may include treatment suggestions and priority strategies for different abnormal indicators. When a match is successful, corresponding intervention prompts are generated and output to the terminal interface. The prompts may include suggestions for further examination, suggestions for adjusting the treatment plan, or prompts for emergency treatment to assist medical staff in carrying out targeted interventions.

[0072] In some embodiments, a pre-defined clinical rule base is used to support the conversion of risk attribution results into clinical intervention suggestions. The rule base can be constructed in one or more of the following ways: Based on the construction of an expert system, that is, clinical experts pre-set rule items according to their experience in the diagnosis and treatment of urinary sepsis, the principles of ureteral obstruction management and infection control procedures; Based on guideline mapping, key judgment nodes in current treatment guidelines, consensus documents, or in-hospital treatment pathways are mapped into rules in the form of conditions-conclusions. Learning-based rule construction involves extracting the correspondence between high-frequency treatment patterns, risk factor combinations, and final outcomes from historical cases to form iteratively updatable rule entries.

[0073] In some implementations, the rule base is managed hierarchically according to risk level, abnormal indicator type, and intervention priority. When a key risk factor matches a condition item in the rule base, the system outputs corresponding suggestions, such as prompting to relieve the obstruction as soon as possible, strengthen anti-infection treatment, or further re-examine relevant indicators.

[0074] The adaptive learning and cross-domain transfer module 206 is used to update the early warning model online based on the doctor's records of handling the visualized early warning information and the patient's final outcome information.

[0075] In some embodiments, the adaptive learning and cross-domain transfer module 206 can be used to collect records of clinicians' handling of visualized early warning information and patients' follow-up and final outcome information. The module constructs the above feedback as training samples and uses them to fine-tune or incrementally train the early warning model online. At the same time, the module adopts a domain adaptation strategy to adjust the model parameters or input mapping based on the differences in data distribution from different medical devices or different hospitals, thereby improving the model's generalization ability in heterogeneous environments and ensuring long-term stable prediction performance.

[0076] In some embodiments, the adaptive learning and cross-domain transfer module 206 can also be used for: Training samples were obtained by constructing feedback data based on doctors' records of handling visual early warning information and patients' final outcomes. The early warning model is updated online based on the training samples to obtain the updated model; The updated model is adaptively adjusted based on different data distributions to obtain an optimized model.

[0077] In some embodiments, feedback data can be constructed based on doctors' records of handling visual warning information and patients' final outcome information to obtain training samples. Specifically, the system records the actual handling behavior of doctors after receiving warning information, including whether intervention measures were taken, the type of intervention, and the time of treatment. Combined with the patient's subsequent diagnosis and treatment process and final outcome (such as whether sepsis occurred or whether the patient entered a critical state), the above information is correlated with the input features of the corresponding time period to construct feedback data containing input features, model prediction results, and actual outcome labels, thereby forming training samples for model optimization.

[0078] Secondly, the early warning model can be updated online based on the training samples to obtain an updated model. Specifically, the training samples are input into the early warning model in chronological order or in batches, and the model is updated through incremental learning or parameter fine-tuning. This allows the model to continuously correct its prediction results based on the latest clinical data and feedback information. During the update process, the learning rate, update frequency, and data filtering strategy can be set to ensure the stability and convergence of the model during continuous learning, thereby obtaining the updated model.

[0079] Finally, the updated model can be adaptively adjusted based on different data distributions to obtain an optimized model. Specifically, the updated model is adapted to the differences in data distribution caused by different medical institutions, different equipment, or different populations. For example, the input feature distribution is recalibrated or the model parameters are adjusted to reduce the impact of distribution differences on the prediction results. Through the above adaptive adjustments, the model can maintain relatively stable performance in different application environments, thereby obtaining an optimized model.

[0080] In some embodiments, the adaptive learning and cross-domain transfer module is also used to receive feedback correction information from doctors regarding the warning results. Specifically, when a doctor confirms that a warning is a false alarm, a missed alarm, or requires adjustment of intervention priority, the system records the feedback information along with the corresponding input sample and uses it for subsequent model updates. The system can adjust the weights of specific features based on the doctor's feedback, or retrain the false alarm samples to gradually reduce the probability of false alarms in similar scenarios.

[0081] In some implementations, doctor feedback includes not only whether the warning is valid, but also whether the intervention is effective and whether the risk threshold needs to be adjusted. Based on this, the system synchronously corrects the model output and rule base, thus forming a closed-loop optimization mechanism.

[0082] Please see Figure 3 The present invention provides a flowchart of a multi-parameter dynamic assessment method for early warning of urosepsis, comprising the following steps: Step 301: Collect data from patients with ureteral stones, and preprocess the collected data to obtain unified time-series multi-source pathological data.

[0083] Step 302: Extract dynamic change features and multi-source index interaction features from multi-source pathological data, and encode them through a long temporal neural network to obtain a temporal evolution feature representation of the patient's pathological state.

[0084] Step 303: Construct an early warning model based on the temporal evolution characteristics, and dynamically weight and fuse the output results of each analysis subnetwork in the model through the early warning model to obtain the patient's dynamic urosepsis risk score.

[0085] Step 304: Normalize and map the dynamic urosepsis risk score, classify the risk level according to the preset risk threshold, and generate corresponding visual early warning information.

[0086] Step 305: Analyze the contribution of key features involved in the calculation based on the dynamic urosepsis risk score, identify major abnormal indicators, generate risk attribution results, and output corresponding clinical intervention prompts.

[0087] Step 306: Update the early warning model online based on the doctor's records of handling the visualized early warning information and the patient's final outcome information.

[0088] Please see Figure 4 , Figure 4 This is a schematic diagram illustrating an embodiment of the electronic device provided in this invention. For example... Figure 4 As shown, an embodiment of the present invention provides an electronic device 400, including a memory 410, a processor 420, and a computer program 411 stored in the memory 410 and executable on the processor 420. When the processor 420 executes the computer program 411, it performs the following steps: Data was collected from patients with ureteral stones, and the collected data was preprocessed to obtain multi-source pathological data with unified time series. Dynamic change features and multi-source index interaction features are extracted from multi-source pathological data and encoded through a long temporal neural network to obtain a temporal evolution feature representation of the patient's pathological state. An early warning model is constructed based on the temporal evolution characteristics, and the output results of each analysis subnetwork in the model are dynamically weighted and fused to obtain the patient's dynamic urosepsis risk score. The dynamic urosepsis risk score is normalized and mapped, and the risk level is divided according to the preset risk threshold to generate corresponding visual early warning information. Based on the dynamic urosepsis risk score, the contribution of key features involved in the calculation is analyzed and major abnormal indicators are identified. Risk attribution results are generated and corresponding clinical intervention prompts are output. The early warning model is updated online based on doctors' records of handling the visualized early warning information and the patient's final outcome information.

[0089] Please see Figure 5 , Figure 5 This is a schematic diagram illustrating an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. For example... Figure 5 As shown, this embodiment provides a computer-readable storage medium 500 on which a computer program 411 is stored. When the computer program 411 is executed by a processor, it performs the following steps: Data was collected from patients with ureteral stones, and the collected data was preprocessed to obtain multi-source pathological data with unified time series. Dynamic change features and multi-source index interaction features are extracted from multi-source pathological data and encoded through a long temporal neural network to obtain a temporal evolution feature representation of the patient's pathological state. An early warning model is constructed based on the temporal evolution characteristics, and the output results of each analysis subnetwork in the model are dynamically weighted and fused to obtain the patient's dynamic urosepsis risk score. The dynamic urosepsis risk score is normalized and mapped, and the risk level is divided according to the preset risk threshold to generate corresponding visual early warning information. Based on the dynamic urosepsis risk score, the contribution of key features involved in the calculation is analyzed and major abnormal indicators are identified. Risk attribution results are generated and corresponding clinical intervention prompts are output. The early warning model is updated online based on doctors' records of handling the visualized early warning information and the patient's final outcome information.

[0090] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0091] Those skilled in the art will understand that embodiments of the present invention can be provided as systems, methods, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0093] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0094] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0095] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0096] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A multi-parameter dynamic assessment early warning system for urosepsis, characterized in that, The system includes: The data acquisition and preprocessing module is used to acquire data from patients with ureteral stones, and to preprocess the acquired data to obtain unified time-series multi-source pathological data. The feature extraction and interaction evolution modeling module is used to extract dynamic change features and multi-source index interaction features based on the multi-source pathological data, and encode them through a long temporal neural network to obtain a temporal evolution feature representation of the patient's pathological state. The risk fusion assessment module is used to construct an early warning model based on the temporal evolution characteristics, and to dynamically weight and fuse the output results of each analysis subnetwork in the model through the early warning model to obtain the patient's dynamic urosepsis risk score. The risk mapping and graded early warning module is used to normalize and map the dynamic urosepsis risk score, classify the risk level according to the preset risk threshold, and generate corresponding level of visual early warning information. The risk attribution and clinical intervention prompt module is used to analyze the contribution of key features involved in the calculation based on the dynamic urosepsis risk score, identify major abnormal indicators, generate risk attribution results, and output corresponding clinical intervention prompt information. The adaptive learning and cross-domain transfer module is used to update the early warning model online based on the doctor's handling records of the visualized early warning information and the patient's final outcome information.

2. The multi-parameter dynamic evaluation early warning system for urosepsis according to claim 1, characterized in that, The data acquisition and preprocessing module is also used for: Based on the patient's vital signs data, laboratory test data, microbiological test data, stone-related parameters, and imaging characteristics data, multi-source data were collected to obtain raw data; Based on the original data, missing values ​​are filled and outliers are processed to obtain cleaned data; The purified data is standardized and time-axis aligned to obtain the multi-source pathological data.

3. The multi-parameter dynamic evaluation early warning system for urosepsis according to claim 2, characterized in that, The feature extraction and interaction evolution modeling module is also used for: Based on the multi-source pathological data, dynamic change characteristics of continuous monitoring indicators are extracted to obtain a dynamic feature set; Based on the multi-source pathological data, multi-source index interaction features are constructed to obtain a fusion feature set; The temporal evolution feature representation is obtained by encoding the fused feature set through a long temporal neural network.

4. The multi-parameter dynamic evaluation early warning system for urosepsis according to claim 3, characterized in that, The risk fusion assessment module is also used for: Based on the temporal evolution characteristics, multiple analysis sub-networks are constructed, and the output results of each sub-network are obtained. Calculate the corresponding dynamic weights based on the output results of each subnetwork to obtain the weighting parameters; The output results of each subnetwork are fused based on the weighting parameters to obtain the dynamic urosepsis risk score.

5. The multi-parameter dynamic evaluation early warning system for urosepsis according to claim 4, characterized in that, The risk fusion assessment module is also used for: Based on the changing trends of key monitoring indicators, abnormal change characteristics are extracted to obtain change enhancement factors; Structural features were extracted based on the degree of ureteral obstruction and hydronephrosis to obtain anatomical risk factors; The dynamic urosepsis risk score is modified based on the change enhancement factor and the anatomical risk factor to obtain an optimized risk score.

6. The multi-parameter dynamic evaluation early warning system for urosepsis according to claim 5, characterized in that, The risk mapping and graded early warning module is also used for: The dynamic urosepsis risk score was normalized to obtain a standardized score. The standardized score is divided into intervals based on a preset risk threshold to obtain the risk level; Based on the risk level, corresponding visual early warning information is generated.

7. The multi-parameter dynamic evaluation early warning system for urosepsis according to claim 6, characterized in that, The risk attribution and clinical intervention prompting module is also used for: The contribution value of each feature was obtained by performing feature contribution analysis based on the dynamic urosepsis risk score. Based on the contribution values ​​of each feature, the main abnormal indicators are screened to obtain the key risk factors; Based on the key risk factors, preset clinical rules are matched to generate clinical intervention prompts.

8. The multi-parameter dynamic evaluation early warning system for urosepsis according to claim 7, characterized in that, The adaptive learning and cross-domain transfer module is also used for: Based on the doctor's records of handling the visualized early warning information and the patient's final outcome information, feedback data is constructed to obtain training samples; The early warning model is updated online based on the training samples to obtain an updated model. The updated model is adaptively adjusted according to different data distributions to obtain an optimized model.

9. The multi-parameter dynamic assessment early warning system for urosepsis according to claim 8, characterized in that, The collected data includes the patient's vital signs data, laboratory test data, microbiological test data, stone-related parameters, and imaging characteristics data; The early warning model includes a laboratory indicator analysis subnetwork, a vital signs analysis subnetwork, and an image and clinical fusion analysis subnetwork.

10. A method for early warning of urosepsis based on multi-parameter dynamic assessment, characterized in that, The method includes: Data was collected from patients with ureteral stones, and the collected data was preprocessed to obtain multi-source pathological data with unified time series. Dynamic change features and multi-source index interaction features are extracted from the multi-source pathological data and encoded through a long temporal neural network to obtain a temporal evolution feature representation of the patient's pathological state. An early warning model is constructed based on the temporal evolution characteristics, and the output results of each analysis subnetwork in the model are dynamically weighted and fused to obtain the dynamic urosepsis risk score of the patient. The dynamic urosepsis risk score is normalized and mapped, and risk levels are divided according to preset risk thresholds to generate corresponding visual early warning information. Based on the dynamic urosepsis risk score, the contribution of key features involved in the calculation is analyzed and major abnormal indicators are identified. Risk attribution results are generated and corresponding clinical intervention prompts are output. The early warning model is updated online based on the doctor's record of handling the visualized early warning information and the patient's final outcome information.

Images