Enterostomy postoperative infection monitoring and early warning method and system

Abstract

The invention relates to the technical field of postoperative monitoring, in particular to an enterostomy postoperative infection monitoring and early warning method and system, and the method comprises the following steps: obtaining an enterostomy result based on the state of skin around an enterostomy and enterostomy excrement feature data collected after the enterostomy of a patient; performing time sequence decomposition on the red and swollen area, the wound humidity, the seepage amount and the pH value of excrement, extracting the trend of the red and swollen area and the fluctuation component of the humidity, and combining the pH value change of the seepage and the pH value change of the excrement. According to the method, through time sequence decomposition and sliding time window analysis, dynamic changes of red and swollen area, humidity, seepage amount and pH value are accurately captured, characteristic trend and offset difference are generated in real time, dynamic parameter adjustment and weight updating enable risk assessment to be more sensitive, infection risk changes are effectively predicted, early diagnosis delay of infection is avoided, and the risk assessment accuracy is improved. Through data grouping and distribution analysis, the monitoring precision and the risk grading accuracy are improved, accurate early warning and intervention are achieved, and all-around support is provided for postoperative management of patients.

Description

Technical Field

[0001] The present invention relates to the technical field of postoperative monitoring, and in particular to a method and system for monitoring and early warning of infection after enterostomy surgery. Background Art

[0002] The field of postoperative monitoring technology includes continuous tracking and diagnostic methods for the health status of patients after surgery, aiming to timely detect potential postoperative complications through real-time or regular monitoring of physiological indicators, wound status and related biological parameters. The core content of this technical field includes the collection, processing and evaluation of patients' postoperative signs, mainly covering the monitoring of basic vital signs such as body temperature, heart rate, blood oxygen, blood pressure, as well as observation of local regional infection, inflammatory response and tissue healing status after specific surgery. The systematic nature of the overall technical field includes hardware sensor design, data acquisition and transmission system and related software platform support, so as to achieve comprehensive monitoring and data analysis of the patient's postoperative recovery process.

[0003] Among them, the post-operative infection monitoring and early warning system for enterostomy refers to a technical system for real-time monitoring and early warning of possible infections in patients after enterostomy surgery. The subject of this patent covers wound status detection at the enterostomy site, patient vital sign data collection, infection risk assessment and early warning information generation. Specifically, data collection is achieved through micro-sensing devices installed at the patient's stoma site, such as humidity sensors, temperature sensors, etc., to detect changes in the wound environment; the collected data is analyzed and risk judged using a preset algorithm model; and the early warning module uses sound, light or other forms to remind nursing staff or patients to pay attention to potential infection risks. This method relies on the synergy of hardware equipment and analysis models to achieve comprehensive technical coverage for postoperative infection monitoring.

[0004] Existing technologies rely on simple physiological indicators and local wound observations, which are unable to capture dynamic changes and subtle trends, and the monitoring dimension is single. Infection risk assessment lacks comprehensive analysis of humidity, exudate, and acid-base changes, and early warning is delayed, making early intervention difficult. The lack of trend prediction and real-time model adjustment makes it difficult to identify potential infection risks in a timely manner, delaying nursing intervention, increasing the uncertainty of postoperative recovery of patients and the consumption of medical resources. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a method and system for monitoring and early warning of post-intestinal ostomy infection.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a method for monitoring and early warning of infection after enterostomy surgery, comprising the following steps: S1: Based on the skin condition around the stoma and the characteristic data of stoma excretion collected after the patient's stoma surgery, the time series of redness and swelling area, wound humidity, exudate volume and excreta pH were decomposed, the trend of redness and swelling area and the fluctuation component of humidity were extracted, and the changes in pH of exudate and excreta were combined to generate the decomposition characteristic trend value; S2: Based on the decomposed characteristic trend value, the sliding time window technology is used to process the swelling diffusion rate, the humidity change rate and the exudate acid-base gradient, the divergence difference and the moving distance of each window are calculated, and it is determined whether the distribution difference between the windows reaches the offset threshold, the time point of the offset is determined, and the characteristic drift difference value is generated; S3: Based on the feature drift difference value, weight calculation is performed on the colostomy healing process data within the feature offset window, the changes in healing time, regional temperature difference and tissue pH are evaluated, the weight components are adjusted to match the distribution changes, the parameter matrix of the prediction model is updated, and a drift update weight matrix is ​​generated; S4: Based on the drift update weight matrix, adjust the parameters of the infection prediction model, integrate the updated healing time, regional temperature difference and body temperature dynamic change data, calculate the dynamic prediction value of infection risk, analyze the distribution of risk data, and generate a dynamic infection risk value; S5: Based on the dynamic infection risk value, combined with the skin condition around the stoma, excrement characteristics and healing process data, the infection risk level is evaluated by grouping, and the risk group data exceeding the warning range is screened and compared with the set warning threshold to generate an infection risk warning indicator.

[0007] The decomposed characteristic trend value specifically includes the trend of redness and swelling area, humidity fluctuation component, and change in pH of exudate. The characteristic drift difference value includes divergence difference, moving distance, and time point determination. The drift update weight matrix specifically refers to the healing time weight, regional temperature difference weight, and tissue pH weight. The dynamic infection risk value includes dynamic data of healing time, dynamic data of regional temperature difference, and dynamic change of body temperature. The infection risk warning indicator specifically includes infection risk level, risk grouping data, and warning threshold comparison.

[0008] As a further solution of the present invention, the step of obtaining the decomposition characteristic trend value is specifically as follows: S111: according to the redness and swelling area time series data of the skin condition around the enterostomy, the difference between the local weighted average value and the global time series is calculated in segments, the time trend feature is judged by using the relationship between the difference and the dynamic threshold, the significant change part of the redness and swelling area in the time series is extracted, and the normalization operation is performed based on the characteristics of the significant change part to generate the redness and swelling area trend feature value; S112: Based on the wound humidity time series data, the humidity fluctuation amplitude is processed by dynamic window smoothing to construct a product sequence of humidity change frequency and amplitude. At the same time, the overall fluctuation intensity characteristic value is calculated by combining the fluctuation amplitude and frequency. The fluctuation characteristic value is used to filter the part exceeding the set threshold and adjust the result to generate the humidity fluctuation characteristic value; S113: Combining the redness and swelling area trend characteristic value and the humidity fluctuation characteristic value, based on the two characteristic values ​​and the time series parameter of the pH change of the excrement, using the formula: ; Generate decomposition feature trend values; in, represents the decomposition characteristic trend value, Represents the characteristic value of the redness and swelling area trend, represents the humidity fluctuation frequency, Represents the humidity fluctuation amplitude, represents the eigenvalue weight, represents the pH change adjustment factor, Represents the range of pH value of excrement. Represents the characteristic component number in the time series, Represents the total number of feature components.

[0009] As a further solution of the present invention, the step of obtaining the characteristic drift difference value is specifically as follows: S211: Based on the redness and swelling diffusion rate, humidity change rate and exudate acid-base gradient in the decomposed characteristic trend value, a sliding time window is set, a divergence difference is calculated according to the characteristic parameters of each window, and a relationship between the divergence difference and the distribution threshold is used to filter the part that meets the conditions in the window, and a divergence difference sequence is extracted to generate a divergence difference sequence of the redness and swelling diffusion rate and the humidity change rate; S212: Based on the divergence difference sequence of the redness and swelling diffusion rate and the humidity change rate, the moving distance of adjacent parameters between windows is calculated, the difference of the moving distance between windows is used to construct a normalized index, and the windows exceeding the offset threshold are screened in combination with the distribution characteristics of the moving distance, and the parameters in the offset window are used to readjust the feature range to generate an offset window sequence; S213: Combining the divergence difference sequence and the offset window sequence, calculating the time point and amplitude of the feature drift in each window, using the formula: ; Generate feature drift difference value; in, represents the feature drift difference value, Represents the divergence difference of the current window, represents the divergence difference of the previous window, represents the weighted weight, Represents the moving distance between windows, represents the drift time point, represents the time adjustment factor, Represents the total number of time windows.

[0010] As a further solution of the present invention, the step of acquiring the drift update weight matrix is ​​specifically as follows: S311: based on the characteristic drift difference value, extract the intestinal stoma healing process data in the characteristic offset window, call the healing time series, regional temperature difference series and tissue pH series, calculate the initial weight value in the window for each characteristic data according to the change trend of time and space distribution, and generate the initial weight value matrix of the characteristic offset window; S312: using the initial weight value matrix of the feature offset window, calculating the local time trend and spatial distribution difference in the feature sequence, adjusting the multi-feature weight values ​​in combination with the dynamic trend of the weight component change, and using the multi-feature component weight change ratio to update the initial value matrix of the global feature distribution to generate a weight update matrix that matches the distribution change; S313: Combining the weight update matrix of the distribution change matching with the parameter matrix of the prediction model, the drift matching amount of the feature distribution to the prediction weight is calculated using the formula: ; Generate drift update weight matrix; in, represents the drift update weight matrix, Represents the initial weight value of the feature offset window, characterizing the initial weight distribution in multiple windows. Represents the prediction model parameter matrix component, which represents the parameter value of the corresponding feature in the model. The adjustment coefficient of the local trend feature is used to dynamically adjust the time trend of the feature. The scaling factor that represents the distribution change is used to adjust the gain of the drift range to the weight. Indicates the offset difference of the feature sequence, characterizing the distribution offset change of the feature in the sequence. Represents the weight matrix smoothing adjustment item, which is used to avoid the calculation instability caused by extremely small or zero values. Represents the normalized offset adjustment coefficient, which is used to control the weight normalization of the overall feature distribution. Represents the total number of features.

[0011] As a further solution of the present invention, the step of obtaining the dynamic infection risk value is specifically as follows: S411: Based on the drift update weight matrix, adjust the parameters of the infection prediction model, call the healing time series, regional temperature difference series and body temperature dynamic change data in the model, calculate the risk impact factor of each feature, and generate a preliminary risk impact matrix; S412: using the preliminary risk impact matrix, analyzing the impact of the dynamic changes of the time characteristic distribution and the regional temperature difference in the healing time series on the body temperature data, calculating the risk correlation between the distributions, dynamically adjusting the risk component weights according to the correlation, and generating a dynamically adjusted risk correlation matrix; S413: Based on the dynamically adjusted risk association matrix, the dynamic prediction value of the infection risk is calculated, and the association calculation is performed in combination with the dynamic weight of the risk component, the time distribution characteristics and the regional temperature difference distribution, using the formula: ; Generate dynamic infection risk value; in, Represents the dynamic infection risk value, characterizing the risk assessment of the characteristic component in dynamic infection prediction, Represents the dynamic weight of the risk component, characterizing the weight ratio of each component in the infection prediction model. The influencing factors representing the time distribution characteristics characterize the dynamic characteristic changes in the healing time series. The risk influencing factor representing the regional temperature difference distribution characterizes the contribution of regional temperature difference in infection prediction. The risk fluctuation amount representing the dynamic change of body temperature, characterizing the impact of body temperature characteristics on the overall infection risk, Represents the smoothing adjustment parameter, which is used to correct the weight effect of feature fluctuations. Represents the total number of components, Represents the number of the current feature component.

[0012] As a further solution of the present invention, the steps for obtaining the infection risk warning indicator are specifically as follows: S511: Based on the dynamic infection risk value, combined with the skin status data around the stoma, excrement characteristics and healing process data, the impact of the differentiated component data on the infection risk level is evaluated in groups, and the change characteristics of the skin status and excrement in each group of data are called to generate a group risk assessment matrix; S512: Using the group risk assessment matrix, filter the group data that exceeds the warning range, calculate the distribution offset amplitude through the dynamic correlation characteristics between risk components, compare the change trends of multi-component dynamic characteristics, and call the adjustment threshold within the warning range to generate a risk warning screening matrix; S513: Based on the risk alert screening matrix, the risk data is compared with the set alert threshold, the distribution of risk data is analyzed, and quantitative analysis is performed in combination with the comparison between the risk level data and the threshold to obtain the risk grouping data that exceeds the alert range, using the formula: ; Calculate and generate infection risk warning indicators; in, It represents the infection risk warning indicator and is a quantitative representation of the risk level. Represents the risk value of each group of data, which is used to determine the absolute amount of risk for each group. Represents the set warning threshold, which is the predetermined upper limit of risk tolerance. Represents the standard deviation of all risk data and is used to measure the volatility of risk. Represents the correction factor for each set of data, which is used to adjust the sensitivity of each set of risk values. Represents the number of data groups used to calculate the average warning indicator.

[0013] A system for monitoring and early warning of infection after colostomy, which is used to implement the above-mentioned method for monitoring and early warning of infection after colostomy, and comprises: The feature extraction module extracts the redness and swelling area trend component and humidity fluctuation component based on the redness and swelling area data, wound humidity data, exudate volume data and excreta pH data collected after enterostomy surgery, combines the trends of exudate and excreta pH changes, and generates decomposition feature trend values; The feature drift module groups the swelling diffusion rate data, the humidity change rate data and the exudate acid-base gradient data based on the decomposed feature trend value, calculates the divergence difference and the moving distance according to the set sliding time window, compares the difference between adjacent windows and determines the offset time point and the drift degree, and generates a feature drift difference value; The infection assessment module calls the enterostomy healing time data, regional temperature difference data and tissue pH change data within the offset window based on the characteristic drift difference value, calculates weight values ​​for multiple data, adjusts the weight distribution to match the drift trend distribution, and generates a drift update weight matrix; The infection prediction module calculates and aggregates the dynamic distribution of infection risk based on the drift update weight matrix, combines the healing time data, the regional temperature difference data and the dynamic change data of body temperature, obtains the dynamic prediction value of infection risk, and generates the dynamic prediction value of infection risk; The early warning analysis module evaluates the infection risk level in groups based on the dynamic prediction value of infection risk, combined with the skin status data around the stoma, the fecal characteristic data and the healing process data, screens the risk data and compares it with the warning threshold to generate an infection risk early warning indicator.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, through time series decomposition and sliding time window analysis, the dynamic changes of redness and swelling area, humidity, exudate volume and pH are accurately captured, and characteristic trends and offset differences are generated in real time. Dynamic parameter adjustment and weight update make risk assessment more sensitive, effectively predict changes in infection risk, and avoid delays in early diagnosis of infection. Through data grouping and distribution analysis, the accuracy of monitoring and risk grading is improved, accurate early warning and intervention are achieved, and all-round support is provided for postoperative management of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 A flowchart of the steps for obtaining the decomposition characteristic trend value of the present invention; Figure 3 This is a flow chart of the steps for obtaining the characteristic drift difference value of the present invention; Figure 4 A flow chart of the steps for obtaining the drift update weight matrix of the present invention; Figure 5 This is a flow chart of the steps for obtaining the dynamic infection risk value of the present invention; Figure 6 This is a flow chart of the steps for obtaining infection risk warning indicators of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0018] Example 1: Please refer to Figure 1 The present invention provides a technical solution: a method for monitoring and early warning of post-intestinal ostomy infection, comprising the following steps: S1: Based on the skin condition around the stoma and the characteristic data of stoma excretion collected after the patient's stoma surgery, the time series of redness and swelling area, wound humidity, exudate volume and excreta pH were decomposed, the trend of redness and swelling area and the fluctuation component of humidity were extracted, and the changes in pH of exudate and excreta were combined to generate the decomposition characteristic trend value; S2: Based on the decomposed characteristic trend value, the sliding time window technology is used to process the swelling diffusion rate, humidity change rate and exudate acid-base gradient, calculate the divergence difference and moving distance of each window, judge whether the distribution difference between windows reaches the offset threshold, determine the offset time point, and generate the characteristic drift difference value; S3: Based on the feature drift difference value, the colostomy healing process data within the feature offset window is weighted, the changes in healing time, regional temperature difference and tissue pH are evaluated, the weight components are adjusted to match the distribution changes, the parameter matrix of the prediction model is updated, and the drift update weight matrix is ​​generated; S4: Based on the drift update weight matrix, adjust the parameters of the infection prediction model, integrate the updated healing time, regional temperature difference and body temperature dynamic change data, calculate the dynamic prediction value of infection risk, analyze the distribution of risk data, and generate a dynamic infection risk value; S5: Based on the dynamic infection risk value, combined with the skin status around the stoma, excrement characteristics and healing process data, the infection risk level is evaluated by group, and the risk group data that exceeds the warning range is screened and compared with the set warning threshold to generate infection risk warning indicators.

[0019] The decomposed characteristic trend values ​​specifically include the trend of redness and swelling area, humidity fluctuation component, and change in exudate pH. The characteristic drift difference values ​​include divergence difference, moving distance, and time point determination. The drift update weight matrix specifically refers to the healing time weight, regional temperature difference weight, and tissue pH weight. The dynamic infection risk value includes the dynamic data of healing time, the dynamic data of regional temperature difference, and the dynamic change of body temperature. The infection risk warning indicators specifically include infection risk level, risk grouping data, and warning threshold comparison.

[0020] See also Figure 2 , the specific steps for obtaining the decomposition feature trend value are: S111: according to the redness and swelling area time series data of the skin condition around the intestinal stoma, the difference between the local weighted average value and the global time series is calculated in segments, and the time trend characteristics are judged by using the relationship between the difference and the dynamic threshold, and the significant change part of the redness and swelling area in the time series is extracted, and the normalization operation is performed based on the characteristics of the significant change part to generate the redness and swelling area trend characteristic value; First, the collected time series data is divided into several small segments per hour, and the local weighted average value is calculated for each segment. Each segment of data is smoothed by using the weighted moving average formula to reduce the interference of local fluctuations on trend characteristics. Then, the weighted average value of each segment is compared with the global time series mean, and the deviation value of each segment from the global mean is calculated. Subsequently, a dynamic threshold is established according to the amplitude of the deviation value. The dynamic threshold is calculated by the proportional factor according to the distribution range of the deviation value. For each segment of data, it is judged whether its deviation value exceeds the dynamic threshold range, and the redness and swelling area data segments that exceed the range are screened out. Then, these data segments that exceed the range are normalized using the normalization formula, and their values ​​are standardized to the same scale. Finally, the significant change part after normalization is feature extracted to generate the redness and swelling area trend characteristic value.

[0021] S112: Based on the wound humidity time series data, the humidity fluctuation amplitude is processed by dynamic window smoothing to construct a product sequence of humidity change frequency and amplitude. At the same time, the overall fluctuation intensity characteristic value is calculated by combining the fluctuation amplitude and frequency. The fluctuation characteristic value is used to filter the part exceeding the set threshold and adjust the result to generate the humidity fluctuation characteristic value; Firstly, the humidity data is preprocessed to remove the obvious outliers that appear in the measurement process, and the missing data is filled by the cubic spline interpolation method. Then, the product sequence of humidity change frequency and amplitude is constructed. The humidity change frequency is obtained by calculating the number of humidity changes between adjacent time points, and the humidity amplitude is obtained by calculating the absolute value of the humidity change between adjacent time points. Then, the fluctuation amplitude is adjusted by combining the window smoothing processing of the time series. The smoothing processing uses the sliding window average formula to locally smooth the humidity fluctuation amplitude, thereby eliminating the influence of random fluctuations in the short term. Then, the fluctuation characteristic value is combined to screen out the part that exceeds the set threshold. The threshold is calculated according to the statistical distribution of humidity fluctuation frequency and amplitude. Finally, the humidity fluctuation characteristic value is optimized by adjusting the dynamic weighting coefficient of frequency and amplitude to generate the humidity fluctuation characteristic value.

[0022] S113: Combining the redness and swelling area trend characteristic value and the humidity fluctuation characteristic value, based on the two characteristic values ​​and the time series parameter of the pH change of the excrement, using the formula: ; Generate decomposition feature trend values; in, represents the decomposition characteristic trend value, Represents the characteristic value of the redness and swelling area trend, represents the humidity fluctuation frequency, Represents the humidity fluctuation amplitude, represents the eigenvalue weight, represents the pH change adjustment factor, Represents the range of pH value of excrement. Represents the characteristic component number in the time series, Represents the total number of feature components.

[0023] formula: ; The benefit of the formula is that it comprehensively considers the multidimensional characteristics in the time series through the nonlinear combination of the trend characteristic values ​​of the redness and swelling area, the frequency and amplitude of humidity fluctuations, and the introduction of constrained adjustment of the pH changes of feces, thereby improving the accuracy and detail of the decomposed characteristic trend values.

[0024] Detailed explanation of the formula and the process of formula calculation and derivation: Assume that the trend characteristic value of the redness and swelling area is , collect and extract the significant change values ​​of the redness and swelling area trend in the time series, humidity fluctuation frequency The humidity fluctuation amplitude is obtained by counting the frequency of humidity changes in unit time. The characteristic value weight is obtained by counting the amplitude of humidity change in unit time. Dynamic allocation based on the importance assessment of the above characteristics, pH change adjustment coefficient According to the influence of pH range on the characteristics, the pH range of excrement is set. It is calculated by the absolute change in the pH value of the excreta in each time period; Assume that the collected data is , the calculation formula is: 1. Calculate the molecular part: ; 2. Calculate the denominator: ; 3. Comprehensive calculation: ; The results show that the decomposition characteristic trend value is 0.4338, which comprehensively reflects the overall characteristic trend of the redness and swelling area trend, humidity fluctuations and changes in the pH value of feces, and can be used for subsequent dynamic monitoring and optimization analysis.

[0025] See also Figure 3 , the specific steps for obtaining the feature drift difference value are: S211: based on the redness and swelling diffusion rate, humidity change rate and exudate acid-base gradient in the decomposed characteristic trend value, a sliding time window is set, a divergence difference is calculated according to the characteristic parameters of each window, and the relationship between the divergence difference and the distribution threshold is used to filter the part that meets the conditions in the window, extract the divergence difference sequence, and generate the divergence difference sequence of the redness and swelling diffusion rate and the humidity change rate; The decomposed characteristic trend value is obtained by decomposing the time series of the skin status and feces characteristic data after colostomy surgery. The trend component of the swelling and redness diffusion rate, the fluctuation component of the humidity change rate, and the characteristic value of the exudate acid-base gradient are extracted respectively. The decomposed characteristic trend value is formed by these characteristic values. Then, a sliding time window is set to divide the characteristic data into several segments according to the time window. Each segment contains data at multiple time points. For each segment of the data, the divergence difference of each window is calculated. The divergence difference is calculated by the formula Completed, including Is the first The characteristic values ​​of each time point are obtained by taking the average of the divergence differences of all time points in the window to obtain the divergence difference result of each window. Then the divergence difference of each window is compared with the set distribution threshold. The windows with significantly exceeded divergence differences are screened out through threshold judgment. These windows are marked as abnormal windows, and the divergence difference sequence of the abnormal windows is extracted as the input of subsequent calculations to generate the divergence difference sequence of redness and swelling diffusion rate and humidity change rate.

[0026] S212: Based on the divergence difference sequence of the redness and swelling diffusion rate and the humidity change rate, the moving distance of the adjacent parameters between the windows is calculated, and the difference of the moving distance between the windows is used to construct a normalized index. The windows exceeding the offset threshold are screened out in combination with the distribution characteristics of the moving distance, and the parameters in the offset window are used to readjust the feature range to generate an offset window sequence; The divergence difference sequence is calculated by extracting the window divergence difference outside the distribution threshold, and taking it as the input of the analysis window moving distance. By calculating the moving distance between adjacent windows, the specific formula is Completed, including It is a window The divergence difference of , and then construct the normalized index of the moving distance, and set the normalized formula as , obtain the moving normalized distance of each window, then determine whether the offset exceeds the offset threshold through the normalized distance, extract all windows whose offsets exceed the threshold as offset windows, further filter out the feature data in these offset windows, adjust the range of the feature values ​​in the filtered offset windows, map the offset window feature data to a unified interval through linear normalization, and obtain the adjusted offset window feature data again, and combine these adjusted feature data to generate an offset window sequence.

[0027] S213: Combine the divergence difference sequence and the offset window sequence to calculate the time point and amplitude of the feature drift in each window, using the formula: ; Generate feature drift difference value; in, represents the feature drift difference value, Represents the divergence difference of the current window, represents the divergence difference of the previous window, represents the weighted weight, Represents the moving distance between windows, represents the drift time point, represents the time adjustment factor, Represents the total number of time windows.

[0028] formula: ; The benefit of the formula is that by introducing the difference in divergence between windows and moving distance The weighted operation, combined with the adjustment coefficient and time point The dynamic constraints can comprehensively evaluate the temporal and spatial differences of feature drift, improving the accuracy and granularity of the results.

[0029] Detailed explanation of the formula and the process of formula calculation and derivation: It is The difference in the divergence of the windows, It is The divergence difference of the windows is calculated by the above formula and input into the calculation. The specific value can be calculated by decomposing the characteristic trend value. For example, assuming and ,but , weight parameter is the weight value of each window, indicating the importance of the window. Its value is calculated based on the significant change of the characteristic parameters in the window. For example, if the redness and swelling diffusion rate is large in a certain window, the weight Increase, assuming , window moving distance The formula Calculate, assuming the moving distance , time point is the time stamp of the window, and the adjustment coefficient Used to adjust the influence of time points on the overall drift difference, assuming , substitute into the formula: ; The results show that the calculation of the feature drift difference value reflects the degree of difference in features between sliding windows. When the result is small, it means that the feature change amplitude is small or the continuity is strong, while a larger value indicates that the feature drift is significant. This value can be used to further screen windows with significant drift features.

[0030] See also Figure 4 , the steps to obtain the drift update weight matrix are as follows: S311: based on the feature drift difference value, extract the colostomy healing process data in the feature offset window, call the healing time series, regional temperature difference series and tissue pH series, calculate the initial weight value in the window for each feature data according to the change trend of time and space distribution, and generate the initial weight value matrix of the feature offset window; Firstly, the healing time series is split into daily observations, the daily change rate is calculated, and the average change rate within 7 days is calculated through a sliding window. The time points with a change rate higher than a certain dynamic threshold are extracted, and the time offset range is marked in the form of intervals. The regional temperature difference series takes the temperature difference data at different time points to establish a time-space distribution matrix. The gradient values ​​of adjacent time points in each region are calculated according to the temperature difference change trend, and regularization is performed with the gradient difference. The high gradient area is marked as the key distribution area. The tissue pH series adopts the time series decomposition method to decompose the time trend component and the fluctuation component. The trend component is used to judge the direction and amplitude of long-term changes, and the fluctuation component is used to screen the time nodes with significant local changes. The gradient deviation of pH is calculated for the significant time nodes and fused with the time trend component. The common offset range of the time trend and local fluctuation is judged according to the fusion result. The above offset time range and spatial key area are combined to generate a distribution map of the initial weight value. The initial value weight is assigned according to the time period and spatial position respectively. All the initial value weight data are integrated to generate the weight initial value matrix of the feature offset window.

[0031] S312: using the initial weight matrix of the feature offset window, calculating the local time trend and spatial distribution difference in the feature sequence, adjusting the multi-feature weight values ​​in combination with the dynamic trend of the weight component change, and updating the initial value matrix of the global feature distribution using the multi-feature component weight change ratio to generate a weight update matrix that matches the distribution change; First, for the characteristic sequence of local time trends, the distribution ratio of each time period is dynamically normalized based on the initial value weight. The normalized ratio data is used to screen out the time periods with weight changes greater than a set value, and dynamic adjustment factors are added to the weight components of these time periods. The dynamic adjustment factors are obtained by comparing the difference gradients of the weight distribution within the time period and the distribution of the feature drift difference values. Then, for the spatial distribution characteristic sequence, the initial value matrix is ​​divided into different weight area blocks according to the region, and the normalization adjustment parameters are calculated based on the gradient distribution within the region. By calculating the normalized gradient values ​​of each area block, it is determined whether it is necessary to add additional weight distribution. The regional weight matrix is ​​adjusted according to the added weight distribution, and the normalization processing is performed again to fuse the time and space components to generate an updated matching matrix. The smoothing process is further performed and the distribution change value is calculated. Combined with the time and space dynamic adjustment parameters, a weight update matrix matching the distribution change is generated.

[0032] S313: Combine the weight update matrix of distribution change matching with the parameter matrix of the prediction model to calculate the drift matching amount of the feature distribution to the prediction weight, using the formula: ; Generate drift update weight matrix; in, represents the drift update weight matrix, Represents the initial weight value of the feature offset window, characterizing the initial weight distribution in multiple windows. Represents the prediction model parameter matrix component, characterizing the parameter value of the corresponding feature in the model, The adjustment coefficient of the local trend feature is used to dynamically adjust the time trend of the feature. The scaling factor that represents the distribution change is used to adjust the gain of the drift range to the weight. Indicates the offset difference of the feature sequence, characterizing the distribution offset change of the feature in the sequence. Represents the weight matrix smoothing adjustment item, which is used to avoid the calculation instability caused by extremely small or zero values. Represents the normalized offset adjustment coefficient, which is used to control the weight normalization of the overall feature distribution. Represents the total number of features.

[0033] formula: ; The benefit of the formula is that, through the multiple combinations of initial value weights, prediction model parameters and local time trends, and the introduction of smoothing adjustment terms and normalized offset adjustment coefficients, the formula's ability to match dynamic changes in distribution can be effectively enhanced, while avoiding the impact of zero or minimum values ​​on calculation stability.

[0034] Detailed explanation of the formula and the process of formula calculation and derivation: It is obtained by calculating the initial weight value of the feature offset window, whose initial value comes from the weight adjustment factor in paragraph 2. Set , is the parameter matrix component in the prediction model, obtained by model fitting , is the adjustment coefficient of the local time trend characteristics, which is calculated by the trend change range within the time period , is the scaling factor of the distribution change, and its dynamic range is set based on the characteristic drift gradient analysis. , is the offset difference of the characteristic sequence, which is calculated by integrating the offset range and amplitude , is the weight smoothing adjustment term, set , is the normalized offset adjustment coefficient, which is dynamically adjusted based on the distribution difference and is set to .

[0035] Calculation formula: 1. Calculate the sum of squares: ; 2. Add a smoothing adjustment and take the square root: ; 3. Calculate the denominator: ; 4. Sum and calculate the weight update matrix: ; The result shows that the drift update weight matrix value calculated by the formula is 14.121. Combined with the formula structure, it can be seen that the distribution of the overall weight matrix is ​​optimized through the dynamic changes of the initial value of the feature offset window, the time trend adjustment coefficient and the normalization adjustment item. This numerical result can be directly used as the input content for subsequent weight matrix optimization.

[0036] See also Figure 5 ,The specific steps for obtaining the dynamic infection risk value are: S411: Update the weight matrix based on drift, adjust the parameters of the infection prediction model, call the healing time series, regional temperature difference series and body temperature dynamic change data in the model, calculate the risk impact factor of each feature, and generate a preliminary risk impact matrix; By extracting time distribution feature points from the healing time series, the change in the time period is calculated with one day as the time unit, and the change is matched with the data to calculate the degree of difference. The data in the regional temperature difference series is divided into high temperature area and low temperature area. By comparing the absolute value of the temperature difference between the high and low temperature areas, the areas with temperature difference fluctuation values ​​greater than a specific threshold are screened. When calculating the dynamic change data of body temperature, hourly data points are used to record the body temperature changes. The fluctuation characteristics are quantified by calculating the variance of the data series, and the fluctuation variance value is used as the indicator quantification value of the dynamic change of body temperature. The above three groups of characteristic values ​​are normalized and used as risk impact factors to participate in the calculation to generate a preliminary risk impact matrix.

[0037] S412: using the preliminary risk impact matrix, analyzing the impact of the dynamic changes of the time characteristic distribution and the regional temperature difference in the healing time series on the body temperature data, calculating the risk correlation between the distributions, dynamically adjusting the risk component weights according to the correlation, and generating a dynamically adjusted risk correlation matrix; By extracting the time distribution eigenvalues ​​and the regional temperature difference eigenvalues, a linear relationship model is established. The regression coefficient is used as the quantitative value of risk association, the residual of the regression model is quantified as the model error, and the offset weight of the regional temperature difference eigenvalue to the time distribution eigenvalue is adjusted. The variance value of the dynamic change of body temperature is normalized and introduced into the above model, the risk component weight is readjusted, the adjusted weight component is brought into the preliminary risk impact matrix, the dynamic risk association quantification value is recalculated, and a dynamically adjusted risk association matrix is ​​formed.

[0038] S413: Based on the dynamically adjusted risk association matrix, the dynamic prediction value of infection risk is calculated, and the association calculation is performed in combination with the dynamic weight of the risk component, the time distribution characteristics and the regional temperature difference distribution, using the formula: ; Generate dynamic infection risk value; in, Represents the dynamic infection risk value, characterizing the risk assessment of the characteristic component in dynamic infection prediction, Represents the dynamic weight of the risk component, characterizing the weight ratio of each component in the infection prediction model. The influencing factors representing the time distribution characteristics characterize the dynamic characteristic changes in the healing time series. The risk influencing factor representing the regional temperature difference distribution characterizes the contribution of regional temperature difference in infection prediction. The risk fluctuation amount representing the dynamic change of body temperature, characterizing the impact of body temperature characteristics on the overall infection risk, Represents the smoothing adjustment parameter, which is used to correct the weight effect of feature fluctuations. Represents the total number of components, Represents the number of the current feature component.

[0039] formula: ; The benefit of the formula is that by introducing the square term of the regional temperature difference distribution and the absolute value of the dynamic change factor of the time characteristic into the calculation, the coupling correlation of regional and time characteristics in the infection risk calculation is effectively enhanced, and the interference of fluctuating data on the overall prediction results is reduced by using smoothing adjustment parameters.

[0040] Detailed explanation of the formula and the process of formula calculation and derivation: In the formula, the process of obtaining each parameter is set as follows: is the dynamic weight of the risk component, which is calculated by extracting the associated distribution weights of the healing time series and the regional temperature difference series data. The distribution weight of the healing time series is calculated by the ratio of the daily change to the standard, and the distribution weight of the regional temperature difference series is calculated by the ratio of the temperature difference between the high temperature area and the low temperature area to the total temperature difference. is the influencing factor of the time distribution characteristics, which is calculated by the slope of the healing time series. The calculation formula is: ; in, is the change in the eigenvalue in the time series, is the time interval; is the risk impact factor of regional temperature difference distribution, which is quantified as the risk impact factor by calculating the standard deviation of temperature difference between high and low temperature areas. The formula is: ; in, For the The regional temperature difference, is the mean temperature difference, is the number of regions; is the risk fluctuation of dynamic changes in body temperature, which is calculated by the normalized variance of body temperature data. The calculation formula is: ; in, is the standard deviation of the body temperature data, is the average value of body temperature data; In order to smoothly adjust the parameters, the residual amount is calculated by the dynamic normalization of the model, which represents the adjustment amount of the influence range of the feature change.

[0041] Set the value: . Substitute into the formula to calculate: The first step is to calculate the numerator: ; The second step is to calculate the denominator: ; The third step is to calculate the formula result: ; The results show that the dynamic infection risk value is approximately 9.447, indicating that under the current data conditions of healing time, regional temperature difference and dynamic changes in body temperature, the dynamic prediction value of infection risk is high, which can be further used as the basis for subsequent dynamic risk assessment and distribution analysis.

[0042] See also Figure 6 , the specific steps for obtaining infection risk warning indicators are: S511: Based on the dynamic infection risk value, combined with the skin status data around the stoma, excrement characteristics and healing process data, the impact of the differentiated component data on the infection risk level is evaluated in groups, and the change characteristics of the skin status and excrement in each group of data are called to generate a group risk assessment matrix; For the grouped data, we first extracted the redness and swelling area and humidity fluctuation values ​​in the skin condition data, and evaluated whether the healing condition was stable by calculating the mean redness and swelling area and humidity change of each group. The specific quantification of the redness and swelling area can be obtained through the area change rate formula Calculate, where and Represents the redness and swelling area values ​​at the current time and the previous time respectively, and the humidity fluctuation is expressed by its standard deviation measure, represents the mean humidity value, Represents the value of each data point of humidity; for the characteristic data of excrement, extract its pH change sequence, calculate the mean and maximum fluctuation value of pH change in each group, and express them as and ; Combining the above data, a feature matrix containing the mean value of redness and swelling area, humidity fluctuation value, pH mean and maximum fluctuation value was constructed for each group. After normalizing each eigenvalue in the matrix, the infection risk score of each group was calculated using the weighted average formula. The formula is: ,in The weight of each feature item is dynamically adjusted according to the influence of each feature on the infection risk. The score values ​​of different groups are compared through the infection risk score to generate a group risk assessment matrix.

[0043] S512: Using the group risk assessment matrix, filter the group data that exceeds the warning range, calculate the distribution deviation amplitude through the dynamic correlation characteristics between the risk components, compare the change trends of the multi-component dynamic characteristics, and call the adjustment threshold within the warning range to generate a risk warning screening matrix; By calculating the variation range of redness and swelling area and humidity fluctuation in the high-risk group, it is evaluated whether there is abnormal fluctuation of redness and swelling and humidity in the group. The variation range of redness and swelling area is calculated by its standard deviation formula The amplitude of humidity fluctuation is calculated by the difference between the maximum and minimum values. , further analyze the abnormal fluctuation characteristics of excrement pH, and use its fluctuation rate formula To measure whether the volatility exceeds the set threshold, the dynamic correlation features are used to analyze the correlation of each feature in the group. The dynamic features of redness and swelling area and humidity fluctuation in the high-risk group are calculated through the covariance formula of the two. Make a judgment, combine the warning range data to filter out the groups with larger deviations of dynamic characteristics, and generate a risk warning screening matrix.

[0044] S513: Based on the risk alert screening matrix, the risk data is compared with the set alert threshold, the distribution of risk data is analyzed, and quantitative analysis is performed in combination with the comparison of risk level data and threshold, and the risk grouping data beyond the alert range is obtained using the formula: ; Calculate and generate infection risk warning indicators; in, It represents the infection risk warning indicator and is a quantitative representation of the risk level. Represents the risk value of each group of data, which is used to determine the absolute amount of risk for each group. Represents the set warning threshold, which is the predetermined upper limit of risk tolerance. Represents the standard deviation of all risk data and is used to measure the volatility of risk. Represents the correction factor for each set of data, which is used to adjust the sensitivity of each set of risk values. Represents the number of data groups used to calculate the average warning indicator.

[0045] formula: ; The benefit of the formula is that by introducing the group correction factor Adjust the sensitivity of risk data, taking into account the risk bias and volatility standard deviation within the group Accurately measure the risk level of each group of data, so as to more accurately calculate the infection risk warning indicators; Detailed explanation of the formula and the process of formula calculation and derivation: Set the risk data group value The warning threshold is [1.8, 2.2, 3.0, 2.5]. is 2.0, standard deviation =0.5, correction factor They are [0.1, 0.15, 0.12, 0.18] respectively, and the total number of groups =4, the calculation is as follows: First, calculate the absolute value of risk deviation for each set of data: ; Then calculate the denominator: ; Calculate the risk ratio item by item: ; Finally, average the results: ; The result shows that the current risk warning index is 0.747. By comparing with the benchmark value or warning index range (0.5~1.0), it shows that the risks of some groups are close to the high warning range. Further key monitoring or adjustments should be carried out on high-risk groups to generate infection risk warning indicators.

[0046] A system for monitoring and early warning of infection after colostomy, which is used to implement the above-mentioned method for monitoring and early warning of infection after colostomy, and comprises: The feature extraction module extracts the redness and swelling area trend component and humidity fluctuation component based on the redness and swelling area data, wound humidity data, exudate volume data and excreta pH data collected after enterostomy surgery, combines the trends of exudate and excreta pH changes, and generates decomposition feature trend values; The feature drift module groups the swelling diffusion rate data, humidity change rate data, and exudate acid-base gradient data based on the decomposed feature trend value, calculates the divergence difference and moving distance according to the set sliding time window, compares the differences between adjacent windows, determines the offset time point and drift degree, and generates the feature drift difference value; Based on the characteristic drift difference value, the infection assessment module calls the colostomy healing time data, regional temperature difference data and tissue pH change data in the offset window, calculates the weight value of multiple data, adjusts the weight distribution to match the drift trend distribution, and generates a drift update weight matrix; The infection prediction module calculates and aggregates the dynamic distribution of infection risk based on the drift update weight matrix, combined with the healing time data, regional temperature difference data and body temperature dynamic change data, obtains the dynamic prediction value of infection risk, and generates the dynamic prediction value of infection risk; The early warning analysis module is based on the dynamic prediction value of infection risk, combined with the skin status data around the stoma, excrement characteristics data and healing process data, to evaluate the infection risk level in groups, screen the risk data and compare it with the warning threshold to generate infection risk warning indicators.

[0047] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for monitoring and early warning of post-intestinal ostomy infection, characterized in that: The following steps are involved: S1: Based on the skin condition around the stoma and the characteristic data of stoma excretion collected after the patient's stoma surgery, the time series of redness and swelling area, wound humidity, exudate volume and excreta pH were decomposed, the trend of redness and swelling area and the fluctuation component of humidity were extracted, and the changes in pH of exudate and excreta were combined to generate the decomposition characteristic trend value; S2: Based on the decomposed characteristic trend value, the sliding time window technology is used to process the swelling diffusion rate, the humidity change rate and the exudate acid-base gradient, the divergence difference and the moving distance of each window are calculated, and it is determined whether the distribution difference between the windows reaches the offset threshold, the time point of the offset is determined, and the characteristic drift difference value is generated; S3: Based on the feature drift difference value, weight calculation is performed on the colostomy healing process data within the feature offset window, the changes in healing time, regional temperature difference and tissue pH are evaluated, the weight components are adjusted to match the distribution changes, the parameter matrix of the prediction model is updated, and a drift update weight matrix is ​​generated; S4: Based on the drift update weight matrix, adjust the parameters of the infection prediction model, integrate the updated healing time, regional temperature difference and body temperature dynamic change data, calculate the dynamic prediction value of infection risk, analyze the distribution of risk data, and generate a dynamic infection risk value; S5: Based on the dynamic infection risk value, combined with the skin condition around the stoma, excrement characteristics and healing process data, the infection risk level is evaluated by grouping, and the risk group data exceeding the warning range is screened and compared with the set warning threshold to generate an infection risk warning indicator.

2. The method for monitoring and early warning of post-operative infection of enterostomy according to claim 1, characterized in that: The decomposed characteristic trend value specifically includes the trend of redness and swelling area, humidity fluctuation component, and change in pH of exudate. The characteristic drift difference value includes divergence difference, moving distance, and time point determination. The drift update weight matrix specifically refers to the healing time weight, regional temperature difference weight, and tissue pH weight. The dynamic infection risk value includes dynamic data of healing time, dynamic data of regional temperature difference, and dynamic change of body temperature. The infection risk warning indicator specifically includes infection risk level, risk grouping data, and warning threshold comparison.

3. The method for monitoring and early warning of post-enterostomy infection according to claim 2, characterized in that: The steps for obtaining the decomposition characteristic trend value are specifically as follows: S111: according to the redness and swelling area time series data of the skin condition around the enterostomy, the difference between the local weighted average value and the global time series is calculated in segments, the time trend feature is judged by using the relationship between the difference and the dynamic threshold, the significant change part of the redness and swelling area in the time series is extracted, and the normalization operation is performed based on the characteristics of the significant change part to generate the redness and swelling area trend feature value; S112: Based on the wound humidity time series data, the humidity fluctuation amplitude is processed by dynamic window smoothing to construct a product sequence of humidity change frequency and amplitude. At the same time, the overall fluctuation intensity characteristic value is calculated by combining the fluctuation amplitude and frequency. The fluctuation characteristic value is used to filter the part exceeding the set threshold and adjust the result to generate the humidity fluctuation characteristic value; S113: Combining the redness and swelling area trend characteristic value and the humidity fluctuation characteristic value, based on the two characteristic values ​​and the time series parameter of the pH change of the excrement, using the formula: ； Generate decomposition feature trend values; in, represents the decomposition characteristic trend value, Represents the characteristic value of the redness and swelling area trend, represents the humidity fluctuation frequency, Represents the humidity fluctuation amplitude, represents the eigenvalue weight, represents the pH change adjustment factor, Represents the range of pH value of excrement. Represents the characteristic component number in the time series, Represents the total number of feature components.

4. The method for monitoring and early warning of post-operative infection of enterostomy according to claim 3, characterized in that: The steps for obtaining the characteristic drift difference value are specifically as follows: S211: Based on the redness and swelling diffusion rate, humidity change rate and exudate acid-base gradient in the decomposed characteristic trend value, a sliding time window is set, a divergence difference is calculated according to the characteristic parameters of each window, and a relationship between the divergence difference and the distribution threshold is used to filter the part that meets the conditions in the window, and a divergence difference sequence is extracted to generate a divergence difference sequence of the redness and swelling diffusion rate and the humidity change rate; S212: Based on the divergence difference sequence of the redness and swelling diffusion rate and the humidity change rate, the moving distance of adjacent parameters between windows is calculated, the difference of the moving distance between windows is used to construct a normalized index, and the windows exceeding the offset threshold are screened in combination with the distribution characteristics of the moving distance, and the parameters in the offset window are used to readjust the feature range to generate an offset window sequence; S213: Combining the divergence difference sequence and the offset window sequence, calculating the time point and amplitude of the feature drift in each window, using the formula: ； Generate feature drift difference value; in, represents the feature drift difference value, Represents the divergence difference of the current window, represents the divergence difference of the previous window, represents the weighted weight, Represents the moving distance between windows, represents the drift time point, represents the time adjustment factor, Represents the total number of time windows.

5. The method for monitoring and early warning of post-enterostomy infection according to claim 4, characterized in that: The steps for obtaining the drift update weight matrix are specifically as follows: S311: based on the characteristic drift difference value, extract the intestinal stoma healing process data in the characteristic offset window, call the healing time series, regional temperature difference series and tissue pH series, calculate the initial weight value in the window for each characteristic data according to the change trend of time and space distribution, and generate the initial weight value matrix of the characteristic offset window; S312: using the initial weight value matrix of the feature offset window, calculating the local time trend and spatial distribution difference in the feature sequence, adjusting the multi-feature weight values ​​in combination with the dynamic trend of the weight component change, and using the multi-feature component weight change ratio to update the initial value matrix of the global feature distribution to generate a weight update matrix that matches the distribution change; S313: Combining the weight update matrix of the distribution change matching with the parameter matrix of the prediction model, the drift matching amount of the feature distribution to the prediction weight is calculated using the formula: ； Generate drift update weight matrix; in, represents the drift update weight matrix, Represents the initial weight value of the feature offset window, characterizing the initial weight distribution in multiple windows. Represents the prediction model parameter matrix component, characterizing the parameter value of the corresponding feature in the model, The adjustment coefficient of the local trend feature is used to dynamically adjust the time trend of the feature. The scaling factor that represents the distribution change is used to adjust the gain of the drift range to the weight. Indicates the offset difference of the feature sequence, characterizing the distribution offset change of the feature in the sequence. Represents the weight matrix smoothing adjustment item, which is used to avoid the calculation instability caused by extremely small or zero values. Represents the normalized offset adjustment coefficient, which is used to control the weight normalization of the overall feature distribution. Represents the total number of features.

6. The method for monitoring and early warning of post-enterostomy infection according to claim 5, characterized in that: The steps for obtaining the dynamic infection risk value are specifically as follows: S411: Based on the drift update weight matrix, adjust the parameters of the infection prediction model, call the healing time series, regional temperature difference series and body temperature dynamic change data in the model, calculate the risk impact factor of each feature, and generate a preliminary risk impact matrix; S412: using the preliminary risk impact matrix, analyzing the impact of the dynamic changes of the time characteristic distribution and the regional temperature difference in the healing time series on the body temperature data, calculating the risk correlation between the distributions, dynamically adjusting the risk component weights according to the correlation, and generating a dynamically adjusted risk correlation matrix; S413: Based on the dynamically adjusted risk association matrix, the dynamic prediction value of the infection risk is calculated, and the association calculation is performed in combination with the dynamic weight of the risk component, the time distribution characteristics and the regional temperature difference distribution, using the formula: ； Generate dynamic infection risk value; in, Represents the dynamic infection risk value, characterizing the risk assessment of the characteristic component in dynamic infection prediction, Represents the dynamic weight of the risk component, characterizing the weight ratio of each component in the infection prediction model. The influencing factors representing the time distribution characteristics characterize the dynamic characteristic changes in the healing time series. The risk influencing factor representing the regional temperature difference distribution characterizes the contribution of regional temperature difference in infection prediction. The risk fluctuation amount representing the dynamic change of body temperature, characterizing the impact of body temperature characteristics on the overall infection risk, Represents the smoothing adjustment parameter, which is used to correct the weight effect of feature fluctuations. Represents the total number of components, Represents the number of the current feature component.

7. The method for monitoring and early warning of post-enterostomy infection according to claim 6, characterized in that: The steps for obtaining the infection risk warning indicator are specifically as follows: S511: Based on the dynamic infection risk value, combined with the skin status data around the stoma, excrement characteristics and healing process data, the impact of the differentiated component data on the infection risk level is evaluated in groups, and the change characteristics of the skin status and excrement in each group of data are called to generate a group risk assessment matrix; S512: Using the group risk assessment matrix, filter the group data that exceeds the warning range, calculate the distribution offset amplitude through the dynamic correlation characteristics between risk components, compare the change trends of multi-component dynamic characteristics, and call the adjustment threshold within the warning range to generate a risk warning screening matrix; S513: Based on the risk alert screening matrix, the risk data is compared with the set alert threshold, the distribution of risk data is analyzed, and quantitative analysis is performed in combination with the comparison between the risk level data and the threshold to obtain the risk grouping data that exceeds the alert range, using the formula: ； Calculate and generate infection risk warning indicators; in, It represents the infection risk warning indicator and is a quantitative representation of the risk level. Represents the risk value of each group of data, which is used to determine the absolute amount of risk for each group. Represents the set warning threshold, which is the predetermined upper limit of risk tolerance. Represents the standard deviation of all risk data and is used to measure the volatility of risk. Represents the correction factor for each set of data, which is used to adjust the sensitivity of each set of risk values. Represents the number of data groups used to calculate the average warning indicator.

8. A post-operative infection monitoring and early warning system for enterostomy, characterized in that: According to the method for monitoring and early warning of post-enterostomy infection according to any one of claims 1 to 7, the system comprises: The feature extraction module extracts the redness and swelling area trend component and humidity fluctuation component based on the redness and swelling area data, wound humidity data, exudate volume data and excreta pH data collected after enterostomy surgery, combines the trends of exudate and excreta pH changes, and generates decomposition feature trend values; The feature drift module groups the swelling diffusion rate data, the humidity change rate data and the exudate acid-base gradient data based on the decomposed feature trend value, calculates the divergence difference and the moving distance according to the set sliding time window, compares the difference between adjacent windows and determines the offset time point and the drift degree, and generates a feature drift difference value; The infection assessment module calls the enterostomy healing time data, regional temperature difference data and tissue pH change data within the offset window based on the characteristic drift difference value, calculates weight values ​​for multiple data, adjusts the weight distribution to match the drift trend distribution, and generates a drift update weight matrix; The infection prediction module calculates and aggregates the dynamic distribution of infection risk based on the drift update weight matrix, combines the healing time data, the regional temperature difference data and the dynamic change data of body temperature, obtains the dynamic prediction value of infection risk, and generates the dynamic prediction value of infection risk; The early warning analysis module evaluates the infection risk level in groups based on the dynamic prediction value of infection risk, combined with the skin status data around the stoma, the fecal characteristic data and the healing process data, screens the risk data and compares it with the warning threshold to generate an infection risk early warning indicator.

Images