Patient sign deterioration evaluation processing method based on intensive care clinical data

Abstract

The invention discloses a patient sign deterioration assessment processing method based on intensive care clinical data, and the method can generate a more accurate grading result based on comprehensive data through the continuous monitoring of a plurality of pieces of physiological index data and the deep analysis of historical patient data. In an analysis process of a grading result, by introducing concepts of a dynamic normal value and a dynamic standard deviation and utilizing an acute change coefficient and a long-term adaptation coefficient, a sign deterioration score can be calculated more accurately, so that changes of physiological states of patients can be captured in time, and meanwhile, by combining individual difference factors such as ages and body mass indexes of the patients, the physical sign deterioration score can be calculated more accurately. The dynamic normal value of each physiological index is calculated, so that personalized evaluation and grading are realized; the structure causal model is utilized, and the intervention strategy suitable for the current patient can be intelligently output based on the grading result by training the multiple pieces of sample intervention strategy information and the corresponding grading result, so that the scientificity and effectiveness of medical decision making are improved.

Description

Technical Field

[0001] The present invention relates to the medical field, and in particular to a method for evaluating and processing patient vital sign deterioration of intensive care clinical data. Background Art

[0002] In modern medical monitoring, continuous monitoring of patients' physiological indicators is an important means to improve the quality of medical services. Existing technologies usually rely on single monitoring of basic physiological indicators, such as heart rate, blood oxygen saturation, body temperature, etc., and judge the patient's health status through the real-time data of these indicators. However, this traditional monitoring method often cannot effectively integrate multiple physiological indicator data, resulting in the inability to form a comprehensive health assessment; for example, the abnormality of a single physiological indicator does not always reflect the patient's overall health status, and the lack of multi-dimensional data analysis may lead to wrong judgments;

[0003] At the same time, current technologies often ignore the dynamic changes of physiological indicators and fail to consider the acute changes and long-term adaptation process of physiological indicators. This makes the model lack accuracy in assessing the health status of patients, which may lead to missed or misdiagnosis;

[0004] In addition, existing technologies usually use fixed thresholds for health assessment, without considering individual differences of patients (such as age, body mass index, etc.), making it difficult to achieve personalized medical intervention, resulting in the inapplicability of intervention strategies;

[0005] In addition, when generating intervention strategies for patients, traditional systems rely on simple rules or rules of thumb and lack intelligent decision support based on data analysis, which may not provide the best treatment options. Summary of the invention

[0006] The present invention provides a method for evaluating and processing patient vital sign deterioration of intensive care clinical data, comprising the following steps:

[0007] Continuously monitor and record multiple physiological index data of patients i ;

[0008] Physiological index data i These include body temperature, mean arterial pressure, heart rate, and blood oxygen saturation;

[0009] Based on physiological index data i Combine historical patient data and the spatial information of the patient's physiological index data to generate a grading result for the current patient, and output intervention strategy information based on the grading result combined with the structural causal model;

[0010] Preferably, based on physiological index data f i Generate a grading result for the current patient in combination with historical patient data, including the following steps:

[0011] Collect and obtain multiple historical patient data; initialize evaluation model parameters based on the historical patient data;

[0012] The evaluation model parameters and physiological index data are used, and the spatial information of the current patient's physiological index data is introduced to calculate the patient's physical sign deterioration score at the target time point;

[0013] The physical sign deterioration score is graded according to the preset threshold to obtain the grading result.

[0014] Preferably, the evaluation model parameters include the basic normal value μ of each physiological index data base,i and the basic standard deviation σ base,i , Acute change coefficient A of physiological index data i and the long-term adaptation coefficient B i , the periodic variation range of physiological index data C i and frequency ω i , Age influence coefficient D i , Standard deviation variation E i and BMI influence coefficient F i , the time constant τ of acute changes in physiological index data i , the time constant T of long-term adaptation of physiological index data i , the time constant τ of the periodic change of physiological index data δ,i And the standard deviation change time constant τ of physiological index data ε,i .

[0015] Preferably, the evaluation model parameters and physiological index data f i , and introduce the spatial information of the current patient's physiological index data to calculate the patient's physical sign deterioration score at the target time point, including the following steps:

[0016] Based on the evaluation model parameters combined with the current patient's age and the patient's body mass index (BMI), the dynamic normal value f of each physiological index data is calculated. i,norm(t) and the dynamic standard deviation σ of physiological index data i(t) ;

[0017] Based on physiological index data i And continuously monitored and recorded physiological index dataf i Calculate the second-order time derivative of each physiological index data at each time point t

[0018] Calculate the spatial gradient of the physiological index data based on the same physiological index data collected from different parts of the current patient's body; obtain the Laplacian operator of the physiological index data based on the spatial gradient calculation

[0019] Using the predefined scoring function, combined with the dynamic normal value f of the physiological index data i ,norm(t) and the dynamic standard deviation σ of physiological index data i (t), the second-order time derivative of physiological index data at each time point t Laplacian operator of physiological index data The sign deterioration score was calculated.

[0020] Preferably, the predefined scoring function is:

[0021]

[0022] Where t represents the time point; f i represents the measured value of the ith physiological index data; f i ,norm(t) represents the dynamic normal value of the ith physiological index data at time t; σ i (t) represents the dynamic standard deviation of the i-th physiological index data at time t; λ, α, β are weight coefficients; is the Laplace operator of the physiological index data; is the second-order time derivative of the ith physiological index data at each time point t.

[0023] Preferably, the dynamic normal value f of each physiological index data is calculated based on the evaluation model parameters combined with the current patient's age and the patient's body mass index BMI i,norm(t) and the dynamic standard deviation σ of physiological index data i(t) , including the following steps:

[0024] Get the current patient's age and body mass index (BMI);

[0025] Dynamic normal values ​​and standard deviations are calculated based on the estimated model parameters.

[0026] Preferably, the calculation of the dynamic normal value and the standard deviation based on the basic parameters of the physiological indicators includes the following steps:

[0027] Basic normal value μ based on physiological index data base,i Acute change coefficient A of physiological index data i and the long-term adaptation coefficient B i and the time constant τ i , T i Calculate the dynamic mean μ of the i-th physiological index data at time t i(t) ;

[0028] The time constant τ of periodic changes based on physiological index data δ,i, the periodic variation range of physiological index data C i , Age influence coefficient D i And age age is used to calculate the periodic variation amplitude δ of each physiological index data at time t i(t) ;

[0029] Standard deviation change time constant τ based on physiological index data ε,i , Standard deviation variation E i , BMI influence coefficient F i , body mass index BMI calculates the standard deviation change factor ε of each physiological index data at time t i(t) ;

[0030] Basic normal value μ based on physiological index data base,i and the basic standard deviation σ base,i , dynamic mean μ of physiological index data at time t i(t) , the periodic variation amplitude δ of physiological index data at time t i(t) And the standard deviation change factor ε of the physiological index data at time t i(t) Calculate the dynamic normal value f i,norm(t) and the dynamic standard deviation σ i(t) .

[0031] Better, dynamic mean μ i(t) The calculation method is:

[0032] μ i(t) =μ base,i +A i ×exp(-t / τ i )+B i ×log(1+t / T i );

[0033] In the formula, μ base,i represents the basic normal value of the i-th physiological index data;

[0034] A i is the acute change coefficient of physiological index data;

[0035] B i is the long-term adaptation coefficient of physiological index data;

[0036] τ i is the time constant of acute changes in physiological index data;

[0037] T i is the time constant for long-term adaptation of physiological indicator data;

[0038] Periodic variation amplitude δ i (t) is calculated as:

[0039]

[0040] In the formula, C i is the periodic variation amplitude of physiological index data, D i is the age effect coefficient;

[0041] τ δ,i Represents the time constant of periodic changes in physiological index data;

[0042] age indicates the patient's age; ref represents the reference age;

[0043] Standard deviation variation factor ε i(t) The calculation method is:

[0044]

[0045] In the formula, E i is the standard deviation variation; F i is the BMI influence coefficient; τ ε,i is the standard deviation change time constant of physiological index data; BMI is body mass index; BMI_ref represents the reference body mass index;

[0046] Dynamic normal value f i,norm(t) and the dynamic standard deviation σ i(t) The calculation formula is:

[0047] f i,norm(t) =μ i(t) +δ i(t) ×sin(ω i ×t+φ i );

[0048] In the formula, ω i is the frequency of periodic change; φ i is the phase of periodic changes;

[0049] σ i(t) =σ base,i ×(1+ε i(t) )×|cos(ω i ×t+φ i )|;

[0050] In the formula, σ base,i is the basic standard deviation of physiological index data; ω i is the frequency of periodic change; φ i is a periodically changing phase.

[0051] Compared with the prior art, the embodiments of the present invention have at least the following technical advantages:

[0052] From the analysis of the above-mentioned patient sign deterioration assessment and processing method for intensive care clinical data provided by the present invention, it can be known that in specific applications, by continuously monitoring multiple physiological indicator data (such as body temperature, mean arterial pressure, heart rate and blood oxygen saturation), and combining historical patient data for in-depth analysis, a more accurate grading result can be generated based on comprehensive data; and in the analysis process of the grading results, by introducing the concepts of dynamic normal value and dynamic standard deviation, and using acute change coefficient and long-term adaptation coefficient, the sign deterioration score can be calculated more accurately, so as to capture the changes in the patient's physiological state in time, and at the same time, combined with individual difference factors such as the patient's age and body mass index, the dynamic normal value of each physiological indicator is calculated, thereby realizing personalized evaluation and grading; using the structural causal model, through the training of multiple sample intervention strategy information and corresponding grading results, it is possible to intelligently output the intervention strategy suitable for the current patient based on the grading results, thereby improving the scientificity and effectiveness of medical decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A schematic diagram of the main process of a method for assessing and processing patient's physical signs deterioration in critical care clinical data;

[0054] Figure 2 A schematic diagram of the operation steps for generating a classification result in a method for evaluating and processing patient vital signs deterioration in intensive care clinical data;

[0055] Figure 3 A schematic diagram of the operation steps for calculating a physical sign deterioration score in a method for evaluating and processing a patient's physical sign deterioration in intensive care clinical data;

[0056] Figure 4 Calculate the dynamic normal value f of each physiological index data in a method for evaluating and processing the patient's physical sign deterioration in intensive care clinical data i,norm(t) and the dynamic standard deviation σ of physiological index data i(t) Schematic diagram of the operation steps;

[0057] Figure 5 A schematic diagram of the operational steps for calculating dynamic normal values ​​and standard deviations in a method for evaluating and processing patient vital sign deterioration in intensive care clinical data. DETAILED DESCRIPTION

[0058] The present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings.

[0059] like Figure 1 As shown, an embodiment of the present invention provides a method for evaluating and processing patient vital sign deterioration of intensive care clinical data, including the following steps:

[0060] Step S10: Continuously monitor and record multiple physiological index data of the patient i ;

[0061] The physiological index data f i These include body temperature, mean arterial pressure, heart rate, and blood oxygen saturation;

[0062] It should be noted that in the above-mentioned embodiment of the present application, the technicians set the data collection frequency for the pre-set equipment for collecting the patient's vital signs and physiological indicators data, calibrate the monitoring equipment, establish a data transmission channel for the data collected by the monitoring, and record and store the data in real time as the data basis for the subsequent assessment of the deterioration of the patient's vital signs.

[0063] Step S20: Based on the physiological index data f i Generate a grading result for the current patient in combination with historical patient data and spatial information of the patient's physiological indicator data, and output intervention strategy information based on the grading result combined with the structural causal model;

[0064] Specifically, the above classification results refer to the classification of the patient's physical condition, such as the patient's condition is stable, needs close attention, may need intervention, and emergency situation, requiring immediate intervention; the above embodiment of the present application is based on physiological indicator data f i Combining historical patient data with current patient care is first based on physiological indicator data f i Determine the clinical intervention strategy for the patient based on historical patient data analysis, and then provide care for the patient based on the clinical intervention strategy;

[0065] Specific clinical intervention strategy information, for example: when the patient's blood pressure is too high, such as when the mean arterial pressure is significantly higher than the normal level (for example, systolic blood pressure exceeds 140 mmHg), the clinical intervention strategy of increasing drug treatment (such as adjusting antihypertensive drug dosages), monitoring the patient's sodium intake, adjusting diet, and increasing exercise; for example, when the patient has difficulty breathing, such as when the oxygenation index and blood oxygen saturation are lower than the normal range, providing oxygen therapy, improving environmental ventilation, and evaluating whether bronchodilators or other respiratory treatments are needed; for example, when the patient's body temperature rises (over 38°C), antipyretic drugs (such as ibuprofen, acetaminophen) are used, body temperature changes are monitored, possible sources of infection are found, and clinical intervention strategies for adjusting antibiotic medications are adjusted;

[0066] It should be noted that the above-mentioned embodiments of the present application can first obtain the patient's health status in real time by continuously monitoring the patient's key physiological indicators (such as body temperature, mean arterial pressure, heart rate and blood oxygen saturation) to ensure rapid identification of abnormalities; the continuously recorded data provides an important basis for the assessment of the patient's physical signs deterioration, which helps to formulate appropriate nursing countermeasures; through real-time monitoring, changes in the patient's physiological indicators can be detected early, and measures can be taken in time to reduce the possibility of worsening of the disease, and the accurate data provided provides an important basis for clinicians to make decisions, which helps to improve the accuracy and effectiveness of patient care;

[0067] Furthermore, by combining the initial physiological indicators of the current patient with their historical data, the changes in the patient's health status are analyzed to provide a scientific basis for subsequent care; based on the above data analysis, a clinical intervention strategy suitable for the patient is determined to make the care more targeted and help improve the treatment effect; the above-mentioned embodiment of the present application can formulate a personalized care plan for each patient by combining historical patient data, thereby improving the effectiveness of treatment, and realizing dynamic tracking and adjustment of nursing strategies, which can respond to changes in the patient's condition in a timely manner and improve the nursing effect;

[0068] The above-mentioned embodiments of the present application can generate more accurate grading results based on comprehensive data by continuously monitoring multiple physiological indicator data (such as body temperature, mean arterial pressure, heart rate and blood oxygen saturation) and conducting in-depth analysis in combination with historical patient data; and in the analysis process of the grading results, by introducing the concepts of dynamic normal value and dynamic standard deviation, and using acute change coefficient and long-term adaptation coefficient, it is possible to more accurately calculate the physical sign deterioration score, so as to timely capture the changes in the patient's physiological state. At the same time, combined with individual difference factors such as the patient's age and body mass index, the dynamic normal value of each physiological indicator is calculated, thereby realizing personalized evaluation and grading; using the structural causal model, through the training of multiple sample intervention strategy information and corresponding grading results, it is possible to intelligently output the intervention strategy suitable for the current patient based on the grading results, thereby improving the scientificity and effectiveness of medical decision-making.

[0069] Specifically, Figure 2 As shown, in step S20, based on the physiological index data f i Combining historical patient data and spatial information of the patient's physiological index data to generate a grading result for the current patient includes the following steps:

[0070] Step S21: collecting and acquiring a plurality of historical patient data; initializing evaluation model parameters based on the historical patient data;

[0071] The evaluation model parameters include the basic normal value μ of each physiological index data base,i and the basic standard deviation σ base,i, Acute change coefficient A of physiological index data i and the long-term adaptation coefficient B i , the periodic variation range of physiological index data C i and frequency ω i , Age influence coefficient D i , Standard deviation variation E i and BMI influence coefficient F i , the time constant τ of acute changes in physiological index data i , the time constant T of long-term adaptation of physiological index data i , the time constant τ of the periodic change of physiological index data δ,i And the standard deviation change time constant τ of physiological index data ε,i ;

[0072] It should be noted that the basic normal value μ of each physiological indicator data in the above-mentioned embodiment of the present application is base,i and the basic standard deviation σ base,i It is the initial reference value of each physiological indicator data, usually determined based on large-scale population statistics or medical literature, and serves as the starting point for calculating dynamic normal values ​​and dynamic standard deviations;

[0073] The acute change coefficient A of the above physiological index data i is the short-term rapid change range of physiological index data; the long-term adaptation coefficient B i is the slow change amplitude of physiological index data over time; the acute change coefficient A of the above physiological index data i and the long-term adaptation coefficient B i Usually determined based on clinical research or expert experience, there may be different values ​​for each physiological indicator data, which will not be described in detail in this application;

[0074] The periodic variation range of the above physiological index data C i and frequency ω i It is determined based on known physiological cycles (such as circadian rhythms), and this application will not go into details;

[0075] The above age influence coefficient D i is the degree of influence of age on physiological index data; the above BMI influence coefficient F i is the influence of BMI on physiological index data, and the above standard deviation changes E i is the basic amplitude of the standard deviation change of the patient; the above age influence coefficient D i , Standard deviation variation E i and BMI influence coefficient F iThe acquisition is obtained by analyzing the age, BMI index and physiological index data of multiple patients, which is common knowledge of those skilled in the art and will not be described in detail in this application;

[0076] The various time constants τ in the above embodiments of the present application are i is the time constant of acute change, indicating the speed of short-term change; T i is the time constant of long-term adaptation, indicating the speed of long-term change; τ δ,i is the time constant of the periodic change of physiological index data, indicating the speed at which the periodic change reaches a stable state; τ ε,i It is the time constant of the standard deviation change of the physiological index data, which indicates the speed of the standard deviation change; the above-mentioned time constants are determined based on physiological research or clinical observation, and will not be repeated in this application.

[0077] It should be noted that the above-mentioned embodiments of the present application collect large-scale historical patient data through crawler technology or from a database recorded by monitoring equipment, and pre-process the physiological indicator data of patients in the historical patient data, including denoising and standardization; further use a machine learning algorithm to train the initial model; then optimize the model parameters through cross-validation; and finally set the optimized parameters as the initial parameters of the evaluation model.

[0078] Step S22: using the evaluation model parameters and the physiological index data, and introducing the spatial information of the current patient's physiological index data to calculate the patient's physical sign deterioration score at the target time point;

[0079] Step S23: grading the physical sign deterioration score according to a preset threshold to obtain a grading result.

[0080] It should be noted that the above-mentioned embodiment of the present application can provide sufficient information basis for the establishment and evaluation of subsequent models by collecting large-scale historical patient data; then convert the historical data into specific evaluation model parameters, including basic normal values, standard deviations, acute and long-term change coefficients, etc., to provide standard references for patient physiological indicators, helping to further understand and evaluate the health status of patients; use machine learning algorithms to train the initial model, and continuously optimize the model parameters through cross-validation to ensure the accuracy and reliability of the evaluation model;

[0081] Furthermore, by combining the evaluation model parameters with the patient's real-time physiological indicator data, a comprehensive physical sign deterioration score is calculated for the patient at a specific time point, which helps to quickly understand the dynamic changes in the patient's health status; the physical sign deterioration score is graded according to the set threshold. This grading process converts the score into an actionable care level to provide guidance for subsequent care measures; appropriate care plans are adopted according to the score grade to ensure that each patient receives appropriate medical care based on the severity of their physical sign deterioration.

[0082] Specifically, Figure 3 As shown, in step S22, the evaluation model parameters and the physiological index data f are used. i , and introduce the spatial information of the current patient's physiological index data to calculate the patient's physical sign deterioration score at the target time point, including the following steps:

[0083] Step S221: Calculate the dynamic normal value f of each physiological index data based on the evaluation model parameters combined with the current patient's age and the patient's body mass index BMI i,norm(t) and the dynamic standard deviation σ of physiological index data i(t) ;

[0084] For example: According to the technical solution adopted in the above embodiment of the present application, the technician monitors a 65-year-old critically ill patient with a BMI index of 28; first, we collect the patient's 7 key physiological index data at time t as follows:

[0085] f1 (oxygenation index PaO2 / FiO2) = 250 mmHg;

[0086] f2(body temperature)=38.5℃;

[0087] f3 (mean arterial pressure) = 70 mmHg;

[0088] f4 (heart rate) = 95 beats / min;

[0089] f5 (respiratory rate) = 22 times / min;

[0090] f6 (body temperature) = 38.5°C (same as f2);

[0091] f7 (blood oxygen saturation) = 94;

[0092] Furthermore, the f of each indicator is calculated by combining the current patient's age and the patient's body mass index (BMI). i,norm(t) and σ i(t) ;

[0093] Take the oxygenation index (f1) as an example:

[0094] Calculated: fi,norm(t) =380 mmHg, σ i(t) =50 mmHg;

[0095] Step S222: Based on the physiological index data f i And the physiological index data f that is continuously monitored and recorded i Calculate the second-order time derivative of each physiological index data at each time point t

[0096] It should be noted that the above embodiment of the present application is to calculate the second-order time derivative of each physiological indicator data at each time point by numerical differentiation method.

[0097] The technical solution adopted in the above-mentioned embodiment of the present application calculates the second-order derivative for each time point. In the scoring function, these second-order derivative values ​​will be used to evaluate the acceleration of the change of physiological indicator data, which may reveal potential acute changes or trend turning points;

[0098] It should be noted that the second-order derivative is very sensitive to noise, so in practical applications, additional measures may be needed to ensure the stability and reliability of the results, such as using more complex numerical methods or adding data smoothing steps;

[0099] The embodiment of the present application is in the medical field, so this calculation is usually performed in real time, and the problems of calculation efficiency and real-time performance need to be considered. In this case, the embodiment of the present application adopts sliding window technology to improve calculation efficiency;

[0100] The calculation method of the above second-order derivative is prior art and will not be described in detail in this application;

[0101] Continuing with the above example, we use numerical methods (such as the central difference method) to calculate the second-order time derivative of each indicator.

[0102] For f1, we get:

[0103] Step S223: Calculate the spatial gradient of the physiological index data according to the collected physiological index data of different body parts of the current patient; and obtain the Laplace operator of the physiological index data based on the spatial gradient calculation. Record the spatial gradient location attribute information of the current patient's collected physiological index data, including the head, chest, and abdomen;

[0104] It should be noted that the spatial gradient in the above-mentioned embodiment of the present application refers to the difference in the same physiological indicator data measured at different positions of the human body. The purpose of calculating the spatial gradient is to capture the changes in the physiological indicator data between different parts of the body, which can reveal important clinical information of the patient; for example, it may be the temperature difference between different parts of the body (such as the head, chest, abdomen). The purpose of calculating the spatial gradient of the same physiological indicator is to capture the changes in the physiological indicator data between different parts of the body (such as the temperature difference between different parts of the body), which can reveal important clinical information of the patient.

[0105] The Laplace operator of the above physiological index data Refers to the divergence of the spatial gradient of physiological indicator data, which can detect "abnormal points" or "singular points" in the distribution of physiological indicator data. The values ​​around these points are significantly different from the surrounding environment;

[0106] The spatial gradient of the above physiological index data and the calculation method of the Laplace operator of the above physiological index data are both common knowledge to those skilled in the art, and will not be described in detail in this application;

[0107] Continuing with the above example, calculate the Laplace operator of physiological index data Considering the spatial relationship between the indicators (ie, the spatial gradient of the above-mentioned physiological indicator data), the spatial gradient of the above-mentioned physiological indicator data is the so-called spatial information of the patient's physiological indicator data.

[0108] get

[0109] Step S224: Using a predefined scoring function, combined with the dynamic normal value f of the physiological index data i ,norm(t) and the dynamic standard deviation σ of physiological index data i (t), the second-order time derivative of physiological index data at each time point t Laplacian operator of physiological index data The sign deterioration score was calculated.

[0110] The predefined scoring function is:

[0111]

[0112] Where t represents the time point; f i represents the measured value of the ith physiological index data; f i ,norm(t) represents the dynamic normal value of the ith physiological index data at time t; σ i (t) represents the dynamic standard deviation of the i-th physiological index data at time t; λ, α, β are weight coefficients; is the Laplace operator of the physiological index data; is the second-order time derivative of the i-th physiological index data at each time point t;

[0113] Continuing with the above example, the dynamic normal value f of the physiological index data calculated in the above embodiment of the present application is i ,norm(t) and the dynamic standard deviation σ of physiological index data i (t), the second-order time derivative of physiological index data at each time point t Laplacian operator of physiological index data Substituting into the scoring function,

[0114] Set the weight coefficients λ = 0.1, α = 0.001, β = 0.1;

[0115] The following results are obtained by calculation:

[0116] Item 1:

[0117]

[0118] Item 2:

[0119] The third term: α∏[i=1to7](1+|f i -f i ,norm(t)| / σ i (t));

[0120] For f1: 1+|250-380| / 50=3.6;

[0121] The product of other indicators is calculated as follows: 3.6×2.1×1.8×2.5×1.9×2.1×1.7=144.66;

[0122] 0.001×144.66=0.14466;

[0123] The fourth term: β∑[i=1to7]exp(|f i -f i ,norm(t)| / σ i (t));

[0124] For f1: exp(|250-380| / 50)=13.46;

[0125] The sum of all 7 indicators is calculated as follows: 13.46+8.21+6.39+10.68+7.10+8.21+5.47=59.52;

[0126] 0.1×59.52=5.952;

[0127] Finally, S(t) = -2.3 + 0.02 + 0.14466 + 5.952 = 3.81666;

[0128] According to the technical solution adopted in the above embodiment of the present application, the following classification thresholds are set:

[0129] S(t)<2: patient's condition is stable;

[0130] 2≤S(t)<4: need to pay close attention;

[0131] 4≤S(t)<6: intervention may be required;

[0132] S(t)≥6: emergency situation, requiring immediate intervention;

[0133] According to the calculated S(t)=3.81666, the patient's status is "needs close attention", which is close to the threshold of "may require intervention", so the clinical intervention strategy in the subsequent operation is implemented;

[0134] It should be noted that, through the technical solution adopted in the above-mentioned embodiment of the present application, the physiological index data of the patient is subjected to a series of calculations to obtain a physical sign deterioration score, which provides a data basis for subsequent clinical intervention of the patient;

[0135] In the above-mentioned embodiments of the present application, the patient's age and BMI are important indicators for assessing health status. Patients of different ages and weights have significant differences in physiological indicators. By calculating dynamic normal values ​​and standard deviations, a personalized benchmark can be provided for subsequent monitoring to more accurately identify abnormal fluctuations; further, the second-order derivative reflects the trend and acceleration of changes in physiological indicators, and can reveal potential acute changes or turning points; by identifying these changes, clinicians can intervene in the patient's health status more promptly, and since the second-order derivative is sensitive to noise, measures such as data smoothing can enhance the reliability of the results, and the use of sliding window technology can improve computational efficiency and ensure real-time monitoring; in the above-mentioned implementation of the present application, the spatial gradient can capture the changes of the same physiological indicator in different parts of the body, which can provide important clinical information to help doctors understand the patient's physiological state. For example, the difference in heart rate in different parts may indicate certain physiological problems, and the Laplace operator helps identify abnormal points or singular points, thereby discovering potential pathological changes;

[0136] The above scoring function combines multiple indicators and calculation results. The scoring function can more comprehensively reflect the health status of patients and provide a quantitative evaluation standard to help clinicians quickly determine whether the patient's condition has deteriorated. It can enhance the monitoring ability of critically ill patients and timely detect changes in health status, so as to take more effective intervention measures.

[0137] In the above-mentioned embodiments of the present application, through dynamic evaluation and real-time calculation, doctors can quickly obtain feedback on the patient's health status, thereby improving the timeliness of intervention; personalized normal value calculations are performed according to the patient's specific circumstances (such as age and BMI) to ensure that each patient receives tailored monitoring standards; by using second-order time derivatives and Laplace operators, changes in physiological indicators can be more accurately identified, reducing the possibility of misjudgment; through comprehensive analysis of multiple indicators, potential health risks can be discovered early, providing doctors with effective decision-making support, and ultimately improving the quality of patient care.

[0138] Specifically, Figure 4 As shown, in step S221, the dynamic normal value f of each physiological index data is calculated based on the evaluation model parameters combined with the current patient's age and the patient's body mass index BMI. i,norm(t) and the dynamic standard deviation σ of physiological index data i(t) , including the following steps:

[0139] Step S2211: obtaining the age and body mass index (BMI) of the current patient;

[0140] It should be noted that the above-mentioned body mass index BMI is an indicator for evaluating whether the weight and height ratio of an adult is appropriate, which helps to more accurately calculate the dynamic standard deviation of each physiological indicator data, thereby improving the accuracy of the overall assessment. The calculation of the above-mentioned body mass index BMI is a prior art and will not be repeated in this application.

[0141] Step S2212: Calculating a dynamic normal value and a standard deviation based on the evaluation model parameters;

[0142] It should be noted that in the above-mentioned embodiment of the present application, BMI is used as an indicator to evaluate the weight-to-height ratio of adults, which is helpful to judge the nutritional status and health risks of patients. Through accurate BMI calculation, the patient's physiological condition can be better understood; the embodiment of the present application uses the patient's age and BMI as input factors, which can more accurately calculate the dynamic standard deviation of physiological indicators in the evaluation model, and can more accurately reflect individual characteristics, thereby improving the accuracy of overall health assessment; through the calculation of dynamic normal values ​​and standard deviations, the patient's health changes can be monitored in real time, so that medical staff can promptly understand the degree of change of the patient's physiological indicators.

[0143] Specifically, Figure 5 As shown, in step S2212, the dynamic normal value and the standard deviation are calculated based on the basic parameters of the physiological indicators, including the following steps:

[0144] Step S22121: Based on the basic normal value μ of the physiological index data base,i The acute change coefficient A of the physiological index data i and the long-term adaptation coefficient B i and the time constant τ i 、T i Calculate the dynamic mean μ of the i-th physiological index data at time t i(t) ;

[0145] The dynamic mean μ i(t) The calculation method is:

[0146] μ i(t) =μ base,i +A i ×exp(-t / τ i )+B i ×log(1+t / T i );

[0147] In the formula, μ base,i represents the basic normal value of the i-th physiological index data;

[0148] A i is the acute change coefficient of physiological index data;

[0149] B i is the long-term adaptation coefficient of physiological index data;

[0150] τ i is the time constant of acute changes in physiological index data;

[0151] T i is the time constant for long-term adaptation of physiological indicator data;

[0152] Step S22122: Based on the periodic change time constant τ of the physiological index data δ,i , the periodic variation range of physiological index data C i , Age influence coefficient D i And the age is calculated to obtain the periodic change amplitude δ of each physiological index data at time t i(t) ;

[0153] The periodic variation amplitude δ i (t) is calculated as:

[0154]

[0155] In the formula, C i is the periodic variation amplitude of physiological index data, D i is the age effect coefficient;

[0156] τ δ,i Represents the time constant of periodic changes in physiological index data;

[0157] age indicates the patient's age; ref represents the reference age;

[0158] Step S22123: Based on the standard deviation of the physiological index data, the time constant τ changes ε,i , Standard deviation variation E i , BMI influence coefficient F i , body mass index BMI calculates the standard deviation change factor ε of each physiological index data at time t i(t) ;

[0159] The standard deviation variation factor ε i(t) The calculation method is:

[0160]

[0161] In the formula, E i is the standard deviation variation; F i is the BMI influence coefficient; τ ε,i is the standard deviation change time constant of physiological index data; BMI is body mass index; BMI_ref represents the reference body mass index;

[0162] Step S22124: Based on the basic normal value μ of the physiological index data base,i and the basic standard deviation σ base,i , the dynamic mean μ of the physiological index data at time t i(t) , the periodic variation amplitude δ of physiological index data at time t i(t) And the standard deviation change factor ε of the physiological index data at time t i(t) Calculate the dynamic normal value f i,norm(t) and the dynamic standard deviation σ i(t) ;

[0163] The dynamic normal value f i,norm(t) and the dynamic standard deviation σ i(t) The calculation formula is:

[0164] f i,norm(t) =μ i(t) +δ i(t) ×sin(ω i ×t+φ i );

[0165] In the formula, ω i is the frequency of periodic change; φ i is the phase of periodic changes;

[0166] σ i(t) =σ base,i ×(1+ε i(t) )×|cos(ω i ×t+φ i )|;

[0167] In the formula, σ base,i is the basic standard deviation of physiological index data; ω i is the frequency of periodic change; φ i is the phase of periodic changes;

[0168] It should be noted that the above embodiments of the present application calculate parameters from various aspects, and finally use multi-parameter comprehensive calculation to obtain the patient's dynamic normal value and dynamic standard deviation, for example:

[0169] According to the technical solution adopted in the above embodiment of the present application, a technician monitors a 65-year-old critically ill patient with a BMI of 28, and calculates the dynamic normal value and dynamic standard deviation of the patient's heart rate (assuming it is the fourth physiological indicator, i.e., i=4) as an example:

[0170] First determine the basic parameters

[0171] Patient age: age=65;

[0172] Patient BMI: BMI = 28;

[0173] Reference age:age ref =50;

[0174] Reference BMI: BMI ref =25;

[0175] Current time: t=48;

[0176] Further, setting the model parameters related to heart rate;

[0177] μ base,4 =75 (basic normal heart rate);

[0178] σ base,4 =10(basal heart rate standard deviation);

[0179] A4=15 (acute change coefficient);

[0180] B4=5 (long-term adaptation coefficient);

[0181] C4=8 (periodic variation amplitude);

[0182] D4=0.2 (age influence coefficient);

[0183] E4 = 0.1 (standard deviation variation);

[0184] F4=0.3 (BMI influence coefficient);

[0185] ω4 = 0.26 (periodic change frequency, corresponding to a 24-hour period);

[0186] φ4=π / 6(periodic variation phase);

[0187] τ4 = 24 (acute change time constant);

[0188] T4 = 240 (long-term adaptation time constant);

[0189] τ δ,4 =72 (periodic change time constant);

[0190] τ ε,4 =120 (standard deviation change time constant);

[0191] Furthermore, we calculate μ 4(t) :

[0192] μ 4(t) =μ base,4 +A4×exp(-t / τ4)+B4×log(1+t / T4)

[0193] =75+15×exp(-48 / 24)+5×log(1+48 / 240)

[0194] ≈75+2.03+0.97

[0195] ≈78bpm;

[0196] Furthermore, we calculate δ 4(t) :

[0197]

[0198] Furthermore, we calculate f i,norm(t) :

[0199] f 4,norm(t) =μ 4(t) +δ 4(t) ×sin(ω4×t+φ4)

[0200] =78+4.12×sin(0.26×48+π / 6)

[0201] ≈78+4.12×0.5

[0202] ≈80.06bpm;

[0203] That is, the above 80.06 bpm is the current patient's dynamic normal heart rate, which means that at the current time point, considering the patient's age, BMI, hospitalization time and other factors, the expected normal heart rate is about 80 beats / minute;

[0204] Furthermore, we calculate ε 4(t) :

[0205]

[0206] Furthermore, we calculate σ i(t) :

[0207] σ 4(t) =σ base,4 ×(1+ε 4(t) )×|cos(ω4×t+φ4)|

[0208] =10×(1+0.034)×|cos(0.26×48+π / 6)|

[0209] ≈10×1.034×0.866

[0210] ≈8.95bpm;

[0211] That is, the above 8.95 bpm is the dynamic standard deviation of the patient's heart rate. At the current time point, the expected fluctuation range of the heart rate is about ±8.95 beats / minute;

[0212] In summary, it can be considered that the normal heart rate range of this patient at this time is about 80.06±8.95bpm, that is, 71.11 to 89.01bpm;

[0213] The above-mentioned embodiment of the present application calculates the dynamic normal value and the dynamic standard deviation according to the technical solution adopted in the present application, taking into account the individual characteristics of the patient (such as age and BMI) and time factors, so that the evaluation of the patient's physiological indicators is more personalized and dynamic, which helps to more accurately identify potential signs of deterioration.

[0214] Specifically, after step S23, the following operation steps are also included:

[0215] Step S231: Based on the grading result, calculating the change trend of the physical sign deterioration score by trend analysis;

[0216] It should be noted that the above trend analysis can be used to obtain the changing trend of the physical sign deterioration score by moving average, exponential smoothing and other methods; using moving average (Moving Average) or exponential smoothing (ExponentialSmoothing) and other methods to smooth the time series data of the physical sign deterioration score, so as to reveal its changing trend, eliminate short-term fluctuations, and emphasize long-term trends; specifically, by calculating the average score within a certain time window to smooth the data; further, by giving a higher weight to the most recent observations, the changing trend of the score is captured;

[0217] During the specific operation process, machine learning models (such as exponential smoothing with adaptive learning rate) can be combined to adjust weights in real time to adapt to individual differences among patients and improve the accuracy of trend prediction.

[0218] Step S232: Analyzing the contribution of each of the physiological indicator data to the physical sign deterioration score through principal component analysis;

[0219] It should be noted that the above embodiment of the present application uses PCA (i.e., the above principal component analysis) to perform dimensionality reduction processing on multidimensional physiological index data, integrating the influence of multiple indicators into a few principal components, thereby identifying the most important indicators; further, by calculating the contribution rate of each principal component in the physical sign deterioration score, it is determined which physiological indicators have the greatest impact on the score change;

[0220] During the specific implementation process, the PCA model can be updated according to real-time data so that it can quickly adapt to changes in the patient's physiological state and thus identify new key physiological indicators in real time.

[0221] Step S233: generating a visualization report based on the change trend and the contribution;

[0222] The visual report shows the changes in the physical sign deterioration score and the contribution of each indicator;

[0223] It should be noted that the above-mentioned embodiments of the present application use data visualization tools (such as Matplotlib, D3.js, etc.) to create charts and graphs to intuitively display the changing trend of the score and the contribution of each physiological indicator to the score; the report includes time series graphs, bar charts or pie charts to facilitate the medical team to quickly understand the data.

[0224] Step S234: Transmit the visualization report to the medical team's mobile device.

[0225] It should be noted that the above-mentioned embodiment of the present application can understand the changing pattern of the physical sign deterioration score by performing trend analysis on the physical sign deterioration score, which is crucial for timely detection of potential health risks; further, by analyzing the contribution, the medical team can give priority to those indicators that have a greater impact on the physical sign deterioration score, thereby improving the efficiency of intervention, and thus conducting targeted health management and intervention; through visual reports, medical staff can more intuitively monitor the patient's health status and respond in a timely manner; finally, the visual report is transmitted to the medical team's mobile devices to ensure that medical staff can access information anytime, anywhere, and enhance the initiative in patient health management.

[0226] Specifically, in step S20, outputting intervention strategy information according to the classification result combined with the structural causal model includes the following steps:

[0227] Using crawler technology to crawl multiple sample intervention strategy information and sample grading results corresponding to each intervention strategy information and sample physiological indicator data;

[0228] Based on the spatial gradient part attribute information of the physiological indicator data collected by the current patient, sample physiological indicator data with consistent spatial gradient part attribute information is selected;

[0229] A structural causal model is established and trained based on the sample intervention strategy information, the sample classification results, the categories of physiological indicator data (the categories of physiological indicator data include body temperature, mean arterial pressure, heart rate and blood oxygen saturation) and the spatial gradient location attribute information;

[0230] Based on the current classification result and the category of the current physiological indicator data, the intervention strategy information is obtained by analyzing and outputting the structural causal model.

[0231] Step S201: crawling multiple sample intervention strategy information and sample grading results and sample physiological index data corresponding to each intervention strategy information using crawler technology; then screening sample physiological index data with consistent spatial gradient site attribute information based on the spatial gradient site attribute information of the physiological index data collected by the current patient;

[0232] Step S202: establishing and training a structural causal model based on the sample intervention strategy information, the sample classification results, the categories of physiological indicator data (the categories of physiological indicator data include body temperature, mean arterial pressure, heart rate and blood oxygen saturation) and the spatial gradient site attribute information;

[0233] It should be noted that the above sample intervention strategy information records different intervention measures and their details; the sample grading results describe the patient's physical condition grading corresponding to the above sample physiological indicator data, such as whether the patient is stable, needs close attention, may need intervention, and is in an emergency, requiring immediate intervention; sample physiological indicator data: including sample patient's sample temperature, sample mean arterial pressure, sample heart rate and sample blood oxygen saturation, etc. At the same time, it should be noted that the "recording of spatial gradient site attribute information of the current patient's collected physiological indicator data, including the head, chest, and abdomen" performed in the aforementioned step S223 is still performed. This step can screen sample physiological indicator data that is more authentic and more comparative and referenceable;

[0234] For example, heart rate can be measured directly in the chest; heart rate can be measured in other parts of the head and abdomen, but usually not by directly measuring the heartbeat, but by indirectly obtaining heart rate information by measuring the arterial pulsation; in the head, the pulsation of the temporal artery can be felt in the temporal part (near the temple). The temporal artery pulsation point is located in front of the ear, and you can feel the pulse by touching it. In addition, the carotid artery in the neck is also a common place to measure the pulse. The carotid artery pulsation point is located in the depression between the trachea and the inner side of the sternocleidomastoid muscle. In the abdomen, the pulsation of the abdominal aorta can be measured and the heart rate can be calculated.

[0235] In addition, there are indeed differences in blood oxygen saturation at different locations of the human body, which is mainly caused by the different oxygen saturation of arterial blood and venous blood. Arterial blood is responsible for transporting oxygen from the lungs to various parts of the body, and its blood oxygen saturation is usually higher, so the blood oxygen saturation near the chest is higher.

[0236] Head oxygen saturation measurement method: Head oxygen saturation is usually measured using near-infrared spectroscopy (NIRS) technology, which can non-invasively monitor brain perfusion, hemodynamics and oxygenation status. Abdominal oxygen saturation measurement: Abdominal oxygen saturation can also be measured using NIRS technology, which can monitor the oxygenation of abdominal tissues, such as changes in intestinal blood flow.

[0237] Therefore, physiological indicator data include body temperature, mean arterial pressure, heart rate and blood oxygen saturation; the embodiment of the present application believes that it is not enough to simply determine the type of the same physiological indicator data, and it is also necessary to additionally verify the spatial gradient part attribute information; some physiological indicator data only measure the head and chest, and some physiological indicator data measure the four spatial gradient part attribute information of the head and chest, abdomen or even a certain part of the lower limbs. Therefore, when establishing a structural causal model, it is also necessary to consider the spatial gradient part attribute information, and introduce it as a parameter to participate in the establishment of the structural causal model.

[0238] Step S203: Based on the current classification result and the category of the current physiological indicator data, the structural causal model is used to analyze and output intervention strategy information.

[0239] It should be noted that the above-mentioned embodiment of the present application selects crawler technology to collect sample data, thereby providing rich and diverse sample data for the subsequent establishment and training of the structural causal model, including intervention strategies and corresponding grading results, thereby ensuring the diversity and authenticity of the data for structural causal model training; further, by converting sample data into an intelligent model, it is enabled to automatically generate intervention strategies, thereby providing intelligent support for practical applications; then, using the trained structural causal model, the corresponding intervention strategy is generated according to the real-time or specific grading results; finally, the medical team can formulate corresponding treatment plans based on the intervention strategy information generated by the model, thereby improving the treatment effect of patients.

[0240] Specifically, in step S202, a structural causal model is established and trained based on the sample intervention strategy information, the sample classification results, and the category and spatial gradient site attribute information of the physiological indicator data, including the following operation steps:

[0241] Step S2021: Encode the sample intervention strategy information to obtain an intervention strategy encoding vector I; encode the sample classification result to obtain a classification encoding vector y; establish a physiological indicator matrix X based on the category of the physiological indicator data; and encode the spatial gradient site attribute information to obtain a spatial information code S;

[0242] The intervention strategy encoding vector I∈Rm; wherein m is the number of sample intervention strategy information; R is a set of encoding vectors of m number of sample intervention strategy information;

[0243] The hierarchical coding vector y∈{1,2,…,K}, where K is the total number of levels;

[0244] The physiological index matrix Where n is the number of samples and p is the number of categories of physiological indicator data.

[0245]

[0246] Among them, x np Represents the category of the pth physiological indicator data of the nth sample;

[0247] The spatial information encoding in, Indicates the number of different spatial gradient sites;

[0248] Taking the intervention strategy encoding vector I as an example, when there are the first three intervention strategies: drug therapy, physical therapy and psychological therapy, the intensity of each strategy is divided into three levels: low, medium and high. According to the above embodiment of the present application, it can be encoded in the following way:

[0249] I = [I drug, I physical, I psychological];

[0250] Each of these I i ∈{0,1,2}, 0 means not using this strategy, 1 means low intensity, 2 means medium intensity, and 3 means high intensity, that is, when the intervention strategy is expressed as I1=[I drug 0, I physical 1, I psychological 2], the current patient is not treated with a drug intervention strategy, and at the same time, a low-intensity physical intervention strategy is used for the current patient, and at the same time, a medium-intensity psychological intervention strategy is used for the current patient;

[0251] For example, the classification coding vector y currently has four classification results: mild, moderate, severe, and critical. The sample classification results can be encoded as: y=1 (mild); y=2 (moderate); y=3 (severe); y=4 (critical);

[0252] Take the physiological index matrix X as an example, x 11 represents the temperature of the first patient, x 12 represents its mean arterial pressure, and so on;

[0253] For example, the spatial information encoding S includes three parts: head, chest, and abdomen. Then the spatial information encoding of the patient's head is encoded as follows: si = [1, 0, 0] (head);

[0254] Step S2022: constructing a graph structure based on the conditional independence of the intervention strategy coding vector I, the hierarchical coding vector y, the physiological indicator matrix X and the spatial information coding S;

[0255] It should be noted that the above-mentioned embodiment of the present application first establishes a completely undirected graph; further, by testing the conditional independence of every two parameters in the intervention strategy coding vector I, the hierarchical coding vector y, the physiological indicator matrix X and the spatial information coding S, when the two parameters are independent under the condition of any given subset, the edge between them is removed; finally, the direction of the edge is determined according to the influence relationship of its conditional independence to avoid introducing directed loops.

[0256] Step S2023: Calculate the conditional probability distribution for each node in the graph structure;

[0257] For example, take the category node of physiological indicator data as an example.

[0258]

[0259] in:

[0260] X i is the ith physiological index;

[0261] Pa(X i ) is X i The parent node set of

[0262] β ij Is the parent node pair X i The influence coefficient of

[0263] γ i Is the intervention strategy effective for X i The influence coefficient vector of

[0264] δ i is the spatial gradient position to X i The influence coefficient vector of

[0265] ∈ i is a random error that follows a normal distribution;

[0266] It should be noted that the above-mentioned embodiments of the present application are based on the structure of the Bayesian network and can clarify the causal relationship between variables. Through this parameterization, the model can capture the interaction between variables, thereby improving the prediction accuracy.

[0267] Step S2024: construct a likelihood function L(θ) based on the conditional probability distribution; and add a regularization term to the likelihood function to obtain a structural causal model L reg (θ);

[0268] The likelihood function is expressed as:

[0269]

[0270] Right now

[0271]

[0272] in, Contains all the parameters to be learned;

[0273] The structural causal model L reg (θ) is expressed as:

[0274] L reg (θ)=L(θ)-λΣ i,j |θ ij |; where λ is the regularization strength.

[0275] It should be noted that the above embodiments of the present application are based on maximum likelihood estimation (MLE), the purpose of which is to find the optimal parameters by maximizing the probability of observed data. (k) , X (k) |I (k) ,S (k) ; θ) is the conditional probability, which represents the probability of the classification result given the features, which enables the model to learn the prediction results based on the input data;

[0276] The above embodiment of the present application uses L1 regularization (i.e., the above structural causal model L reg By penalizing large parameter values ​​in the model, λ in (θ) can force some parameters to be zero, thereby achieving feature selection, helping to improve the generalization ability of the model and avoid overfitting on the training data. This completes the establishment and training of the structural causal model.

[0277] In the subsequent operation process, the structural causal model L can also be adjusted as needed. reg (θ) performs optimization operations, evaluation operations, causal effect estimation operations, and model update operations, which will not be described in detail.

[0278] In summary, a method for evaluating and processing the deterioration of patient vital signs in intensive care clinical data proposed in an example of the present invention can generate more accurate grading results based on comprehensive data by continuously monitoring multiple physiological indicator data (such as body temperature, mean arterial pressure, heart rate and blood oxygen saturation) and conducting in-depth analysis in combination with historical patient data; and in the analysis process of the grading results, by introducing the concepts of dynamic normal value and dynamic standard deviation, and using acute change coefficient and long-term adaptation coefficient, it is possible to more accurately calculate the vital sign deterioration score, thereby timely capturing changes in the patient's physiological state. At the same time, combined with individual difference factors such as the patient's age and body mass index, the dynamic normal value of each physiological indicator is calculated, thereby achieving personalized evaluation and grading; using a structural causal model, through training of multiple sample intervention strategy information and corresponding grading results, it is possible to intelligently output an intervention strategy suitable for the current patient based on the grading results, thereby improving the scientificity and effectiveness of medical decision-making.

[0279] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. A person skilled in the art may modify the technical solutions described in the above embodiments, or replace part or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating and processing patient vital signs deterioration in intensive care clinical data, characterized in that: The steps are as follows: Continuously monitor and record multiple physiological index data of patients i ; The physiological index data f i These include body temperature, mean arterial pressure, heart rate, and blood oxygen saturation; Based on the physiological index data f i A grading result is generated for the current patient in combination with historical patient data and spatial information of the patient's physiological indicator data, and intervention strategy information is output based on the grading result combined with the structural causal model.

2. A method for evaluating and processing patient vital signs deterioration of intensive care clinical data according to claim 1, characterized in that: The physiological index data based on i Generate a grading result for the current patient in combination with historical patient data, including the following steps: Collect and obtain multiple historical patient data; Initializing evaluation model parameters based on the historical patient data; Utilizing the evaluation model parameters and the physiological index data, and introducing the spatial information of the current patient's physiological index data, to calculate the patient's physical sign deterioration score at the target time point; The physical sign deterioration score is graded according to a preset threshold to obtain a grading result.

3. A method for evaluating and processing patient vital signs deterioration of intensive care clinical data according to claim 2, characterized in that: The evaluation model parameters include the basic normal value μ of each physiological index data base,i and the basic standard deviation σ base,i , Acute change coefficient A of physiological index data i and the long-term adaptation coefficient B i , the periodic variation range of physiological index data C i and frequency ω i , Age influence coefficient D i , Standard deviation variation E i and BMI influence coefficient F i , the time constant τ of acute changes in physiological index data i , the time constant T of long-term adaptation of physiological index data i , the time constant τ of the periodic change of physiological index data δ,i And the standard deviation change time constant τ of physiological index data ε,i .

4. A method for evaluating and processing patient vital sign deterioration of intensive care clinical data according to claim 3, characterized in that: The method uses the evaluation model parameters and the physiological index data f i , and introduce the spatial information of the current patient's physiological index data to calculate the patient's physical sign deterioration score at the target time point, including the following steps: The dynamic normal value f of each physiological index data is calculated based on the evaluation model parameters combined with the current patient's age and the patient's body mass index BMI. i,norm(t) and the dynamic standard deviation σ of physiological index data i(t) ; Based on the physiological index data f i And the physiological index data f that is continuously monitored and recorded i Calculate the second-order time derivative of each physiological index data at each time point t Calculate the spatial gradient of the physiological index data according to the same physiological index data collected from different parts of the current patient's body; and obtain the Laplace operator of the physiological index data based on the spatial gradient calculation. Record the spatial gradient location attribute information of the current patient's collected physiological index data, including the head, chest, and abdomen; Using a predefined scoring function, combined with the dynamic normal value f of the physiological index data i ,norm(t) and the dynamic standard deviation σ of physiological index data i (t), the second-order time derivative of physiological index data at each time point t Laplacian operator of physiological index data The sign deterioration score was calculated.

5. A method for evaluating and processing patient vital sign deterioration of intensive care clinical data according to claim 4, characterized in that: The predefined scoring function is: Where t represents the time point; f i represents the measured value of the ith physiological index data; f i ,norm(t) represents the dynamic normal value of the ith physiological index data at time t; σ i (t) represents the dynamic standard deviation of the ith physiological index data at time t; λ, α, β are weight coefficients; is the Laplace operator of the physiological index data; is the second-order time derivative of the ith physiological index data at each time point t.

6. A method for evaluating and processing patient vital sign deterioration of intensive care clinical data according to claim 5, characterized in that: The dynamic normal value f of each physiological index data is calculated based on the evaluation model parameters combined with the current patient's age and the patient's body mass index BMI i,norm(t) and the dynamic standard deviation σ of physiological index data i(t) , including the following steps: Get the current patient's age and body mass index (BMI); Dynamic normal values ​​and standard deviations are calculated based on the estimated model parameters.

7. A method for evaluating and processing patient vital sign deterioration of intensive care clinical data according to claim 6, characterized in that: The step of calculating the dynamic normal value and the standard deviation based on the basic parameters of the physiological indicators comprises the following steps: Based on the basic normal value μ of the physiological index data base,i The acute change coefficient A of the physiological index data i and the long-term adaptation coefficient B i and the time constant τ i , T i Calculate the dynamic mean μ of the i-th physiological index data at time t i(t) ; Based on the periodic change time constant τ of the physiological index data δ,i , the periodic variation range of physiological index data C i , Age influence coefficient D i And the age is calculated to obtain the periodic change amplitude δ of each physiological index data at time t i(t) ; Based on the standard deviation change time constant τ of the physiological index data ε,i , Standard deviation variation E i , BMI influence coefficient F i , body mass index BMI calculates the standard deviation change factor ε of each physiological index data at time t i(t) ; Based on the basic normal value μ of the physiological index data base,i and the basic standard deviation σ base,i , the dynamic mean μ of the physiological index data at time t i(t) , the periodic variation amplitude δ of physiological index data at time t i(t) And the standard deviation change factor ε of the physiological index data at time t i(t) Calculate the dynamic normal value f i,norm(t) and the dynamic standard deviation σ i(t) .

8. A method for evaluating and processing patient vital sign deterioration of intensive care clinical data according to claim 7, characterized in that: The dynamic mean μ i(t) The calculation method is: m i(t) =μ base,i +A i ×exp(-t / τ i )+B i ×log(1+t / T i ); In the formula, μ base,i represents the basic normal value of the i-th physiological index data; A i is the acute change coefficient of physiological index data; B i is the long-term adaptation coefficient of physiological index data; τ i is the time constant of acute changes in physiological index data; T i is the time constant for long-term adaptation of physiological indicator data; The periodic variation amplitude δ i (t) is calculated as: In the formula, C i is the periodic variation amplitude of physiological index data, D i is the age effect coefficient; τ δ,i Represents the time constant of periodic changes in physiological index data; age indicates the patient's age; ref represents the reference age; The standard deviation variation factor ε i(t) The calculation method is: In the formula, E i is the standard deviation variation; F i is the BMI influence coefficient; τ ε,i is the standard deviation change time constant of physiological index data; BMI is body mass index; BMI_ref represents the reference body mass index; The dynamic normal value f i,norm(t) and the dynamic standard deviation σ i(t) The calculation formula is: f i,norm(t) =μ i(t) +d i(t) ×sin(ω i ×t+φ i ); In the formula, ω i is the frequency of periodic change; φ i is the phase of periodic changes; s i(t) =s base,i ×(1+ε i(t) )×|cos(ω i ×t+φ i )|; In the formula, σ base,i is the basic standard deviation of physiological index data; ω i is the frequency of periodic change; φ i is a periodically changing phase.

9. A method for evaluating and processing patient vital sign deterioration of intensive care clinical data according to claim 8, characterized in that: After the physical sign deterioration score is graded according to the preset threshold and the grading result is obtained, the following operation steps are also included: Based on the grading results, calculating the changing trend of the physical sign deterioration score by trend analysis; Analyze the contribution of each of the physiological indicator data to the physical sign deterioration score through principal component analysis; Generate a visualization report based on the change trend and the contribution; The visual report shows the changes in the physical sign deterioration score and the contribution of each indicator; Stream visual reports to medical team’s mobile devices.

10. A method for evaluating and processing patient vital sign deterioration of intensive care clinical data according to claim 9, characterized in that: Outputting intervention strategy information according to the classification results combined with the structural causal model specifically includes the following steps: Using crawler technology to crawl multiple sample intervention strategy information and sample grading results corresponding to each intervention strategy information and sample physiological indicator data; Based on the spatial gradient part attribute information of the physiological indicator data collected by the current patient, sample physiological indicator data with consistent spatial gradient part attribute information is selected; A structural causal model is established and trained based on the sample intervention strategy information, the sample classification results, and the category and spatial gradient location attribute information of the physiological indicator data; Based on the current classification result and the category of the current physiological indicator data, the structural causal model is used to analyze and output the intervention strategy information.

Images