A prediction support system for high-risk levels in patients with sepsis

Abstract

The present invention relates to the technical field of disease risk assessment, and specifically to a prediction support system for the high-risk level of sepsis patients. The present invention obtains the importance of each category based on the abnormal degree of physiological data in the abnormal period of each category and the time interval between the abnormal period and its adjacent abnormal period; selects feature categories and target periods based on the importance, and obtains the severity of the target period based on the difference in importance between all categories and feature categories in the target period and the number of categories with abnormal physiological data in the target period; obtains a high-risk value based on the change trend and time interval of the severity between the target period and its adjacent target period, and uses it to adjust and analyze the physiological data of all categories in the time period, thereby predicting the high-risk risk of sepsis in the patient to be tested. The present invention dynamically expands the difference in physiological data in the time period with different abnormal manifestations through the high-risk value, thereby improving the accuracy of predicting high-risk situations of sepsis.

Description

Technical Field

[0001] The present invention relates to the technical field of disease risk assessment, and in particular to a prediction support system for high-risk levels of sepsis patients. Background Art

[0002] Sepsis may develop from a relatively stable state to high-risk conditions such as septic shock in a short period of time. Early identification of high-risk sepsis patients and intervention are the key to treating sepsis patients.

[0003] Existing methods typically use random forest models to predict the risk of high-risk conditions in patients with sepsis. Sepsis is a highly heterogeneous disease. Because the order of symptom onset varies in different patients, the abnormal manifestations of physiological data at different time periods vary. However, some physiological data are relatively close when the patient's condition becomes ill and when the condition improves, making the physiological data during periods with different abnormal manifestations likely to be relatively close. This makes it difficult for the random forest model to effectively distinguish between high-risk and non-high-risk conditions of sepsis, thereby reducing the accuracy of predicting high-risk conditions of sepsis. Summary of the Invention

[0004] In order to solve the technical problem of low accuracy in predicting high-risk sepsis due to similar physiological data during periods of different abnormal manifestations, the present invention aims to provide a prediction support system for the high-risk level of sepsis patients. The technical solutions adopted are as follows:

[0005] The present invention proposes a prediction support system for high-risk levels of sepsis patients, the system comprising:

[0006] A data acquisition module is used to collect several categories of physiological data of the patient at each moment during the analysis period;

[0007] The importance analysis module is used to obtain the abnormal time period of each category; according to the abnormal degree of physiological data in each category of abnormal time period, and the time interval between the abnormal time period and its adjacent abnormal time periods, the importance of physiological data of each category is obtained;

[0008] a severity analysis module configured to select a feature category based on the importance, record the abnormal period of the feature category as a target period, and obtain the severity of each target period based on the difference in importance between all categories and the feature category in each target period, and the number of categories of abnormal physiological data in each target period;

[0009] The sepsis high-risk prediction module is used to obtain a high-risk value for each target period based on the severity change trend and time interval between each target period and its adjacent target period; and predict the high-risk risk of sepsis for the patient to be tested based on all categories of physiological data in the analysis time period adjusted by the high-risk value.

[0010] Furthermore, obtaining the abnormal time period of each category includes:

[0011] Normalize the absolute value of the difference between the physiological data of each category at each moment and the normal physiological preset value to obtain the abnormality degree of the physiological data of each category at each moment;

[0012] The abnormal moments of each category are selected based on the abnormality degree; and the time period consisting of the continuous abnormal moments of each category is used as the abnormal period of each category.

[0013] Furthermore, obtaining the importance of each category of physiological data includes:

[0014] For each abnormal period of each category, the average of the abnormality degrees of the physiological data of each category at all times within the abnormal period is used as the abnormal comprehensive value of the abnormal period; the sum of the time intervals between the abnormal period and all its first adjacent periods is used as the first period interval value of the abnormal period; and the cumulative sum of the ratios of the abnormal comprehensive value of all abnormal periods of each category to the first period interval values ​​is calculated as the local importance value of the physiological data of each category;

[0015] The proportion of the abnormal moments of each category in the analysis time period is used as the abnormal occupancy value of each category;

[0016] The importance of each category of physiological data is obtained according to the local importance value and the abnormal occupancy value, and both the local importance value and the abnormal occupancy value are positively correlated with the importance.

[0017] Furthermore, obtaining the severity of each target period includes:

[0018] Obtaining an abnormal category in each target time period, wherein the abnormal category exists and the abnormal moment is within each target time period; obtaining a correlation between the physiological data of each abnormal category and the characteristic category in each target time period based on the difference in importance between each abnormal category and the characteristic category in each target time period, and the number of abnormal moments of each abnormal category in each target time period;

[0019] Selecting a reference category from the abnormal categories in each target time period based on the correlation;

[0020] The severity of each target period is obtained according to the correlation between the reference category and the feature category in each target period, the importance of the reference category, and the number of reference categories in each target period.

[0021] Furthermore, obtaining the correlation between the physiological data of each abnormal category and the characteristic category in each target time period includes:

[0022] The proportion of abnormal moments of each abnormal category in each target time period is counted; the absolute value of the difference between the importance of each abnormal category and the feature category in each target time period is negatively correlated and mapped, and the product of the mapping result and the proportion is used as the correlation between the physiological data of each abnormal category and the feature category in each target time period.

[0023] Furthermore, obtaining the severity of each target period according to the correlation between the reference category and the feature category in each target period, the importance of the reference category, and the number of reference categories in each target period includes:

[0024] The product of the correlation between each reference category and the feature category and the importance of the reference category in each target period is used as the reference severity value of each reference category in each target period; the cumulative sum of the reference severity values ​​of all reference categories in each target period is used as the local severity value of each target period;

[0025] The severity of each target period is obtained according to the total number of abnormal categories and the local severity value in each target period; the total number of abnormal categories and the local severity value in each target period are both positively correlated with the severity.

[0026] Furthermore, obtaining the high-risk value for each target period includes:

[0027] Performing a straight line fit on the severity of each target period and its second adjacent period, normalizing the slope of the fitted line to obtain a severity trend value for each target period; and taking the sum of the time intervals between each target period and all its second adjacent periods as the second period interval value for each target period;

[0028] According to the severity of each target time period, the severe trend value and the second time period interval value, a high-risk value for each target time period is obtained; the severity and the severe trend value are both positively correlated with the high-risk value, and the second time period interval value is negatively correlated with the high-risk value.

[0029] Furthermore, the method of predicting the high risk of sepsis of the patient to be tested based on all categories of physiological data within the analysis time period adjusted by the high risk value includes:

[0030] The sum of the high-risk value and a constant 1 is used as an adjustment coefficient for each target period; all categories of physiological data within the target period are weighted based on the adjustment coefficient to obtain adjusted physiological data for each category at each moment within the target period;

[0031] All categories of adjusted physiological data of the patients to be tested during the analysis period are input into the trained random forest model to predict the high-risk risk of sepsis in the patients to be tested.

[0032] Furthermore, obtaining the high-risk value for each target period includes:

[0033] Performing a straight line fit on the severity of each target period and its second adjacent period, normalizing the slope of the fitted line to obtain a severity trend value for each target period; and taking the sum of the time intervals between each target period and all its second adjacent periods as the second period interval value for each target period;

[0034] According to the severity of each target time period, the severe trend value and the second time period interval value, a high-risk value for each target time period is obtained; the severity and the severe trend value are both positively correlated with the high-risk value, and the second time period interval value is negatively correlated with the high-risk value.

[0035] Furthermore, the importance of the feature category is a maximum value among the importances of all categories.

[0036] Furthermore, the straight line fitting method is the least squares method.

[0037] The present invention has the following beneficial effects:

[0038] In the embodiment of the present invention, the time interval between each abnormal period and its adjacent abnormal period and the abnormal degree of physiological data in the abnormal period are used to analyze the continuous abnormal changes of physiological data of each category during the onset of sepsis and the abnormal degree of physiological data, and obtain the importance of the abnormal changes in the sepsis signs. Sepsis has many complications. In order to accurately analyze the high-risk state of sepsis, feature categories and target periods are selected. The difference in importance between each category and feature category in the target period shows the abnormal changes in physical signs caused by sepsis in each category. Combined with the number of categories of abnormal physiological data in the target period, the sepsis symptoms at the target time are analyzed. The severity of the target period is obtained by calculating the degree of impact of the target period on the patient's health; the changing trend of the severity of the target period and its adjacent target period shows the good or bad condition of the patient's sepsis during the target period, and combined with the time interval between the target period showing continuous abnormal conditions and its adjacent target period, the possibility of the patient being at high risk of sepsis during the target period is analyzed, and a high-risk value is obtained. The high-risk value is used to dynamically expand the difference in physiological data of periods with different abnormal manifestations, emphasize the physiological data of periods with greater abnormality, improve the recognition between physiological data of different abnormal manifestations, and thus improve the accuracy of prediction of high-risk sepsis. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 This is a system structure diagram of a prediction support system for high-risk levels of sepsis patients provided by one embodiment of the present invention;

[0041] Figure 2 A structural diagram of a severity analysis module provided by one embodiment of the present invention;

[0042] Figure 3 A schematic diagram of a computer device for predicting the high-risk level of a sepsis patient provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0043] To further illustrate the technical means and effectiveness of the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effectiveness of a system for predicting the high-risk level of sepsis patients proposed by the present invention. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0044] Unless defined otherwise, 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 belongs.

[0045] The specific scheme of the prediction support system for high-risk level of sepsis patients provided by the present invention is described in detail below with reference to the accompanying drawings.

[0046] Example 1:

[0047] See also Figure 1 , which shows a system block diagram of a prediction support system for the high-risk level of sepsis patients provided by an embodiment of the present invention. The system includes: a data acquisition module 110, an importance analysis module 120, a severity analysis module 130, and a sepsis high-risk prediction module 140.

[0048] The data collection module 110 is used to collect several categories of physiological data of the patient at each moment in the analysis time period.

[0049] Sepsis presents with symptoms such as high fever or low temperature, increased heart rate, rapid breathing, and decreased blood pressure. In this embodiment, the categories of physiological data include body temperature, heart rate, blood pressure, and respiratory rate. A temperature sensor is used to measure body temperature, an electrocardiogram monitor to measure heart rate, a sphygmomanometer to measure blood pressure, and a respiratory rate sensor to measure respiratory rate. Physiological data of each category is obtained for the patient at each moment during the analysis period. The patient is suffering from sepsis.

[0050] In one implementation of the embodiment of the present invention, the data acquisition frequency of the temperature sensor, electrocardiogram monitor, blood pressure monitor and respiratory rate sensor is the same. The data acquisition frequency of the above instruments in the embodiment of the present invention is once every 5 seconds, and the duration of the analysis time period is an empirical value of 1 hour. The implementer can set it according to the specific situation.

[0051] The importance analysis module 120 is used to obtain the abnormal time period of each category; and obtain the importance of each category of physiological data according to the abnormal degree of physiological data in each category of abnormal time period and the time interval between the abnormal time period and its adjacent abnormal time periods.

[0052] There are large differences in the magnitude, range, and distribution of physiological data of different categories. In order to better analyze the abnormal changes of physiological data of each category, the time period of abnormal changes of physiological data of each category is screened to obtain the abnormal time period of the corresponding category.

[0053] Because different categories of physiological data exhibit different performance during sepsis, it's necessary to analyze whether changes in each category clearly reflect the changes in physical signs caused by sepsis. During sepsis, each category of physiological data will continuously exhibit abnormal changes, and these changes may deviate from those of healthy individuals. The importance of each category of physiological data is determined based on the time interval between each abnormal period and its adjacent abnormal periods, as well as the degree of abnormality within each abnormal period. The greater the importance, the more each category of physiological data reflects the abnormal changes in physical signs caused by sepsis, and the more important this category of physiological data is for analyzing abnormal changes in physical signs caused by sepsis.

[0054] The severity analysis module 130 is used to select feature categories based on importance and record the abnormal period of the feature category as the target period; according to the difference in importance between all categories and the feature category in each target period, and the number of categories of abnormal physiological data in each target period, the severity of each target period is obtained.

[0055] Sepsis has numerous complications, and when sepsis reaches a high-risk stage, the onset of symptoms varies from patient to patient. For example, some patients may experience abnormalities in only one physiological data category early in the disease, followed by abnormalities in multiple categories later on; others may experience abnormalities in multiple categories early in the disease. Regardless of the onset of the disease, the earlier the physiological data with the most obvious abnormalities are detected, the more accurate the high-risk warning for sepsis patients can be.

[0056] Categories with greater importance are more crucial for analyzing abnormal changes in sepsis-related physical signs. Changes in physiological data from these categories can more quickly and accurately reflect a high-risk sepsis state. Therefore, the category with the highest importance among all categories of physiological data is considered the feature category, and the abnormal period of this feature category is recorded as the target period.

[0057] Sepsis is characterized by the concurrent occurrence of multiple symptoms, that is, different symptoms gradually develop. The more complex the categories of abnormal physiological data, the more serious the impact of sepsis on the patient's health. Among them, the physiological data of the feature category can best reflect the abnormal changes in physical signs caused by sepsis. The difference in importance between each category and the feature category in the target time period shows that each category reflects the abnormal changes in physical signs caused by sepsis. If the category of abnormal physiological data in the target time period can clearly reflect the abnormal changes in physical signs caused by sepsis, the more serious the impact of sepsis on the patient's health in the target time period is, and the severity of the target time period is obtained.

[0058] The sepsis high-risk prediction module 140 is configured to obtain a high-risk value for each target period based on the severity change trend and time interval between each target period and its adjacent target periods; and predict the high-risk risk of sepsis for the patient to be tested based on all categories of physiological data within the analysis time period adjusted using the high-risk value.

[0059] The condition of sepsis changes dynamically, and the body of a sepsis patient is usually in a continuous abnormal state. The changing trend of the severity of the target period and its adjacent target period shows the severity of the patient's sepsis condition during the target period. Combined with the time interval between the target period showing a continuous abnormal state and its adjacent target period, the possibility of the patient being at high risk of sepsis during the target period is analyzed to obtain a high-risk risk value.

[0060] The higher the high-risk value, the greater the risk of sepsis occurring in the target period. The high-risk value is used to adjust all categories of physiological data in the target period, adaptively expand the differences in physiological data in periods with different abnormal manifestations, enhance the generalization ability of the model, facilitate timely detection of dynamic changes in the disease, and improve the accuracy of sepsis high-risk level prediction.

[0061] Preferably, in some implementations of the embodiments of the present invention, the abnormal period is obtained by normalizing the absolute value of the difference between the physiological data of each category at each moment and the normal physiological preset value to obtain the abnormality degree of the physiological data of each category at each moment; selecting the abnormal moment of each category based on the abnormality degree; and defining the time period consisting of consecutive abnormal moments of each category as the abnormal period of each category. The normal physiological preset value of each category refers to the physiological data of a healthy population. The greater the difference between the physiological data of each category at each moment and the normal physiological preset value, the greater the possibility that the physiological data deviates from the healthy population, and the greater the degree of abnormality of the physiological data.

[0062] In a specific manner of the embodiment of the present invention, for the abnormality of each category at all moments within the analysis time period, the moment corresponding to the abnormality greater than a preset abnormality threshold is used as the abnormal moment of each category.

[0063] In a specific embodiment of the present invention, the normal physiological preset values ​​of body temperature, heart rate, blood pressure and respiratory rate are taken as the empirical values ​​of 36.5 degrees Celsius, 80 times per minute, 114.5 mmHg and 16 times per minute, and the preset abnormal threshold is taken as the empirical value of 0.6, which can be set by the implementer according to the specific situation; the Norm function is used for normalization.

[0064] Preferably, in some possible implementation methods of the embodiments of the present invention, the method for obtaining the importance includes: for each abnormal time period of each category, taking the average of the abnormality of the physiological data of each category at all moments in the abnormal time period as the abnormal comprehensive value of the abnormal time period; taking the sum of the time intervals between the abnormal time period and all its first adjacent time periods as the first time period interval value of the abnormal time period; calculating the cumulative sum of the ratios of the abnormal comprehensive value of all abnormal time periods of each category to the first time period interval value as the local importance value of the physiological data of each category; taking the proportion of the abnormal moments of each category in the analysis time period as the abnormal occupancy value of each category; and obtaining the importance of the physiological data of each category based on the local importance value and the abnormal occupancy value.

[0065] The abnormal comprehensive value reflects the overall abnormality of the physiological data in each category during the abnormal period. The larger the abnormal comprehensive value, the more serious the abnormality of the physiological data in each category during the abnormal period, and the more it deviates from the physiological data of the healthy population; the smaller the first period interval value, the more continuous the abnormal period of each category. The physiological data of all abnormal periods of each category are integrated, and the importance of the physiological data of each category to the analysis of abnormal changes in physical signs caused by sepsis is considered to obtain the local importance value of the physiological data of each category. The larger the local importance value, the more important the physiological data of each category is to the analysis of abnormal changes in physical signs; the larger the abnormal occupancy value, the more physiological data with abnormal changes in each category, and the more serious the abnormality of the physiological data of each category. Therefore, the local importance value and the abnormal occupancy value are positively correlated with the importance. The specific formula is expressed as:

[0066] ;

[0067] Where H is the importance of each category of physiological data; N is the total number of abnormal moments in each category; Num is the total number of all moments in the analysis time period; is the abnormal occupancy value of each category; A is the total number of abnormal time periods of each category; is the abnormal comprehensive value of the a-th abnormal period for each category; is the first period interval value of the a-th abnormal period of each category; is the local importance value of physiological data for each category.

[0068] In a specific embodiment of the present invention, the adjacent previous and next abnormal time periods of each abnormal time period of each category are recorded as first adjacent time periods; the time interval between any two abnormal time periods of each category is equal to the length of time between the end time of the previous abnormal time period and the start time of the next abnormal time period.

[0069] In a specific embodiment of the present invention, the category corresponding to the maximum importance value of all categories of physiological data is used as the feature category.

[0070] See also Figure 2 , which shows a structural diagram of a severity analysis module provided by an embodiment of the present invention. The severity analysis module includes: a correlation analysis unit 131, a reference category screening unit 132, and a severity analysis unit 133.

[0071] Correlation analysis unit 131: used to obtain the abnormal category in each target time period, and the abnormal moment of the abnormal category in each target time period; according to the difference in importance between each abnormal category and the feature category in each target time period, and the number of abnormal moments of each abnormal category in each target time period, obtain the correlation between the physiological data of each abnormal category and the feature category in each target time period.

[0072] Because the onset of sepsis varies among patients, not all categories of physiological data within the target period can reflect abnormal changes in physical signs. Therefore, it is necessary to screen for categories with abnormal physiological data within the target period to obtain abnormal categories. Physiological data from characteristic categories are most important for analyzing abnormal changes in physical signs caused by sepsis. The closer the importance of the abnormal category and physiological data from characteristic categories in analyzing abnormal changes in physical signs caused by sepsis during the target period, and the closer the abnormal duration of the abnormal category within the target period is to the duration of the target period, the greater the correlation between the abnormal category and physiological data from characteristic categories during the target period, and the correlation degree is obtained.

[0073] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the correlation includes: counting the proportion of abnormal moments of each abnormal category in each target time period in the corresponding target time period; performing negative correlation mapping on the absolute value of the difference between the importance of each abnormal category and the feature category in each target time period, and multiplying the mapping result by the proportion as the correlation between the physiological data of each abnormal category and the feature category in each target time period. In the embodiment of the present invention, the negative of the absolute value of the difference between the importance of each abnormal category and the feature category in the target time period is first taken, and the negative number is used as the exponent of an exponential function with a natural constant as the base to achieve negative correlation mapping of the absolute value of the difference; in other embodiments, negative correlation mapping can also be achieved by taking the negative number and function conversion.

[0074] The greater the proportion of abnormal moments for each abnormal category within the target period, the closer the abnormal duration of each abnormal category within the target period is to the duration of the target period, and the higher the correlation between each abnormal category and the feature category within the target period. The smaller the difference in importance between each abnormal category and the feature category within each target period, and the closer the importance of the abnormal category is to the importance of the feature category, the higher the correlation between each abnormal category and the feature category within the target period.

[0075] Reference category screening unit 132: selects a reference category from the abnormal categories in each target time period based on the correlation.

[0076] To reduce computational complexity and improve the accuracy of sepsis high-risk analysis, a reference category with a high correlation with the physiological data of the characteristic category within the target period was selected from the abnormal categories. For each target period, the correlation between all abnormal categories and the characteristic category was greater than a preset correlation threshold, which was used as the reference category for each target period.

[0077] In a specific embodiment of the present invention, the preset relevant threshold is an empirical value of 0.7, and the implementer can set it according to the specific situation.

[0078] The severity analysis unit 133 is configured to obtain the severity of each target period according to the correlation between the reference category and the feature category, the importance of the reference category, and the number of reference categories in each target period.

[0079] Preferably, in some possible implementation methods of the embodiments of the present invention, the method for obtaining the severity includes: taking the product of the correlation between each reference category and the feature category and the importance of the reference category in each target time period as the reference severity value of each reference category in each target time period; taking the cumulative sum of the reference severity values ​​of all reference categories in each target time period as the local severity value of each target time period; and obtaining the severity of each target time period based on the total number of abnormal categories and the local severity value in each target time period.

[0080] The greater the correlation between the reference category and the feature category during the target period, and the greater the importance of the reference category during the target period, the more severe the patient's physical abnormalities during the target period, and the more significant the abnormal changes in the reference category's physiological data during the target period. This indicates that sepsis has a more severe impact on the patient's health during the target period, and the larger the local severity value. The greater the total number of abnormal categories during the target period, the richer the categories of physiological data experiencing abnormal changes during the target period, and the more severe the impact of sepsis on the patient's health during the target period. Therefore, the total number of abnormal categories and the local severity value in each target period are positively correlated with the severity.

[0081] In a specific implementation of the embodiment of the present invention, the severity of each target period is expressed as follows:

[0082] ;

[0083] Where Z is the severity of each target period; M is the total number of abnormal categories in each target period; B is the total number of reference categories in each target period; is the correlation between the bth reference category and the feature category in each target period; is the importance of the bth reference category in each target period; is the reference severity value for each reference category under each target period; is the local severity value of each target period; Norm is the normalization function.

[0084] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining high-risk values ​​includes: performing linear fitting on the severity of each target period and its second adjacent period, normalizing the slope of the fitted line, and obtaining the severity trend value of each target period; taking the sum of the time intervals between each target period and all its second adjacent periods as the second period interval value of each target period; and obtaining the high-risk value of each target period based on the severity, severity trend value, and second period interval value of each target period. In a specific implementation of the embodiments of the present invention, a two-dimensional coordinate system is established with time as the horizontal axis and severity as the vertical axis, and the severity of each target period and its second adjacent period is mapped into the two-dimensional coordinate system to obtain the corresponding coordinate points, and then the least squares method is used to perform linear fitting to obtain a fitted straight line. It should be noted that the horizontal coordinate of the coordinate point of the target period is equal to the middle moment of the target period.

[0085] The greater the severity, the greater the likelihood that a patient will develop a high-risk sepsis condition during the target period. A larger severity trend value indicates a gradually increasing likelihood of developing a high-risk sepsis condition within the local area of ​​the target period, indicating a rapid deterioration of sepsis during the target period. The smaller the second-period interval value, the greater the patient's continuous abnormal state within the local area of ​​the target period, the greater the patient's continued exposure to sepsis-induced inflammation, and the greater the likelihood that sepsis will develop a high-risk sepsis condition during the target period. Therefore, both severity and severity trend values ​​are positively correlated with high-risk values, while second-period interval values ​​are negatively correlated with high-risk values.

[0086] The high-risk value F for each target period is expressed as follows:

[0087] ;

[0088] Where Z is the severity of each target period; k is the slope of the fitted line obtained by linear fitting of the severity of each target period and its second adjacent period; is the severe trend value of each target period; T2 is the second period interval value of each target period; exp is the exponential function with a natural constant as the base; Norm is the normalization function.

[0089] In one specific embodiment of the present invention, the first three and last three target periods of each target period are recorded as second adjacent periods. It should be noted that in this embodiment, the number of first adjacent periods of the abnormal period is less than the number of second adjacent periods of the target period. Therefore, the high-risk values ​​of the first three and last three target periods within the analysis time period are directly set to a constant of 1.

[0090] In one implementation of an embodiment of the present invention, a method for predicting a high-risk risk of sepsis is as follows: using the sum of a high-risk risk value and a constant of 1 as an adjustment coefficient for each target period; weighting all categories of physiological data within the target period based on the adjustment coefficient to obtain adjusted physiological data for each category at each moment within the target period; and inputting all categories of adjusted physiological data of a patient to be tested within the analysis time period into a trained random forest model to predict the patient's high-risk risk of sepsis.

[0091] The training process of the random forest model is as follows: physiological data of several categories from two training patients over a certain time period is obtained. Training patients at high risk for sepsis are labeled as high risk, while training patients at non-high risk for sepsis are labeled as non-high risk. This physiological data is then input into the random forest model to train the model. Adjusted physiological data from all categories of the test patients during the analysis time period are then input into the trained random forest model. The model output is the predicted probability of each label. When the predicted probability of the high-risk label is lower, the likelihood that the test patient is at a high risk level for sepsis is greater. The training process of the random forest model is well known to those skilled in the art and will not be described in detail here.

[0092] It should be noted that all categories of adjusted physiological data within the analysis period include all categories of adjusted physiological data within the target period and all categories of physiological data at other times within the analysis period except the target period.

[0093] So far, the present invention is completed.

[0094] Example 2:

[0095] Figure 3 A computer device diagram of a device for predicting the high-risk level of sepsis patients provided by an embodiment of the present invention. For example, Figure 3As shown, the computer device includes: a memory 201, a processor 202, and a computer program 203 stored in the memory 201 and running on the processor 202, wherein when the processor 202 executes the computer program 203, the computer device can execute any one of the aforementioned prediction support systems for high-risk levels of sepsis patients.

[0096] In addition, an embodiment of the present application also protects a device, which may include a memory and a processor, wherein the memory stores executable program code, and the processor is used to call and execute the executable program code to execute a prediction support system for high-risk levels of sepsis patients provided by an embodiment of the present application.

[0097] In this embodiment, the device can be divided into functional modules based on the above-described method examples. For example, each functional module can be mapped to a specific functional module, or two or more functions can be integrated into a single processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and represents only a logical functional division. In actual implementation, other division methods may be used.

[0098] It should be understood that the device provided in this embodiment is used to implement the above-mentioned prediction support system for high-risk level of sepsis patients, and thus can achieve the same effect as the above-mentioned implementation method.

[0099] In the case of an integrated unit, the device may include a processing module and a storage module. When the device is applied to a device, the processing module may be used to control and manage the operation of the device. The storage module may be used to support the device in executing mutual program codes, etc.

[0100] The processing module may be a processor or controller that implements or executes the various exemplary logic blocks, modules, and circuits disclosed herein. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a digital signal processing (DSP) and a microprocessor, and the storage module may be a memory.

[0101] Example 3:

[0102] This embodiment also provides a computer-readable storage medium having computer program code stored therein. When the computer program code is executed on a computer, the computer executes the above-mentioned related method steps to implement a prediction support system for high-risk levels of sepsis patients provided in the above embodiment.

[0103] Among them, the device and computer-readable storage medium provided in this embodiment are used to execute the corresponding system provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding system provided above, and will not be repeated here.

[0104] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0105] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0106] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A prediction support system for high-risk level of sepsis patients, characterized by: The system includes: A data acquisition module is used to collect several categories of physiological data of the patient at each moment during the analysis period; The importance analysis module is used to obtain the abnormal time period of each category; obtain the importance of the physiological data of each category according to the abnormal degree of the physiological data in the abnormal time period of each category and the time interval between the abnormal time period and its adjacent abnormal time periods, including: for each abnormal time period of each category, taking the average of the abnormal degree of the physiological data of each category at all moments in the abnormal time period as the abnormal comprehensive value of the abnormal time period; taking the sum of the time intervals between the abnormal time period and all its first adjacent time periods as the first time interval value of the abnormal time period; calculating the cumulative sum of the ratios of the abnormal comprehensive value of all abnormal time periods of each category to the first time interval value as the local importance value of the physiological data of each category; taking the proportion of the abnormal moments of each category in the analysis time period as the abnormal occupancy value of each category; obtaining the importance of the physiological data of each category according to the local importance value and the abnormal occupancy value, and the local importance value and the abnormal occupancy value are both positively correlated with the importance; The severity analysis module is used to select feature categories based on importance, and record the abnormal period of the feature category as the target period; according to the difference in importance between all categories and the feature category in each target period, and the number of categories of abnormal physiological data in each target period, obtain the severity of each target period, including: obtaining the abnormal category in each target period, and the abnormal moment of the abnormal category in each target period; according to the difference in importance between each abnormal category and the feature category in each target period, and the number of abnormal moments of each abnormal category in each target period, obtain the correlation between the physiological data of each abnormal category and the feature category in each target period; based on the correlation, select a reference category from the abnormal category in each target period; according to the correlation between the reference category and the feature category in each target period, the importance of the reference category, and the number of reference categories in each target period, obtain the severity of each target period; The sepsis high-risk prediction module is used to obtain the high-risk value of each target period based on the severity change trend and time interval between each target period and its adjacent target periods; and predict the high-risk risk of sepsis for the patient to be tested based on all categories of physiological data in the analysis time period adjusted by the high-risk value.

2. A prediction support system for high-risk level of sepsis patients according to claim 1, characterized in that: Get the abnormal period for each category, including: Normalize the absolute value of the difference between the physiological data of each category at each moment and the normal physiological preset value to obtain the abnormality degree of the physiological data of each category at each moment; The abnormal moments of each category are selected based on the abnormality degree; the time period consisting of the continuous abnormal moments of each category is taken as the abnormal period of each category.

3. The prediction support system for high-risk level of sepsis patients according to claim 1, characterized in that: Obtain the correlation between physiological data of each abnormal category and feature category in each target period, including: The proportion of abnormal moments of each abnormal category in each target period is counted; the absolute value of the difference between the importance of each abnormal category and the feature category in each target period is negatively correlated and mapped, and the product of the mapping result and the proportion is used as the correlation between the physiological data of each abnormal category and the feature category in each target period.

4. A prediction support system for high-risk level of sepsis patients according to claim 1, characterized in that: According to the correlation between the reference category and the characteristic category in each target period, the importance of the reference category, and the number of reference categories in each target period, the severity of each target period is obtained, including: The product of the correlation between each reference category and the characteristic category and the importance of the reference category in each target period is used as the reference severity value of each reference category in each target period; the cumulative sum of the reference severity values ​​of all reference categories in each target period is used as the local severity value of each target period; The severity of each target period is obtained based on the total number of abnormal categories and local severity values ​​in each target period; the total number of abnormal categories and local severity values ​​in each target period are positively correlated with the severity.

5. The prediction support system for high-risk level of sepsis patients according to claim 1, characterized in that: Obtain high-risk risk values ​​for each target period, including: Perform a straight line fit on the severity of each target period and its second adjacent period, normalize the slope of the fitted line, and obtain the severity trend value of each target period; take the sum of the time intervals between each target period and all its second adjacent periods as the second period interval value of each target period; According to the severity, severe trend value and interval value of the second period of each target period, the high-risk value of each target period is obtained; the severity and severe trend value are positively correlated with the high-risk value, and the interval value of the second period is negatively correlated with the high-risk value.

6. The prediction support system for high-risk level of sepsis patients according to claim 1, characterized in that: The high-risk risk of sepsis in the patient to be tested is predicted based on all categories of physiological data during the analysis period adjusted by the high-risk risk value, including: The sum of the high-risk value and the constant 1 is used as the adjustment coefficient for each target period; based on the adjustment coefficient, all categories of physiological data within the target period are weighted to obtain the adjusted physiological data for each category at each moment within the target period; All categories of adjusted physiological data of the patients to be tested during the analysis period are input into the trained random forest model to predict the high-risk risk of sepsis in the patients to be tested.

7. The prediction support system for high-risk level of sepsis patients according to claim 1, characterized in that: The importance of a feature class is the maximum value among the importances of all classes.

8. The prediction support system for high-risk level of sepsis patients according to claim 5, characterized in that: The method of linear fitting is the least squares method.

Images