A method for dynamically monitoring and early warning analysis of traditional Chinese medicine syndromes of diabetes

Abstract

This application discloses a method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes, including the following steps: (1) construction of basic weights for primary and secondary syndromes; (2) time-series data acquisition and preprocessing; (3) time-series weight drift correction; (4) data lag correction; (5) gradual graded monitoring; (6) removal of interference from secondary syndromes and accurate early warning output. According to the method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes of this application, by constructing a time-series weight drift correction model, the dynamic weights of primary and secondary syndromes are updated in real time using a sliding window iterative algorithm, accurately capturing the time-series drift characteristics of syndrome weights with disease course and body state, effectively solving the problem of syndrome primary and secondary discrimination bias caused by fixed weights in the prior art, and introducing a time-series lag compensation factor, determining the value of the compensation factor through time-series correlation analysis and establishing a lag correction model.

Description

Technical Field

[0001] This application relates to the field of intelligent syndrome differentiation and medical time-series data analysis in traditional Chinese medicine, and in particular to a method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes. Background Technology

[0002] Diabetes mellitus is a highly prevalent chronic metabolic disease with a long course and complex pathogenesis. Long-term glucose metabolism disorders can easily lead to dysfunction of multiple organs. Traditional Chinese medicine (TCM) has unique advantages in the intervention of diabetes mellitus and the stage-by-stage control of the disease. The core of clinical diagnosis and treatment relies on the TCM syndrome differentiation system, which analyzes the body's deficiency and excess, cold and heat, phlegm and dampness, blood stasis and other pathogenesis states to achieve individualized conditioning and disease management.

[0003] Currently, most clinical TCM diagnostic methods rely on in-person consultations with physicians, using periodic questioning and manual interpretation of the tongue and pulse to determine static symptoms. This only reflects the patient's instantaneous physical state at a specific moment during a consultation and cannot achieve long-term, continuous dynamic tracking of symptoms. The pathogenesis of diabetes is not fixed; it generally exhibits a pattern of constant primary symptoms, overlapping and continuously changing secondary symptoms, influenced by diet, lifestyle, blood sugar fluctuations, medication interventions, and disease progression. Conventional static diagnostic methods struggle to capture the slow, evolving process of symptoms and are insufficient in identifying potential worsening trends.

[0004] Existing intelligent monitoring solutions for TCM syndromes in diabetes mainly employ fixed syndrome classification standards and fixed weighting models. These solutions lack the ability to differentiate between complex and concurrent syndromes, cannot quantify the proportion of influence of the primary and secondary syndromes on the overall pathogenesis, and cannot adapt to the dynamic changes in syndrome influence weights throughout the disease course. Furthermore, existing monitoring models often use threshold-triggered mutation warning mechanisms, identifying only obvious abnormal states of the syndromes while ignoring the gradual worsening process, making it difficult to effectively capture early, latent pathological abnormalities.

[0005] Therefore, a method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes is proposed. Summary of the Invention

[0006] This application aims to at least partially solve one of the technical problems in the aforementioned technologies.

[0007] To achieve the above objectives, the first aspect of this application proposes a method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes, comprising the following steps:

[0008] (1) Construction of basic weights for primary and secondary syndromes: Based on the clinical syndrome database of diabetes, the basic types of common primary syndromes and secondary syndromes of diabetes are divided, quantitative indicators corresponding to each syndrome are extracted, and the initial basic weights of each syndrome are determined by the analytic hierarchy process, and a basic weight matrix of primary and secondary syndromes is established.

[0009] (2) Time series data acquisition and preprocessing: Time series monitoring data of diabetic patients are accessed. The time series monitoring data includes quantitative scores of TCM symptoms, quantitative parameters of tongue and pulse, and physicochemical indicators of Western medicine. Outliers in the data are removed using the 3σ criterion. The data after removing outliers is standardized by the Z-score standardization method to obtain standardized time series data. Among them, the physicochemical indicators of Western medicine include at least blood glucose and glycated hemoglobin indicators.

[0010] (3) Time series weight drift correction: Based on the standardized time series data, a time series weight drift correction model is constructed. With time series as input, a sliding window iterative algorithm is used to iteratively update the dynamic weights of the main certificate and the secondary certificate in real time, and generate the time series weight change curves of the main certificate and the secondary certificate.

[0011] (4) Data lag correction: A time lag compensation factor is introduced. The time lag compensation factor is determined by the time correlation analysis between Western medicine physicochemical indicators and TCM syndrome quantitative indicators. A lag correction model is established based on the time lag compensation factor to correct the time misalignment of the standardized time series data and eliminate data correlation distortion.

[0012] (5) Gradual grading monitoring: Based on the time-series weight change curve of the main and concurrent symptoms, at least three gradual grading thresholds for the deterioration of symptoms are set. By calculating the slope of the time-series weight change of the main and concurrent symptoms and the amplitude of weight fluctuation, early minor abnormalities in the gradual grading process of symptoms are identified, and the grading determination of the evolution state of symptoms is completed.

[0013] (6) Concurrent syndrome interference removal and precise early warning output: Set a concurrent syndrome weight threshold, remove concurrent syndrome weights below the concurrent syndrome weight threshold, retain the core weight of the main syndrome and the weight of high-impact concurrent syndromes, lock the deterioration trend of the main syndrome, and output precise early warning information by grade and syndrome based on the graded judgment results of the syndrome evolution state, and simultaneously output TCM intervention tendency prompts corresponding to the early warning level and syndrome type.

[0014] In addition, the method for dynamic monitoring and early warning analysis of TCM syndromes of diabetes proposed in this application may also have the following additional technical features:

[0015] As a further description of the above technical solution:

[0016] The primary syndrome types mentioned in step (1) include Qi and Yin deficiency syndrome, damp-heat syndrome, and turbid toxin accumulation syndrome. The secondary syndrome types include blood stasis syndrome, phlegm-dampness syndrome, and liver stagnation syndrome. The quantitative indicators corresponding to each syndrome include the quantitative score of TCM symptoms and the characteristic values ​​corresponding to the quantitative parameters of tongue and pulse.

[0017] As a further description of the above technical solution:

[0018] The sliding window iterative algorithm described in step (3) has a window size of 7-30 days and an iteration step size of 1 day. Each iteration is based on the standardized time series data in the current window to correct the dynamic weights of the main certificate and the secondary certificate.

[0019] As a further description of the above technical solution:

[0020] The time-series correlation analysis described in step (4) uses the Pearson correlation coefficient method to calculate the correlation coefficient between Western medicine physicochemical indicators and TCM syndrome quantitative indicators. Based on the correlation coefficient, the time difference between the evolution of Western medicine physicochemical indicators and TCM syndrome is determined, and then the value of the time-series lag compensation factor is determined.

[0021] As a further description of the above technical solution:

[0022] The gradual symptom deterioration classification thresholds mentioned in step (5) include mild warning threshold, moderate warning threshold and severe warning threshold. The slope of the weight change corresponding to the mild warning threshold is 0.02-0.05 / day, the slope of the weight change corresponding to the moderate warning threshold is 0.05-0.08 / day, and the slope of the weight change corresponding to the severe warning threshold is ≥0.08 / day.

[0023] As a further description of the above technical solution:

[0024] The threshold for the concurrent certification weight in step (6) is set to 0.1-0.2. Concurrent certification weights below the threshold are judged as redundant interference weights and are removed.

[0025] As a further description of the above technical solution:

[0026] The tongue pulse quantification parameters mentioned in step (2) include tongue color quantification value, tongue coating thickness quantification value, pulse frequency, and pulse amplitude. The TCM symptom quantification score is based on the TCM syndrome diagnosis standard, and each symptom is assigned a value according to its severity.

[0027] As a further description of the above technical solution:

[0028] The TCM intervention tendency prompts mentioned in step (6) include dietary conditioning suggestions, TCM conditioning directions, and emotional regulation suggestions, and the intervention tendency prompts correspond to the current main symptom type and warning level.

[0029] Advantages of this invention:

[0030] According to this application, a method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes is proposed. By constructing a time-series weight drift correction model, the dynamic weights of the primary and secondary syndromes are updated in real time using a sliding window iterative algorithm. This accurately captures the time-series drift characteristics of syndrome weights with disease progression and body state, effectively solving the problem of syndrome primary and secondary discrimination bias caused by fixed weights in existing technologies. At the same time, a time-series lag compensation factor is introduced. The value of the compensation factor is determined through time-series correlation analysis, and a lag correction model is established to specifically correct the time misalignment between Western medicine physicochemical indicators and TCM syndrome evolution. This completely eliminates the monitoring and analysis distortion caused by data synchronization in existing technologies. Furthermore, time-series data preprocessing ensures the accuracy of monitoring data. Based on the time-series weight change curves of primary and secondary syndromes, a gradual grading standard is set to identify early micro-abnormalities of syndromes to avoid missed early warnings. By setting a weight threshold for secondary syndromes, redundant interference weights are removed to lock the deterioration trend of the primary syndrome and output accurate early warning information and corresponding TCM intervention tendency suggestions by grade and syndrome. This significantly improves the accuracy of syndrome monitoring and the sensitivity of early warning, and enhances the clinical applicability of the early warning results.

[0031] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0032] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0033] Figure 1 This is a general flowchart of a method for dynamic monitoring and early warning analysis of TCM syndromes of diabetes according to an embodiment of this application;

[0034] Figure 2 This is a flowchart of the main and secondary syndrome time-series weight drift correction method for a dynamic monitoring and early warning analysis method for diabetes TCM syndromes according to an embodiment of this application;

[0035] Figure 3 This is a flowchart of the time lag correction process for a method for dynamic monitoring and early warning analysis of diabetes TCM syndromes according to an embodiment of this application;

[0036] Figure 4 This is a flowchart of syndrome classification and concurrent syndrome interference removal determination of a method for dynamic monitoring and early warning analysis of diabetes TCM syndromes according to an embodiment of this application. Detailed Implementation

[0037] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0038] The following describes a method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes according to an embodiment of this application, with reference to the accompanying drawings.

[0039] like Figure 1-4 As shown in Embodiment 1 of this application, a method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes includes the following steps:

[0040] (1) Construction of the basic weights of the primary and secondary securities:

[0041] Based on the clinical syndrome database of diabetes, the specific types of primary and secondary syndromes are identified, quantitative indicators corresponding to each syndrome are extracted, and the initial basic weights of each syndrome are quantified by the analytic hierarchy process (AHP). A basic weight matrix of primary and secondary syndromes is constructed to provide a benchmark for subsequent time-series weight drift correction and to solve the defects of fixed weights and lack of quantitative basis in the existing technology.

[0042] Based on the "Guidelines for the Prevention and Treatment of Diabetes in Traditional Chinese Medicine" and clinical diagnosis and treatment data, common primary syndrome types and concurrent syndrome types of diabetes are classified. The primary syndrome types include three categories: Qi and Yin deficiency syndrome (A1), damp-heat syndrome (A2), and turbid toxin accumulation syndrome (A3). The concurrent syndrome types include three categories: blood stasis syndrome (B1), phlegm-dampness syndrome (B2), and liver stagnation syndrome (B3). This covers more than 95% of the concurrent syndrome types of diabetes in clinical practice, ensuring the comprehensiveness and clinical applicability of syndrome classification.

[0043] Quantitative indicators were extracted for each syndrome. The quantitative indicators for both the primary and secondary syndromes included the feature values ​​corresponding to the quantitative scores of TCM symptoms and the quantitative parameters of tongue and pulse, as detailed below:

[0044] Qi and Yin deficiency syndrome (A1): The corresponding symptom quantitative scores include fatigue (X11), dry mouth and throat (X12), excessive appetite and hunger (X13), and soreness and weakness of the lower back and knees (X14). The tongue and pulse quantitative parameters include the quantitative value of pale red tongue color (Y11), the quantitative value of thin white tongue coating (Y12), and the quantitative value of weak pulse (Z11).

[0045] Damp-heat syndrome (A2): The corresponding symptom quantitative scores include bitter taste in the mouth (X21), sticky stool (X22), dark yellow urine (X23), and heaviness in the limbs (X24). The tongue and pulse quantitative parameters include the quantitative value of red tongue color (Y21), the quantitative value of yellow and greasy tongue coating (Y22), and the quantitative value of slippery pulse (Z21).

[0046] Turbid toxin accumulation syndrome (A3): The corresponding symptom quantitative scores include halitosis (X31), abdominal distension (X32), skin itching (X33), and constipation (X34). The tongue and pulse quantitative parameters include the purplish-dark tongue color quantitative value (Y31), the thick and greasy tongue coating quantitative value (Y32), and the deep and hesitant pulse quantitative value (Z31).

[0047] Blood stasis syndrome (B1): The corresponding symptom quantitative scores include limb numbness (X41) and sallow complexion (X42). The tongue and pulse quantitative parameters include the quantitative value of ecchymosis on the tongue (Y41) and the quantitative value of hesitant pulse (Z41).

[0048] Phlegm-dampness syndrome (B2): Corresponding symptom quantitative scores include cough with excessive phlegm (X51) and chest tightness (X52), and tongue and pulse quantitative parameters include white and greasy tongue coating quantitative value (Y51) and soft and slow pulse quantitative value (Z52).

[0049] Liver Qi stagnation syndrome (B3): The corresponding symptom quantitative scores include irritability and anger (X61) and hypochondriac distending pain (X62). The tongue and pulse quantitative parameters include the quantitative value of red tongue color (Y61) and the quantitative value of wiry and tense pulse (Z61).

[0050] The quantitative scores for each symptom mentioned above are based on the diagnostic criteria for syndromes in Traditional Chinese Medicine (TCM) and are assigned according to severity: 0 points for no symptom, 1 point for mild, 2 points for moderate, and 3 points for severe. The quantitative parameters for tongue and pulse are collected by a tongue and pulse quantitative acquisition device and converted into quantitative values ​​between 0 and 1 by the device's built-in algorithm. The larger the quantitative value, the more obvious the corresponding syndrome characteristics.

[0051] Then, the Analytic Hierarchy Process (AHP) is used to determine the initial basic weights of each primary and secondary syndrome. The core principle is to construct a judgment matrix to quantify the relative importance of each syndrome. After consistency testing, reasonable initial weights are obtained, avoiding weight deviations caused by subjective assignment. The specific calculation process is as follows:

[0052] Constructing a judgment matrix: Five senior TCM physicians were invited to construct judgment matrices for the main syndrome and concurrent syndromes based on the degree of influence of each syndrome on the evolution of diabetes pathogenesis. The 1-9 scale method was used (1 indicates that the two syndromes are of equal importance, 3 indicates that the former is slightly more important than the latter, 5 indicates that the former is significantly more important than the latter, 7 indicates that the former is strongly more important than the latter, 9 indicates that the former is extremely more important than the latter, and 2, 4, 6, and 8 are the median values ​​of the above adjacent scales). The average score of the five physicians was taken as the final judgment matrix.

[0053] The judgment matrix M1 between the main evidence is as follows:

[0054] M1 = [[1, 2, 3],[1 / 2, 1, 2],[1 / 3, 1 / 2, 1]]

[0055] The judgment matrix M2 between the two evidences is as follows:

[0056] M2 = [[1, 3, 2],[1 / 3, 1, 1 / 2],[1 / 2, 2, 1]]

[0057] Weight Calculation: The eigenvectors of the judgment matrix are calculated using the sum-product method, serving as the initial basic weights for each syndrome. The specific formula is as follows:

[0058] First, normalize each column of the judgment matrix to obtain the normalized matrix. ,in ( The number of primary or secondary evidences, and the primary evidence. =3, further proof =3);

[0059] Secondly, summation is performed on each row of the normalized matrix N to obtain the row sum vector. , ;

[0060] Finally, the row and vector S are normalized to obtain the initial basic weight vector. , .

[0061] Calculate the initial base weight of the main certificate using the formula above:

[0062] Normalizing column M1 yields N1:

[0063] N1 First Column:

[0064] [1 / (1+0.5+0.333)=0.545, 0.5 / (1+0.5+0.333)=0.273, 0.333 / (1+0.5+0.333)=0.182]

[0065] N1, second column:

[0066] [2 / (2+1+0.5)=0.571, 1 / (2+1+0.5)=0.286, 0.5 / (2+1+0.5)=0.143]

[0067] N1, third column:

[0068] [3 / (3+2+1)=0.5, 2 / (3+2+1)=0.333, 1 / (3+2+1)=0.167]

[0069] Calculate row sum S1:

[0070] S11=0.545+0.571+0.5=1.616, S12=0.273+0.286+0.333=0.892, S13=0.182+0.143+0.167=0.492

[0071] Normalization yields the initial basic weight W1 of the main evidence:

[0072] W11=1.616 / (1.616+0.892+0.492)=0.539,

[0073] W12=0.892 / (1.616+0.892+0.492)=0.297,

[0074] W13=0.492 / (1.616+0.892+0.492)=0.164.

[0075] The initial base weights for Qi and Yin deficiency syndrome (A1) are 0.539, for damp-heat syndrome (A2) it is 0.297, and for turbid toxin accumulation syndrome (A3) it is 0.164.

[0076] Similarly, calculate the initial foundation weights for both verification and application:

[0077] Normalizing column M2, we get N2:

[0078] N2 First Column:

[0079] [1 / (1+0.333+0.5)=0.545, 0.333 / (1+0.333+0.5)=0.182, 0.5 / (1+0.333+0.5)=0.273]

[0080] N2 second column:

[0081] [3 / (3+1+2)=0.5, 1 / (3+1+2)=0.167, 2 / (3+1+2)=0.333]

[0082] N2, third column:

[0083] [2 / (2+0.5+1)=0.571, 0.5 / (2+0.5+1)=0.143, 1 / (2+0.5+1)=0.286]

[0084] Calculate row sum S2:

[0085] S21=0.545+0.5+0.571=1.616, S22=0.182+0.167+0.143=0.492, S23=0.273+0.333+0.286=0.892

[0086] Normalization yields the initial basic weights W2, which also support the following:

[0087] W21=1.616 / (1.616+0.492+0.892)=0.539,

[0088] W22=0.492 / (1.616+0.492+0.892)=0.164,

[0089] W23=0.892 / (1.616+0.492+0.892)=0.297

[0090] The initial baseline weights for blood stasis syndrome (B1) were 0.539, for phlegm-dampness syndrome (B2) 0.164, and for liver stagnation syndrome (B3) 0.297.

[0091] Consistency check: To ensure the rationality of the judgment matrix, a consistency check is performed. The check formula is as follows: ,in To determine the largest eigenvalue of a matrix, The average random consistency index ( When =3, =0.58). When When <0.1, the judgment matrix meets the consistency requirement and the weight allocation is reasonable;

[0092] Calculate the largest eigenvalue of the main evidence judgment matrix M1. : ;

[0093] M1×W1=[1×0.539+2×0.297+3×0.164=1.617,0.5×0.539+1×0.297+2×0.164=0.89

[0094] [30.333×0.539+0.5×0.297+1×0.164=0.492];

[0095] =(1.617 / 0.539)+(0.893 / 0.297)+(0.492 / 0.164) ≈3.003;

[0096] 1 = (3.003-3) / (3-1) = 0.0015, CR1 = 0.0015 / 0.58 ≈ 0.0026 < 0.1, which meets the consistency requirement;

[0097] Similarly, calculate the largest eigenvalue of the concurrent judgment matrix M2. 2≈3.003, 2 = 0.0015, 2 = 0.0026 < 0.1, which meets the consistency requirement;

[0098] Constructing the basic weight matrix: Based on the initial basic weights of the main and secondary warrants calculated above, construct the basic weight matrix W of the main and secondary warrants. The matrix dimension is 6×1, as follows:

[0099]

[0100] Among them, the matrix elements correspond to the initial basic weights of Qi and Yin deficiency syndrome (A1), damp-heat syndrome (A2), turbid toxin accumulation syndrome (A3), blood stasis syndrome (B1), phlegm-dampness syndrome (B2), and liver stagnation syndrome (B3).

[0101] (2) Time series data acquisition and preprocessing:

[0102] By collecting patients' time-series monitoring data throughout the entire process, and eliminating data noise and unifying data scale through outlier removal and standardization, we can ensure the accuracy of subsequent weight drift correction and lag correction, and solve the problem of analysis distortion caused by poor data quality and inconsistent scale in existing technologies.

[0103] One hundred subjects were monitored daily for 90 days. The time-series monitoring data collected included quantitative scores of TCM symptoms, quantitative parameters of tongue and pulse, and physicochemical indicators of Western medicine. The specific data collection methods are as follows:

[0104] Traditional Chinese Medicine Symptom Quantification Scoring: Every day, patients use the TCM symptom quantification collection terminal to score the symptoms corresponding to each syndrome based on their own symptoms on that day, generating daily symptom quantification score data. The data format is the average of the scores of 4 items (main syndrome) or 2 items (concurrent syndrome) of 6 syndromes, that is, each syndrome corresponds to 1 symptom quantification score value per day.

[0105] Every day, medical staff use a tongue and pulse quantitative acquisition device to collect quantitative parameters corresponding to the patient's tongue color, tongue coating, and pulse. The device automatically outputs a quantitative value between 0 and 1. Each syndrome corresponds to an average value of one tongue and pulse quantitative parameter per day.

[0106] Blood glucose (GLU) data were collected 2 hours after breakfast daily, and glycated hemoglobin (HbA1c) data were collected every 15 days. Blood glucose was measured in mmol / L, and HbA1c in %%.

[0107] Taking one patient (P1, male, 56 years old, 6-year course of disease) with both Qi and Yin deficiency and blood stasis as an example, the partial time-series monitoring data for the first 15 days are shown in Table 1 below (only core data are shown; the complete data is for 90 days):

[0108] Number of monitoring days Qi and Yin Deficiency Syndrome (Symptoms / Tongue and Pulse) Blood stasis syndrome (symptoms / tongue and pulse) Blood glucose (mmol / L) Glycated hemoglobin (%) 1 1.8 / 0.62 1.2 / 0.58 8.2 7.5 2 1.7 / 0.60 1.3 / 0.59 8 - 3 1.9 / 0.63 1.1 / 0.57 8.5 - 4 1.7 / 0.61 1.2 / 0.58 7.9 - 5 1.8 / 0.62 1.3 / 0.59 8.1 - 6 2.5 / 0.75 1.4 / 0.60 9.8 - 7 2.4 / 0.73 1.3 / 0.59 9.5 - 8 2.3 / 0.71 1.4 / 0.60 9.2 - 9 2.2 / 0.69 1.3 / 0.59 8.9 - 10 2.1 / 0.67 1.2 / 0.58 8.6 - 11 2.0 / 0.65 1.3 / 0.59 8.3 - 12 1.9 / 0.63 1.2 / 0.58 8.1 - 13 1.8 / 0.62 1.1 / 0.57 7.9 - 14 1.7 / 0.60 1.2 / 0.58 7.8 - 15 1.8 / 0.62 1.2 / 0.58 7.7 7.4

[0109] Outlier removal: using The core principle of the criteria for removing outliers in time series data is: assuming the data follows a normal distribution, data exceeding the mean ± 3σ are considered outliers and removed to prevent them from interfering with subsequent analysis. The specific calculation process is as follows:

[0110] Calculate the mean of each monitoring indicator. and standard deviation The formula is: ,in For the number of monitoring days ( =90), For the first Daily monitoring data;

[0111] Set an outlier threshold: x_i that exceeds this range is considered an outlier and is removed.

[0112] Outlier replenishment: For outliers that have been removed, the average of the data from the next two days is used to replenish them, ensuring the continuity of the time series data;

[0113] Taking the symptom score of Qi and Yin deficiency syndrome in patient P1 as an example, the calculation process is as follows:

[0114] Symptom score data for the first 15 days:

[0115]

[0116] Calculate the mean : ;

[0117] Calculate the standard deviation σ: ;

[0118] Outlier thresholds: 1.947 - 3 × 0.245 ≈ 1.212, 1.947 + 3 × 0.245 ≈ 2.682;

[0119] In the data from the first 15 days, the maximum value was 2.5, which is less than 2.682, and the minimum value was 1.7, which is greater than 1.212. There were no outliers. If the data for a certain day was 3.0, it was determined to be an outlier, removed, and then supplemented with the average of the data from the two adjacent days.

[0120] The Z-score standardization method is used to standardize the time series data after removing outliers, unify the data scale, eliminate the differences in measurement units between different indicators, and ensure the rationality of subsequent weight calculations. The standardization formula is as follows:

[0121]

[0122] in, For the first Heavenly Standardized values ​​of each monitoring indicator For the first Heavenly The raw values ​​of each monitoring indicator For the first The average of the monitoring indicators, For the first Standard deviation of each monitoring indicator;

[0123] Taking the symptom score of patient P1 on day 1 of Qi Yin deficiency as an example, the standardized calculation is as follows:

[0124]

[0125] Similarly, the standardized values ​​of all monitoring data can be calculated to obtain standardized time series data. The standardized data range is between [-3, 3], ensuring that each indicator is comparable.

[0126] (3) Temporal weight drift correction:

[0127] Based on standardized time-series data, a time-series weight drift correction model is constructed. A sliding window iterative algorithm is adopted to iteratively update the dynamic weights of the primary and secondary symptoms in real time with time as the axis. The model captures the time-series drift characteristics of the weights with the course of the disease and the state of the body, which solves the problem of bias in the discrimination of the primary and secondary symptoms caused by the fixed weights in the existing technology. Its core logic is: as the monitoring data accumulates, the sliding window captures the recent data change trend and dynamically corrects the initial basic weights so that the weights can adapt to the evolution of the symptoms.

[0128] The specific implementation process is as follows:

[0129] A time-series weighted drift correction model is constructed. The model input is standardized time-series data (quantified scores of TCM symptoms, quantitative parameters of tongue and pulse, and standardized values ​​of Western medicine physicochemical indicators). The output is the dynamic weights of the main symptom and concurrent symptoms. The core formula of the model is as follows:

[0130]

[0131] in, For the first The dynamic weight matrix of the main and secondary evidence of the heavens, This is the dynamic weight matrix of the primary and secondary securities on day t-1. Weighting adjustment coefficient ( =0.05, which can be adjusted according to clinical practice to control the magnitude of weight correction and avoid over- or under-correction. For the first Standardized time series data difference matrix between day t and day t-1;

[0132] The calculation formula is:

[0133] in, For the first The standardized time-series data matrix of the day (dimension 6×1, corresponding to the standardized monitoring data of 6 syndromes in sequence). This is the standardized time-series data matrix for day t-1;

[0134] The sliding window iterative algorithm is adopted, with a window size of 14 days (i.e., each iteration is based on the standardized time series data of the most recent 14 days for weight correction), an iteration step size of 1 day (i.e., dynamic weight is updated once a day), and the window sliding method is forward scrolling. After each slide, the earliest day's data is removed and the latest day's data is added to ensure that the weight correction can reflect the recent trend of syndrome evolution.

[0135] Taking patient P1 as an example, dynamic weighting iterative calculation is performed using standardized time-series data from the previous 15 days. The specific process is as follows:

[0136] Initial weight settings: Dynamic weights for day 1 =Main and Concurrent Support Basic Weight Matrix ;

[0137] Day 2 weighting calculation:

[0138] First, calculate ,in This is the standardized data matrix for day 1. Standardized data matrix for day 2:

[0139] (Standardized monitoring data corresponding to 6 syndromes in sequence);

[0140]

[0141] ;

[0142] Then, substitute the values ​​into the weight correction formula to calculate. :

[0143] Calculate each element: ;

[0144] Right now ;

[0145] Weight calculation for days 3-15: Following the same method described above, calculate the dynamic weights for days 3 through 15 sequentially. The dynamic weight W_15 for day 15 (where the sliding window covers data from days 2-15) is calculated as follows:

[0146]

[0147]

[0148] ;

[0149] (Dynamic weighting on day 14);

[0150] ;

[0151] Calculate each element:

[0152] 0.528×1.02=0.538, 0.291×1.02=0.297, 0.168×1.005=0.169, 0.551×1.01=0.557, 0.167×1.005=0.168, 0.301×1.005=0.303;

[0153] Right now

[0154] Based on the daily calculated dynamic weights, a time-series weight change curve for the primary and secondary symptoms is generated. Taking patient P1's Qi and Yin deficiency syndrome (A1) and blood stasis syndrome (B1) as an example, the weight change curve data for the first 15 days is shown in the table below. The curves clearly show the time-series drift characteristics of the weights: the weight of Qi and Yin deficiency syndrome decreases slightly from days 6 to 10 and gradually increases from days 11 to 15; the weight of blood stasis syndrome shows a slow upward trend overall, which is consistent with the actual evolution of the patient's syndrome.

[0155] Number of monitoring days Qi and Yin Deficiency Syndrome (A1) Weight Blood stasis syndrome (B1) weight 1 0.539 0.539 2 0.531 0.547 3 0.535 0.543 4 0.533 0.545 5 0.531 0.549 6 0.525 0.553 7 0.523 0.555 8 0.521 0.557 9 0.524 0.555 10 0.526 0.553 11 0.528 0.551 12 0.53 0.553 13 0.532 0.555 14 0.528 0.551 15 0.538 0.557

[0156] (4) Data lag correction:

[0157] There is a natural time lag between the fluctuations of Western medical physicochemical indicators (blood glucose, glycated hemoglobin) and the internal evolution of TCM syndromes. That is, after the TCM syndromes change, the Western medical physicochemical indicators will show significant fluctuations. Existing technology directly correlates the data of the two synchronously, which will lead to analysis distortion. This step introduces a time lag compensation factor to calculate the time difference between Western medical indicators and TCM syndromes, establishes a lag correction model, corrects the standardized time series data, and eliminates the distortion problem caused by time misalignment.

[0158] The specific implementation process is as follows:

[0159] The Pearson correlation coefficient method was used to calculate the time-series correlation between Western medicine physicochemical indicators and TCM syndrome quantitative indicators, and to determine the time lag difference between the two. Then, the time lag compensation factor is calculated. The specific formula is as follows:

[0160] Pearson correlation coefficient Calculation:

[0161] in, For the first Standardized values ​​of TCM syndrome quantification indicators for the day For the first Standardized values ​​of Western medical physicochemical indicators for the day. The mean of the quantitative indicators of TCM syndromes. This represents the average of Western medical physicochemical indicators. Time lag difference ( Tests were conducted sequentially over days 1, 2, ..., 7, and the day with the highest correlation coefficient was selected. As the optimal time lag difference);

[0162] Time lag compensation factor Calculation: ,in The absolute value of the correlation coefficient. The value range of is [0,1]. The closer the correlation coefficient is to 1, the stronger the correlation between the two, the smaller the lag compensation factor, and the smaller the correction magnitude; conversely, the closer the correlation coefficient is to 1, the larger the correction magnitude.

[0163] Taking the quantitative indicators of Qi and Yin deficiency syndrome and blood glucose level of patient P1 as an example, the time lag compensation factor is calculated:

[0164] Standardized values ​​of quantitative indicators for Qi and Yin deficiency syndrome were selected from the previous 30 days. and standardized blood glucose values Test in sequence =Correlation coefficient from 1 to 7 days :

[0165] when When the time is 2 days, the correlation coefficient maximum, =0.82, the calculation process is as follows (data from the first 10 days):

[0166] (Qi and Yin Deficiency Syndrome): -0.6, -0.9, -0.3, -0.7, -0.6, 1.8, 1.5, 1.2, 0.9, 0.6;

[0167] (blood sugar, =2): -0.3, -0.7, -0.6, 1.8, 1.5, 1.2, 0.9, 0.6, 0.3, 0.0;

[0168] ≈0.12, ≈0.21;

[0169] molecular ;

[0170] denominator ;

[0171] =8.76 / 10.74≈0.82;

[0172] Calculate the time lag compensation factor : ;

[0173] Similarly, calculate the time lag compensation factor for other syndromes and Western medical physicochemical indicators, including the relationship between damp-heat syndrome and blood glucose. =3 days =0.78, =0.22, the syndrome of internal accumulation of turbid toxins and glycated hemoglobin =5 days =0.75, =0.25; The lag compensation factor for concurrent symptoms and Western medicine indicators is calculated similarly, and finally the time-series lag compensation factor matrix corresponding to all symptoms is obtained. (The remaining compensatory factors are calculated based on clinical data.)

[0174] In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0175] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0176] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes, characterized in that, Includes the following steps: (1) Construction of basic weights for primary and secondary syndromes: Based on the clinical syndrome database of diabetes, the basic types of common primary syndromes and secondary syndromes of diabetes are divided, quantitative indicators corresponding to each syndrome are extracted, and the initial basic weights of each syndrome are determined by the analytic hierarchy process, and a basic weight matrix of primary and secondary syndromes is established. (2) Time series data acquisition and preprocessing: Time series monitoring data of diabetic patients are accessed. The time series monitoring data includes quantitative scores of TCM symptoms, quantitative parameters of tongue and pulse, and physicochemical indicators of Western medicine. Outliers in the data are removed using the 3σ criterion. The data after removing outliers is standardized by the Z-score standardization method to obtain standardized time series data. Among them, the physicochemical indicators of Western medicine include at least blood glucose and glycated hemoglobin indicators. (3) Time series weight drift correction: Based on the standardized time series data, a time series weight drift correction model is constructed. With time series as input, a sliding window iterative algorithm is used to iteratively update the dynamic weights of the main certificate and the secondary certificate in real time, and generate the time series weight change curves of the main certificate and the secondary certificate. (4) Data lag correction: A time lag compensation factor is introduced. The time lag compensation factor is determined by the time correlation analysis between Western medicine physicochemical indicators and TCM syndrome quantitative indicators. A lag correction model is established based on the time lag compensation factor to correct the time misalignment of the standardized time series data and eliminate data correlation distortion. (5) Gradual grading monitoring: Based on the time-series weight change curve of the main and concurrent symptoms, at least three gradual grading thresholds for the deterioration of symptoms are set. By calculating the slope of the time-series weight change of the main and concurrent symptoms and the amplitude of weight fluctuation, early minor abnormalities in the gradual grading process of symptoms are identified, and the grading determination of the evolution state of symptoms is completed. (6) Concurrent syndrome interference removal and precise early warning output: Set a concurrent syndrome weight threshold, remove concurrent syndrome weights below the concurrent syndrome weight threshold, retain the core weight of the main syndrome and the weight of high-impact concurrent syndromes, lock the deterioration trend of the main syndrome, and output precise early warning information by grade and syndrome based on the graded judgment results of the syndrome evolution state, and simultaneously output TCM intervention tendency prompts corresponding to the early warning level and syndrome type.

2. The method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes according to claim 1, characterized in that, The primary syndrome types mentioned in step (1) include Qi and Yin deficiency syndrome, damp-heat syndrome, and turbid toxin accumulation syndrome. The secondary syndrome types include blood stasis syndrome, phlegm-dampness syndrome, and liver stagnation syndrome. The quantitative indicators corresponding to each syndrome include the quantitative score of TCM symptoms and the characteristic values ​​corresponding to the quantitative parameters of tongue and pulse.

3. The method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes according to claim 1, characterized in that, The sliding window iterative algorithm described in step (3) has a window size of 7-30 days and an iteration step size of 1 day. Each iteration is based on the standardized time series data in the current window to correct the dynamic weights of the main certificate and the secondary certificate.

4. The method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes as described in claim 1, characterized in that, The time-series correlation analysis described in step (4) uses the Pearson correlation coefficient method to calculate the correlation coefficient between Western medicine physicochemical indicators and TCM syndrome quantitative indicators. Based on the correlation coefficient, the time difference between the evolution of Western medicine physicochemical indicators and TCM syndrome is determined, and then the value of the time-series lag compensation factor is determined.

5. The method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes according to claim 1, characterized in that, The gradual symptom deterioration classification thresholds mentioned in step (5) include mild warning threshold, moderate warning threshold and severe warning threshold. The slope of the weight change corresponding to the mild warning threshold is 0.02-0.05 / day, the slope of the weight change corresponding to the moderate warning threshold is 0.05-0.08 / day, and the slope of the weight change corresponding to the severe warning threshold is ≥0.08 / day.

6. The method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes according to claim 1, characterized in that, The threshold for the concurrent certification weight in step (6) is set to 0.1-0.

2. Concurrent certification weights below the threshold are judged as redundant interference weights and are removed.

7. The method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes according to claim 1, characterized in that, The tongue pulse quantification parameters mentioned in step (2) include tongue color quantification value, tongue coating thickness quantification value, pulse frequency, and pulse amplitude. The TCM symptom quantification score is based on the TCM syndrome diagnosis standard, and each symptom is assigned a value according to its severity.

8. The method for dynamic monitoring and early warning analysis of TCM syndromes in diabetes according to claim 1, characterized in that, The TCM intervention tendency prompts mentioned in step (6) include dietary conditioning suggestions, TCM conditioning directions, and emotional regulation suggestions, and the intervention tendency prompts correspond to the current main symptom type and warning level.

Images