Intelligent early identification method and system for ICU patient sepsis

Abstract

This invention relates to the field of personal health risk assessment technology, specifically to a method and system for intelligent early identification of sepsis in ICU patients. The method includes: obtaining the abnormal progression trend of each dimension by analyzing the degree of abnormality in the abnormal periods of the physiological data of ICU patients at the current moment; obtaining the correlation of abnormal trends between any two dimensions by combining the time period overlap characteristics of the intersection of abnormal periods of any two dimensions; clustering each dimension to obtain key clusters; and obtaining an early risk assessment coefficient for sepsis based on the number of dimensions contained in the key clusters, their rate of change, and the correlation of abnormal trends of the key clusters. This invention considers the interference of other factors on the physiological data of each dimension, and obtains the early risk assessment coefficient for sepsis in ICU patients based on the characteristics of the abnormal physiological data caused by interference, which can improve the accuracy of the early risk assessment coefficient for sepsis.

Description

Technical Field

[0001] This invention relates to the field of personal health risk assessment technology, specifically to a method and system for intelligent early identification of sepsis in ICU patients. Background Technology

[0002] Sepsis is a life-threatening organ dysfunction caused by a dysregulation of the host response to infection. It is one of the most common acute and critical illnesses seen in emergency departments and intensive care units (ICUs) worldwide. Therefore, early identification of sepsis in the emergency setting is crucial.

[0003] Existing methods determine the early risk assessment coefficient for sepsis in ICU patients based on the degree of abnormality in various dimensions of their physiological data. Physicians then use this coefficient, combined with other physical data from the ICU patient, for comprehensive analysis to achieve early identification of sepsis. However, the process of obtaining the early risk assessment coefficient does not consider other factors that might interfere with the ICU patient's physiological data, thus affecting the accuracy of the coefficient. Summary of the Invention

[0004] To address the low accuracy of existing early risk assessment coefficients for sepsis in ICU patients, this invention aims to provide an intelligent early identification method and system for sepsis in ICU patients. The specific technical solution adopted is as follows:

[0005] In a first aspect of the present invention, a method for intelligent early identification of sepsis in ICU patients is provided, comprising:

[0006] By analyzing the degree of abnormality in the abnormal periods of various physiological data of ICU patients at the current moment, the progressive trend of abnormality in each dimension can be obtained.

[0007] Determine the time overlap characteristics of the intersection time of the abnormal time periods in any two dimensions, and combine them with the abnormal progression trend to obtain the correlation of abnormal trends in any two dimensions.

[0008] Clustering is performed on each dimension based on the aforementioned abnormal trend correlation to obtain key clusters; the key clusters are those with the highest abnormal trend correlation.

[0009] Based on the number of dimensions contained in the key clusters and their rate of change, as well as the correlation of abnormal trends in the key clusters, the early risk assessment coefficient of sepsis is obtained.

[0010] In an exemplary embodiment, the process of obtaining the abnormal time period includes:

[0011] Obtain the abnormal indicators of any dimension at each time point within a reference time period at the current time; the abnormal indicators represent the differences between the physiological data of any dimension and the normal range;

[0012] The abnormal time periods in any dimension are defined as consecutive abnormal moments in time sequence; the abnormal moment is the moment when the abnormal indicator meets the preset conditions.

[0013] In an exemplary embodiment, the process of obtaining the degree of abnormality includes:

[0014] The degree of abnormality during an abnormal period is determined by the maximum abnormality index and the duration of the abnormal period; the degree of abnormality is positively correlated with both the maximum abnormality index and the duration of the abnormal period.

[0015] In one exemplary embodiment, after obtaining the degree of abnormality, the intelligent early identification method for sepsis in ICU patients further includes:

[0016] Determine the overall increasing trend of abnormal indicators during abnormal periods, and the time interval between the time of the maximum abnormal indicator and its initial time during abnormal periods;

[0017] Based on the overall increasing trend and time interval, the degree of influence of intervention measures during abnormal periods is obtained; the degree of influence is positively correlated with the overall increasing trend and inversely correlated with the time interval.

[0018] The degree of abnormality during the abnormal period is reversed based on the degree of influence to obtain the corrected degree of abnormality during the abnormal period;

[0019] Update the abnormal time period based on the corrected degree of abnormality.

[0020] In an exemplary embodiment, the process of obtaining the abnormal progression trend includes:

[0021] Identify the increasing trend of abnormality during abnormal periods in any dimension;

[0022] Determine the overall interval of abnormal time periods for any of the aforementioned dimensions; the overall interval of abnormal time periods is obtained by the time interval between any two adjacent abnormal time periods.

[0023] Based on the increasing trend of the degree of abnormality and the overall interval of the abnormal period, the progressive trend of abnormality in any dimension is obtained; the progressive trend of abnormality is positively correlated with the increasing trend of the degree of abnormality and negatively correlated with the overall interval of the abnormal period.

[0024] In an exemplary embodiment, the process of obtaining the time period overlap feature includes:

[0025] Determine the percentage of time spent at the intersection of any two dimensions within the abnormal time period.

[0026] Calculate the average of the duration proportions of all intersecting time periods of any two dimensions, and use this as the time period overlap feature of any two dimensions.

[0027] In an exemplary embodiment, the process of obtaining the correlation of the abnormal trend includes:

[0028] Determine the difference in abnormal progressive trends between any two dimensions;

[0029] Based on the time period overlap characteristics and the abnormal progressive trend differences, the abnormal trend correlation between any two dimensions is obtained; the abnormal trend correlation is positively correlated with the time period overlap characteristics and negatively correlated with the abnormal progressive trend differences.

[0030] In an exemplary embodiment, clustering each dimension based on the abnormal trend correlation includes:

[0031] The clustering distance between any two dimensions is obtained from the correlation of the abnormal trends; the clustering distance is inversely correlated with the correlation of the abnormal trends.

[0032] Clustering is performed on each dimension based on the clustering distance between any two dimensions.

[0033] In an exemplary embodiment, the process of obtaining the early sepsis risk assessment coefficient includes:

[0034] Based on the number of dimensions contained in the key cluster at the current moment, the rate of change of the number of dimensions at the current moment, and the correlation of abnormal trends of the key cluster at the current moment, the early risk assessment coefficient of sepsis at the current moment is obtained; the early risk assessment coefficient of sepsis is positively correlated with the number of dimensions, the rate of change, and the correlation of abnormal trends.

[0035] In a second aspect of the present invention, an intelligent early identification system for sepsis in ICU patients is provided, comprising: a memory and a processor; the memory is connected to the processor; the memory is used to store program instructions; the processor is used to implement the above-described intelligent early identification method for sepsis in ICU patients when the program instructions are executed.

[0036] The present invention has the following beneficial effects: When analyzing the physiological data of ICU patients in various dimensions at the current moment, the present invention considers the interference of other factors on the physiological data in various dimensions. Based on the characteristics of the abnormal physiological data caused by interference, the early risk assessment coefficient of sepsis in ICU patients is obtained, which can improve the accuracy of the early risk assessment coefficient of sepsis. Attached Figure Description

[0037] Figure 1 This is a flowchart of an intelligent early identification method for sepsis in ICU patients provided in one embodiment of the present invention;

[0038] Figure 2 This is a flowchart of the process for obtaining abnormal time periods provided in one embodiment of the present invention;

[0039] Figure 3 This is an update flowchart for abnormal time periods provided in one embodiment of the present invention;

[0040] Figure 4 This is a flowchart of the process for obtaining an abnormal progression trend provided in one embodiment of the present invention;

[0041] Figure 5 This is a flowchart of the process for obtaining time period overlap features according to an embodiment of the present invention;

[0042] Figure 6 This is a schematic diagram illustrating the acquisition of the intersection time period provided in one embodiment of the present invention;

[0043] Figure 7 This is a flowchart illustrating the process of obtaining abnormal trend correlations according to an embodiment of the present invention;

[0044] Figure 8 This is a clustering flowchart provided in one embodiment of the present invention. Detailed Implementation

[0045] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the specific implementation methods, structures, features, and effects of the present invention are described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. All data and information collected in this application have been obtained with full consent.

[0047] This embodiment provides an intelligent early identification method for sepsis in ICU patients. The specific scenario addressed is that various interventions can cause temporary fluctuations in the physiological data of ICU patients. To eliminate the "interference" of interventions on physiological data and accurately distinguish between "true abnormalities" caused by the progression of sepsis pathology and "false abnormalities" caused by interventions, this embodiment provides an intelligent early identification method for sepsis in ICU patients that can differentiate between false abnormalities caused by interventions, thereby improving the accuracy of early sepsis risk assessment coefficients.

[0048] Sepsis affects multiple physiological systems in the body, including the cardiovascular, respiratory, immune, and renal systems. Single-dimensional physiological data often fails to fully reflect the complex pathological process of sepsis; for example, blood pressure, heart rate, and respiratory rate can all be affected by sepsis. Therefore, it is necessary to collect physiological data from multiple dimensions of ICU patients simultaneously to improve the accuracy of early sepsis risk assessment.

[0049] In environments such as intensive care units (ICUs), doctors typically use multi-parameter monitors to collect physiological data from ICU patients across multiple dimensions. These monitors can record physiological data in real time, including heart rate, respiratory rate, and blood pressure. It should be understood that each physiological data point has a corresponding normal range; for example, the normal range for body temperature is 36°C-37°C, and the normal range for blood pressure is 90-139 mmHg. Physiological data from all dimensions are sampled simultaneously, with the sampling frequency set according to the actual monitoring needs, such as 500Hz.

[0050] It should be understood that for the multi-dimensional physiological data of ICU patients, the physiological data of each dimension can be standardized to facilitate better data processing. In an exemplary embodiment, this embodiment uses the most commonly used maximum and minimum value normalization method to standardize the physiological data of each dimension. Taking any dimension as an example, the maximum and minimum allowable values ​​in the physiological data of that dimension are obtained, and then the physiological data of that dimension is normalized according to the maximum and minimum allowable values, so that the physiological data of that dimension is stabilized within the numerical range of 0-1, while eliminating dimensions.

[0051] Due to the unique nature of ICU patients' conditions, various medical interventions are necessary to maintain stable vital signs, prevent deterioration, or treat existing complications. For example, for ICU patients at risk of infection, regular suctioning and catheter changes are required. The start time of each intervention should be recorded.

[0052] For the current moment, a reference time period is obtained. In an exemplary embodiment, the current moment is taken as the last moment of the reference time period, and a preset duration is taken as the duration of the reference time period, thus constituting the reference time period for the current moment. The preset duration is set according to actual needs. In an exemplary embodiment, the preset duration is 100 moments, so the reference time period for the current moment is a time period with 100 moments.

[0053] like Figure 1 As shown in the figure, the intelligent early identification method for sepsis in ICU patients provided in this embodiment includes the following steps:

[0054] Step S1: Based on the degree of abnormality in the abnormal periods of the physiological data of ICU patients at the current moment, obtain the progressive trend of abnormality in each dimension;

[0055] Step S2: Determine the time overlap characteristics of the intersection time of any two abnormal time periods, and combine them with the abnormal progression trend to obtain the correlation of abnormal trends in any two dimensions.

[0056] Step S3: Cluster the data across all dimensions based on the correlation of abnormal trends to obtain key clusters;

[0057] Step S4: Based on the number of dimensions contained in the key clusters and their rate of change, as well as the correlation of abnormal trends in the key clusters, obtain the early risk assessment coefficient of sepsis.

[0058] The following detailed explanation of each step, in conjunction with the accompanying drawings, is provided.

[0059] Step S1: Based on the degree of abnormality in the abnormal periods of the physiological data of ICU patients at the current moment, obtain the progressive trend of abnormality in each dimension.

[0060] The onset of sepsis manifests differently in different dimensions of physiological data. Therefore, it is necessary to analyze whether the changes in physiological data in each dimension can clearly and completely reflect the changes in signs caused by sepsis, so as to obtain the abnormality of physiological data in each dimension of ICU patients at the current moment. The greater the degree of abnormality, the more clearly and completely the changes in signs caused by sepsis can be reflected.

[0061] Since the processing procedure for physiological data of ICU patients is the same across all dimensions, for ease of explanation, we will take any one dimension as an example and define it as the w-th dimension.

[0062] First, the abnormal time periods of the w-th dimension of the physiological data of the ICU patient at the current moment are obtained. These abnormal time periods characterize the periods during which abnormal physiological data exists, thus determining the degree of abnormality in each abnormal time period. In an exemplary embodiment, such as... Figure 2 As shown, the following is a specific process for obtaining abnormal time periods:

[0063] Step S11: Obtain the abnormal indicators of any dimension in the reference time period at the current time.

[0064] For the physiological data of the w-th dimension, obtain the abnormal indicators of the physiological data of the w-th dimension at each time point in the reference time period at the current time. The abnormal indicators represent the difference between the physiological data of the w-th dimension and the normal range of the physiological data of the w-th dimension. The greater the difference between the physiological data of the w-th dimension and the corresponding normal range, the larger the abnormal indicator.

[0065] In an exemplary embodiment, the median of the normal range of the physiological data for the w-th dimension is set as the standard value of the physiological data for the w-th dimension. For any moment in the reference time period of the current moment, the absolute value of the difference between the actual value of the physiological data for the w-th dimension at that moment and the standard value of the w-th dimension is calculated, and this absolute value of the difference is used as the anomaly indicator of the w-th dimension at that moment. It should be understood that if the physiological data for each dimension has not been standardized above, then the absolute value of the difference needs to be normalized here, and the result after normalization is used as the anomaly indicator of the w-th dimension at that moment.

[0066] Through the above process, the abnormal indicators of the w-th dimension of physiological data at each time step within the reference time period at the current moment are obtained. For the w-th dimension, preset conditions are determined, and an abnormal indicator is considered large when it meets these preset conditions. Then, it is determined whether the abnormal indicators of the w-th dimension of physiological data at each time step within the reference time period at the current moment meet the preset conditions. If the preset conditions are met, the abnormal indicator is considered large; if not, the abnormal indicator is considered small.

[0067] In an exemplary embodiment, for the w-th dimension, a preset anomaly indicator threshold is determined. The value of this preset anomaly indicator threshold ranges from 0 to 1, and the specific value is set according to the judgment requirements. If a more stringent judgment logic is needed, the preset anomaly indicator threshold can be set smaller, such as 0.5. Furthermore, the preset anomaly indicator thresholds for each dimension are independent of each other, and the preset anomaly indicator thresholds for different dimensions may be different. Therefore, the specific preset condition for the w-th dimension is: the anomaly indicator is greater than or equal to the preset anomaly indicator threshold.

[0068] Compare the abnormal indicators of the physiological data of the w-th dimension at each time point in the reference time period at the current time with the preset abnormal indicator threshold of the w-th dimension. Obtain the time points corresponding to the abnormal indicators that are greater than or equal to the preset abnormal indicator threshold, and use these times points as the abnormal times for the w-th dimension in the reference time period at the current time.

[0069] Step S12: Construct an abnormal time period in any dimension from consecutive abnormal moments in time sequence.

[0070] For the w-th dimension, consecutive anomalous moments within the reference time period at the current moment are considered as anomalous time periods, thus obtaining several anomalous time periods within the reference time period for the w-th dimension. It should be understood that anomalous time periods may differ across dimensions. Furthermore, based on the patterns of physiological data variation, isolated anomalous moments are likely noisy data and are therefore not considered separate anomalous time periods. Time periods within the reference time period other than anomalous time periods are defined as normal time periods.

[0071] For ease of explanation, any abnormal time period in the w-th dimension is designated as the y-th abnormal time period. The maximum abnormal index among the abnormal indicators for each abnormal moment in the y-th abnormal time period is obtained, along with the duration of the abnormal time period, which represents the number of abnormal moments within the y-th abnormal time period. A larger maximum abnormal index indicates a higher degree of abnormality in the y-th abnormal time period; the two are positively correlated. Similarly, a longer duration of the abnormal time period indicates a higher degree of abnormality in the y-th abnormal time period; the two are also positively correlated. Therefore, based on the maximum abnormal index and duration of the abnormal time period, the degree of abnormality in the y-th abnormal time period is obtained. Based on the above logical analysis, a specific calculation method for the degree of abnormality is given below:

[0072] ;

[0073] in, This represents the degree of abnormality in the y-th abnormal time period of the w-th dimension. This represents the maximum anomaly index in the y-th anomaly period of the w-th dimension. This represents the duration of the abnormal period in the y-th abnormal period of the w-th dimension. `norm` represents the normalization function, for example: , where exp represents an exponential function with the natural constant as its base.

[0074] This yields the degree of abnormality for each abnormal time period in the w-th dimension, and further, the degree of abnormality for each abnormal time period in the physiological data of the ICU patient in each dimension at the current moment.

[0075] ICU patients often require frequent interventions, such as suctioning, platelet transfusions, or clotting factor infusions to correct surgical bleeding tendencies. These interventions directly affect physiological indicators such as blood oxygen and heart rate. For example, suctioning may cause a temporary drop in blood oxygen, but this does not indicate sepsis; it is merely a "false abnormality." If the impact of these interventions is not considered, these changes may be mistakenly identified as early signs of sepsis, thus affecting the accuracy of the early sepsis risk assessment coefficient. Accordingly, in an exemplary embodiment, after obtaining the degree of abnormality for each abnormal time period of the ICU patient's physiological data across various dimensions at the current moment, the intelligent early identification method for sepsis in ICU patients provided in this embodiment also includes an update process for the abnormal time periods to remove "false abnormal" periods.

[0076] like Figure 3 As shown, the update process during abnormal periods includes the following steps:

[0077] Step S13: Determine the overall increasing trend of abnormal indicators during the abnormal period, and the time interval between the time corresponding to the time of the maximum abnormal indicator during the abnormal period and its initial time.

[0078] Taking the y-th abnormal time period of the w-th dimension as an example, the overall increasing trend of the abnormal indicators in the y-th abnormal time period is obtained. In an exemplary embodiment, the abnormal indicators at each abnormal moment in the y-th abnormal time period are arranged chronologically and curve fitted to obtain the abnormal indicator change curve of the y-th abnormal time period. Then, the slope of the tangent line at each abnormal moment in the abnormal indicator change curve is obtained, the average value of the slope of the tangent line at each abnormal moment in the abnormal indicator change curve of the y-th abnormal time period is calculated, and this average value is normalized by the sigmoid function. The normalized result is taken as the overall increasing trend of the abnormal indicators in the y-th abnormal time period. The larger the overall increasing trend, the more likely the y-th abnormal time period is a sudden change, more likely to be a rapid response caused by intervention, and not a true abnormality. The degree of influence of the intervention measures on the y-th abnormal time period is higher, and the degree of influence is positively correlated with the overall increasing trend.

[0079] Obtain the abnormal time corresponding to the maximum abnormal indicator in the y-th abnormal time period, and determine the time interval between the abnormal time corresponding to the maximum abnormal indicator and the initial time of the y-th abnormal time period (i.e., the first time of the y-th abnormal time period). The shorter this time interval, the more likely the abnormal indicator in the y-th abnormal time period reached its peak rapidly after the intervention, and the more likely it is a rapid response caused by the intervention rather than a true abnormality. The y-th abnormal time period is more affected by the intervention measures, and the degree of influence is inversely correlated with this time interval.

[0080] Step S14: Based on the overall increasing trend and time interval, determine the degree of impact of intervention measures on abnormal periods.

[0081] Based on the overall increasing trend of the y-th abnormal period and the time interval between the time of the maximum abnormal indicator in the y-th abnormal period and its initial time, the degree of influence of the intervention measures on the y-th abnormal period is obtained. Based on the above logical analysis, a specific calculation method for the degree of influence is given below:

[0082] ;

[0083] in, This indicates the degree of influence of intervention measures on the y-th abnormal time period of the w-th dimension. This indicates the overall increasing trend of the y-th abnormal period in the w-th dimension. This represents the time interval between the moment corresponding to the maximum abnormal indicator in the y-th abnormal time period and its initial moment. This represents the proportion of the time interval between the time of the maximum abnormal indicator in the y-th abnormal time period and its initial time to the duration of the y-th abnormal time period. Essentially, it represents... Normalization.

[0084] Step S15: The degree of abnormality during the abnormal period is reversed based on the degree of influence to obtain the corrected degree of abnormality during the abnormal period.

[0085] The higher the degree of influence of intervention measures on the y-th abnormal time period in the w-th dimension, the more likely the y-th abnormal time period is to be a "false anomaly," and the more necessary it is to reduce the anomaly level of the y-th abnormal time period in the w-th dimension. Therefore, the anomaly level of the y-th abnormal time period is reverse-corrected based on the degree of influence of intervention measures, resulting in the corrected anomaly level of the y-th abnormal time period. In an exemplary embodiment, a specific correction method is given below:

[0086] ;

[0087] in, This represents the corrected anomaly level for the y-th anomaly period in the w-th dimension. Using the above process, the corrected anomaly level for each anomaly period in each dimension is obtained.

[0088] Step S16: Update the abnormal time period according to the corrected abnormality level.

[0089] For the y-th abnormal time period in the w-th dimension, the determination of whether the y-th abnormal time period in the w-th dimension is definitely an abnormal time period is based on the corrected abnormality level. In an exemplary embodiment, an abnormality level threshold is preset. This preset abnormality level threshold is used to determine whether the corrected abnormality level of the y-th abnormal time period is high, thereby determining whether the y-th abnormal time period belongs to an abnormal time period based on the judgment result. The value range of the preset abnormality level threshold is 0-1, and the specific value is set according to the judgment needs. For example, 0.5. If the corrected abnormality level of the y-th abnormal time period is greater than or equal to the preset abnormality level threshold, the y-th abnormal time period is determined to be an abnormal time period; if the corrected abnormality level of the y-th abnormal time period is less than the preset abnormality level threshold, the y-th abnormal time period is determined not to be an abnormal time period and is modified to a normal time period. According to the above process, each abnormal time period in the w-th dimension is traversed to complete the update of the abnormal time periods in the w-th dimension, thereby completing the update of the abnormal time periods in all dimensions. The abnormal time periods mentioned below are all updated abnormal time periods.

[0090] Then, based on the degree of abnormality in the abnormal periods of the ICU patient's physiological data in various dimensions at the current moment, the abnormal progression trend in each dimension is obtained, which represents the severity trend of the abnormality. In an exemplary embodiment, such as... Figure 4 As shown, the following is a specific process for obtaining an abnormal progressive trend:

[0091] Step S17: Determine the increasing trend of abnormality in any dimension during abnormal periods.

[0092] Taking the w-th dimension as an example, the anomaly severity of each abnormal time period in the w-th dimension is arranged chronologically and fitted with a straight line. The slope of the fitted line is then obtained; the larger the slope, the more pronounced the increasing trend of the anomaly severity. This slope is then normalized using the sigmoid function, and the normalized result is taken as the increasing trend of the anomaly severity in the w-th dimension. The stronger the increasing trend of the anomaly severity, the stronger the trend of anomaly progression; the two are positively correlated.

[0093] Step S18: Determine the overall interval of abnormal time periods in any dimension.

[0094] Obtain the time interval between any two adjacent abnormal time periods in the w-th dimension. Then, calculate the average time interval between any two adjacent abnormal time periods in the w-th dimension as the overall interval of abnormal time periods in the w-th dimension. The longer the overall interval of abnormal time periods, the less discontinuous the abnormal time periods in the w-th dimension are, and the smaller the trend of abnormal progression; the shorter the overall interval of abnormal time periods, the more continuous the abnormal time periods are in the w-th dimension, and the greater the trend of abnormal progression. The trend of abnormal progression is inversely correlated with the overall interval of abnormal time periods.

[0095] Step S19: Based on the increasing trend of the degree of abnormality and the overall interval of the abnormal period, obtain the abnormal progression trend in any dimension.

[0096] Based on the increasing trend of the anomaly severity in the w-th dimension and the overall interval of the anomaly period in the w-th dimension, the anomaly progression trend in the w-th dimension is obtained. Based on the above logical analysis, a specific quantification method for the anomaly progression trend is given below:

[0097] ;

[0098] in, This indicates an abnormal progressive trend in the w-th dimension. This indicates an increasing trend in the degree of anomaly in the w-th dimension. This represents the overall interval of the abnormal time period in the w-th dimension.

[0099] Using the above process, the abnormal progression trends in each dimension are obtained.

[0100] Step S2: Determine the time period overlap characteristics of the intersection of abnormal time periods in any two dimensions, and combine the abnormal progression trend to obtain the correlation of abnormal trends in any two dimensions.

[0101] Sepsis typically affects multiple physiological systems, including the cardiovascular, respiratory, metabolic, and coagulation systems, simultaneously or sequentially, through mechanisms such as the spread of inflammatory factors and microcirculatory disturbances. For example, in the early stages of infection, an increased heart rate may initially occur, followed by a decrease in blood pressure due to vasodilation, while respiratory rate increases to compensate for hypoxia. Furthermore, as sepsis progresses, the synchronicity of these multidimensional abnormalities gradually strengthens, evolving from "heart rate abnormality alone" to "synchronous abnormality of heart rate and respiratory rate," and then to "synchronous abnormality of heart rate, blood pressure, and lactate." Analyzing the correlation of abnormal trends across multiple dimensions can provide abnormal signals that truly reflect disease progression for early sepsis identification, improving the accuracy of early sepsis risk assessment coefficients. Moreover, ICU patients may exhibit isolated single-dimensional abnormalities due to underlying diseases and postoperative stress. If these abnormalities do not form a "temporal correlation" with other dimensions, the risk of sepsis is low. However, when multiple dimensions of abnormalities overlap temporally, it indicates an increased likelihood of systemic inflammatory response, and the risk of sepsis is higher.

[0102] First, determine the time overlap characteristics of the intersection time period of any two abnormal time periods. In an exemplary embodiment, such as... Figure 5 As shown, the following is a specific process for obtaining time period overlap features:

[0103] Step S21: Determine the percentage of time of each intersection period of any two dimensions within the abnormal period.

[0104] Let the w-th dimension and the v-th dimension represent any two dimensions, and determine the intersection of the anomalous time periods of the w-th dimension and the v-th dimension. Typically, a particular anomalous time period in the w-th dimension intersects with only one anomalous time period in the v-th dimension; however, in special cases, it may intersect with two adjacent anomalous time periods in the v-th dimension simultaneously. For example... Figure 6 As shown, time axis w represents the w-th dimension, and the thick lines in the w-th dimension represent the various abnormal time periods of the w-th dimension; time axis v represents the v-th dimension, and the thick lines in the v-th dimension represent the various abnormal time periods of the v-th dimension. Time axis wv represents the intersection axis of the w-th and v-th dimensions, where the dashed lines indicate the intersection of the abnormal time periods of the w-th and v-th dimensions, thus obtaining the various intersection time periods of the abnormal time periods of the w-th and v-th dimensions, which are displayed as the time periods formed by the thick lines on the intersection axis.

[0105] Arrange the intersection periods of the abnormal periods in the w-th and v-th dimensions chronologically. For any intersection period, taking the s-th intersection period as an example, obtain the abnormal period in the w-th dimension and the abnormal period in the v-th dimension. Calculate the ratio of the duration of the s-th intersection period to the duration of its abnormal period in the w-th dimension, as its duration percentage for the w-th dimension. Calculate the ratio of the duration of the s-th intersection period to the duration of its abnormal period in the v-th dimension, as its duration percentage for the v-th dimension. Then, obtain the maximum value between the duration percentages for the w-th and v-th dimensions, as the duration percentage of the s-th intersection period within its respective abnormal period. This yields the duration percentage of each intersection period in the w-th and v-th dimensions within its respective abnormal period.

[0106] Step S22: Calculate the average of the duration percentages of all intersecting time periods in any two dimensions, and use this as the time period overlap feature of any two dimensions.

[0107] After obtaining the duration percentage of each intersection period of the w-th and v-th dimensions in the abnormal period, the average duration percentage of each intersection period of the w-th and v-th dimensions in the abnormal period is calculated as the time period overlap feature of the w-th and v-th dimensions.

[0108] Sepsis does not occur suddenly, but rather is a dynamic process that gradually progresses from "local inflammation" to "systemic multi-system dysfunction." For example, in the early stages, there may only be slight fluctuations in body temperature and occasional abnormalities in white blood cells, with a short overlap period, meaning the overlap period accounts for a small proportion of the total abnormal period. As the disease progresses, the periods of sustained fever and significant white blood cell increases gradually overlap, and the overlap period gradually lengthens, meaning the overlap period accounts for a larger proportion of the total abnormal period. This indicates that the sepsis is becoming more severe and requires closer attention.

[0109] Based on the time overlap characteristics of the intersection of abnormal time periods in any two dimensions, and the abnormal progression trend in any two dimensions, the correlation of abnormal trends in any two dimensions is obtained. In an exemplary embodiment, such as... Figure 7 As shown below, a specific process for obtaining the correlation of abnormal trends is given:

[0110] Step S23: Determine the difference in abnormal progressive trends between any two dimensions;

[0111] Step S24: Based on the overlapping time period characteristics and the difference in abnormal progressive trends, obtain the correlation of abnormal trends in any two dimensions.

[0112] Specifically, the difference in the anomalous progression trend between the w-th and v-th dimensions is the absolute value of the difference between their respective anomalous progression trends. The smaller the difference in the anomalous progression trend between the w-th and v-th dimensions, the more similar the severity trends of the anomalousness in the w-th and v-th dimensions, and the stronger the correlation between their anomalous trends. In this case, the correlation between anomalous trends is inversely correlated with the difference in anomalous progression trend.

[0113] The greater the time overlap between the w-th and v-th dimensions, the more severe the temporal overlap between the anomalies in the w-th and v-th dimensions. This makes it more likely that inflammatory responses in the physiological systems corresponding to the w-th and v-th dimensions will occur simultaneously. The stronger the correlation between the abnormal trends in the w-th and v-th dimensions, the more positively correlated the abnormal trend correlation is with the time overlap characteristics.

[0114] Based on the time-period overlap characteristics and abnormal progressive trend differences between the w-th and v-th dimensions, the abnormal trend correlation between the w-th and v-th dimensions is obtained. Based on the above logical analysis, a specific calculation method for the abnormal trend correlation is given below:

[0115] ;

[0116] in, This indicates the correlation between abnormal trends in the w-th and v-th dimensions. This represents the time-segment overlap feature between the w-th and v-th dimensions. This indicates the difference in the abnormal progression trend between the w-th dimension and the v-th dimension.

[0117] By following the above method, the correlation between abnormal trends in any two dimensions can be obtained.

[0118] Step S3: Cluster the data across all dimensions based on the correlation of abnormal trends to obtain key clusters.

[0119] The progression of sepsis follows a "gradual" pattern. In the early stages, it may only manifest as weak synchronous abnormalities in a few dimensions. As the condition worsens, the synchronicity of multidimensional abnormalities gradually increases, and the number of synchronous abnormal dimensions gradually increases. For example, it may develop from "only inflammatory markers are abnormal" to "inflammation + circulatory + respiratory markers are synchronously abnormal." This indicates that the wider the spread of sepsis, the greater the harm to ICU patients, and the more attention is needed.

[0120] Based on the correlation of abnormal trends between any two dimensions, clustering is performed on each dimension, grouping similar dimensions into the same cluster. In an exemplary embodiment, such as... Figure 8 As shown, the following is a clustering process:

[0121] Step S31: Obtain the clustering distance between any two dimensions from the correlation of abnormal trends between any two dimensions.

[0122] Taking the w-th and v-th dimensions as an example, the greater the correlation between their abnormal trends, the shorter their cluster distance. Therefore, the cluster distance is inversely correlated with the correlation of abnormal trends. In an exemplary embodiment, the cluster distance between the w-th and v-th dimensions is equal to... This allows us to obtain the clustering distance between any two dimensions.

[0123] Step S32: Cluster each dimension based on the clustering distance between any two dimensions.

[0124] Based on the clustering distance between any two dimensions, the K-means clustering algorithm is used to cluster each dimension, resulting in several clusters. This ensures that clusters with similar clustering distances are grouped together, meaning that similar abnormal trend correlations are grouped into the same cluster. It should be understood that since abnormal trend correlations are related to two dimensions, each cluster contains several dimensions, and different clusters may contain the same dimensions. The K value in the K-means clustering algorithm can be set manually or determined using the elbow rule or silhouette coefficient method.

[0125] For any cluster, calculate the average of the outlier correlations within that cluster as the outlier correlation of that cluster, thus obtaining the outlier correlation of each cluster. Then, find the cluster with the largest outlier correlation among all clusters, and define the cluster corresponding to the largest outlier correlation as the key cluster at the current moment. Finally, obtain the number of dimensions contained in the key cluster at the current moment.

[0126] Step S4: Based on the number of dimensions contained in the key clusters and their rate of change, as well as the correlation of abnormal trends in the key clusters, obtain the early risk assessment coefficient of sepsis.

[0127] The greater the correlation of the abnormal trends of the key clusters at the current moment, the more synchronous and stronger the abnormal changes of physiological data in multiple dimensions at the current moment, and the greater the risk of sepsis. In other words, the greater the early risk assessment coefficient of sepsis at the current moment, the more positively correlated the early risk assessment coefficient of sepsis at the current moment is with the abnormal trend of the key clusters at the current moment.

[0128] The more dimensions a key cluster contains at a given moment, the wider the spread of the inflammatory response and the more abnormal dimensions involved, indicating a greater risk of sepsis. In other words, the higher the early risk assessment coefficient of sepsis at a given moment, the more positively correlated the early risk assessment coefficient of sepsis at a given moment is with the number of dimensions contained in the key cluster at that moment.

[0129] To obtain the rate of change of the number of dimensions of key clusters at the current moment, in an exemplary embodiment, the process described above is used to obtain the number of dimensions of key clusters at multiple historical moments prior to the current moment. Then, the number of dimensions of key clusters at multiple historical moments and the current moment are sorted and subjected to curve fitting according to time sequence to obtain a curve of change in the number of dimensions. Next, the slope of the tangent line at the current moment in this curve of change in the number of dimensions is obtained, and this slope represents the rate of change of the number of dimensions of key clusters at the current moment. For ease of subsequent processing, the slope of the tangent line at the current moment in this curve of change in the number of dimensions is normalized using the sigmoid function; the normalized result represents the rate of change of the number of dimensions of key clusters at the current moment.

[0130] The greater the rate of change in the number of dimensions of the key clusters at the current moment, the faster the data of the abnormal dimensions at the current moment grows, the more rapidly the sepsis progresses, and the greater the risk of sepsis. In other words, the higher the early risk assessment coefficient of sepsis at the current moment, the more positively correlated the early risk assessment coefficient of sepsis at the current moment is with the rate of change in the number of dimensions of the key clusters at the current moment.

[0131] Therefore, based on the number of dimensions of the key clusters at the current moment, the rate of change of the number of dimensions at the current moment, and the correlation of abnormal trends in the key clusters at the current moment, the early risk assessment coefficient of sepsis at the current moment is obtained. Based on the above logical analysis, the following is a quantification method for the early risk assessment coefficient of sepsis at the current moment:

[0132] ;

[0133] in, This represents the early risk assessment coefficient for sepsis at the current moment. This indicates the rate of change of the number of dimensions of the key clusters at the current moment. This indicates the number of dimensions of the key clusters at the current moment. This represents the total number of dimensions of ICU patients obtained in this embodiment. Indicates to Normalization, This indicates the anomalous trend correlation of key clusters at the current moment.

[0134] This provides the early sepsis risk assessment coefficient for the ICU patient at the current moment. Subsequently, doctors can use this early sepsis risk assessment coefficient, combined with other physical data of the ICU patient, for comprehensive analysis to achieve early identification of sepsis in ICU patients.

[0135] In addition, this embodiment can also set three numerical ranges: a high-priority warning range, a medium-priority warning range, and a low-priority warning range. The upper and lower limits of these three numerical ranges are between 0 and 1, and the specific values ​​are set according to actual needs. As an example, the high-priority warning range is greater than or equal to 0.7, the medium-priority warning range is greater than or equal to 0.4 and less than 0.7, and the low-priority warning range is less than 0.4. The numerical range of the early risk assessment coefficient of sepsis at the current moment is determined. If it is in the high-priority warning range, a high-priority warning signal (such as an audible and visual alarm) is immediately sent to ICU medical staff, and the physiological data of each dimension in the key cluster that triggered the warning is simultaneously displayed. If it is in the medium-priority warning range, a medium-priority warning signal is sent to the ICU medical team (such as a system pop-up window or nurse station screen prompt), and the sampling frequency of the physiological data of each dimension in the key cluster of ICU patients can be increased to enhance the monitoring of ICU patients. If it is in the low-priority warning range, the current monitoring intensity is maintained, and no other additional arrangements are required.

[0136] This embodiment also provides an intelligent early identification system for sepsis in ICU patients, including: a memory and a processor; the memory is connected to the processor, and the memory is used to store program instructions; the processor is used to implement the steps in the above-described intelligent early identification method embodiment for sepsis in ICU patients when the program instructions are executed.

[0137] In one exemplary embodiment, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the above-described embodiment of the intelligent early identification method for sepsis in ICU patients.

[0138] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0139] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. An ICU patient sepsis intelligent early identification method, characterized in that, include: By analyzing the degree of abnormality in various dimensions of physiological data of ICU patients at the current moment, the progressive trend of abnormality in each dimension is obtained. The method for obtaining the degree of abnormality is as follows: determining the overall increasing trend of abnormal indicators within the abnormal period, and the time interval between the moment corresponding to the maximum abnormal indicator within the abnormal period and its initial moment; based on the overall increasing trend and the time interval, the degree of influence of intervention measures on the abnormal period is obtained; the degree of influence is positively correlated with the overall increasing trend and inversely correlated with the time interval; the degree of influence is used to reversely correct the degree of abnormality of the abnormal period, resulting in the corrected degree of abnormality; the abnormal period is updated based on the corrected degree of abnormality. Determine the time overlap characteristics of the intersection time of the abnormal time periods in any two dimensions, and combine them with the abnormal progression trend to obtain the correlation of abnormal trends in any two dimensions. Clustering is performed on each dimension based on the aforementioned abnormal trend correlation to obtain key clusters; the key clusters are those with the highest abnormal trend correlation. Based on the number of dimensions contained in the key clusters and their rate of change, as well as the correlation of abnormal trends in the key clusters, the early risk assessment coefficient of sepsis is obtained.

2. The ICU patient sepsis intelligent early identification method of claim 1, wherein, The process of obtaining the abnormal time period includes: Obtain the abnormal indicators of any dimension at each time point within a reference time period at the current time; the abnormal indicators represent the differences between the physiological data of any dimension and the normal range; The abnormal time periods in any dimension are defined as consecutive abnormal moments in time sequence; the abnormal moment is the moment when the abnormal indicator meets the preset conditions.

3. The intelligent early identification method for sepsis in ICU patients as described in claim 2, characterized in that, The process of obtaining the degree of abnormality includes: The degree of abnormality during an abnormal period is determined by the maximum abnormality index and the duration of the abnormal period; the degree of abnormality is positively correlated with both the maximum abnormality index and the duration of the abnormal period.

4. The intelligent early identification method for sepsis in ICU patients as described in claim 1, characterized in that, The process of obtaining the abnormal progression trend includes: Identify the increasing trend of abnormality during abnormal periods in any dimension; Determine the overall interval of abnormal time periods for any of the aforementioned dimensions; the overall interval of abnormal time periods is obtained by the time interval between any two adjacent abnormal time periods. Based on the increasing trend of the degree of abnormality and the overall interval of the abnormal period, the progressive trend of abnormality in any dimension is obtained; the progressive trend of abnormality is positively correlated with the increasing trend of the degree of abnormality and negatively correlated with the overall interval of the abnormal period.

5. The intelligent early identification method for sepsis in ICU patients as described in claim 1, characterized in that, The process of obtaining the time period overlap feature includes: Determine the percentage of time spent at the intersection of any two dimensions within the abnormal time period. Calculate the average of the duration proportions of all intersecting time periods of any two dimensions, and use this as the time period overlap feature of any two dimensions.

6. The intelligent early identification method for sepsis in ICU patients as described in claim 1, characterized in that, The process of obtaining the correlation of the abnormal trends includes: Determine the difference in abnormal progressive trends between any two dimensions; Based on the time period overlap characteristics and the abnormal progressive trend differences, the abnormal trend correlation between any two dimensions is obtained; the abnormal trend correlation is positively correlated with the time period overlap characteristics and negatively correlated with the abnormal progressive trend differences.

7. The intelligent early identification method for sepsis in ICU patients as described in claim 1, characterized in that, The clustering of each dimension based on the abnormal trend correlation includes: The clustering distance between any two dimensions is obtained from the correlation of the abnormal trends; the clustering distance is inversely correlated with the correlation of the abnormal trends. Clustering is performed on each dimension based on the clustering distance between any two dimensions.

8. The intelligent early identification method for sepsis in ICU patients as described in claim 1, characterized in that, The process of obtaining the early risk assessment coefficient for sepsis includes: Based on the number of dimensions contained in the key cluster at the current moment, the rate of change of the number of dimensions at the current moment, and the correlation of abnormal trends of the key cluster at the current moment, the early risk assessment coefficient of sepsis at the current moment is obtained; the early risk assessment coefficient of sepsis is positively correlated with the number of dimensions, the rate of change, and the correlation of abnormal trends.

9. An intelligent early identification system for sepsis in ICU patients, characterized in that it includes: Memory and processor; The memory is connected to the processor; The memory is used to store program instructions; The processor is configured to implement the intelligent early identification method for sepsis in ICU patients according to any one of claims 1-8 when the program instructions are executed.

Images