Method for assessing risk of diabetic complications by fusing metabolomics and clinical indicators

Abstract

The present application relates to the technical field of health risk assessment, in particular to a diabetes complication risk assessment method fusing metabolomics and clinical indicators, which comprises obtaining records of blood glucose, insulin, fatty acid chain length, etc., constructing a time series, identifying an insulin trend reversal point, screening a sugar metabolism trend linkage section, marking a fatty acid chain length transfer area, and classifying to generate a diabetes complication risk grade result. The present application integrates multiple types of metabolic indicators through a continuous time series, establishes a synchronous correlation of blood glucose, insulin and fatty acid chain length, depicts metabolic change rhythm with the aid of a trend reversal node, forms a linkage change range in combination with a multi-index consistent section, and introduces a chain length structure transfer feature to subdivide and classify the sugar-fat relationship, thereby distinguishing risk grades from the aspects of time coverage and linkage characteristics, and improving the adaptability and distinguishing ability of complication risk assessment in dynamic change scenarios.

Description

Technical Field

[0001] This invention relates to the field of health risk assessment technology, and in particular to a method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators. Background Technology

[0002] The field of health risk assessment technology encompasses multi-dimensional data, including individual bioinformatics, clinical testing data, and behavioral and environmental factors, to determine and predict the risk of chronic diseases, acute events, or other health problems. Its core content involves integrating medical and health-related data from various sources, utilizing statistical models, bioinformatics analysis, and algorithmic logic to construct a risk assessment system. This system quantitatively assesses the likelihood of an individual or population developing a particular disease or health problem within a specific timeframe. Overall, health risk assessment technology is widely applied in disease prediction, personalized treatment recommendations, and public health interventions. It covers key aspects such as data collection, indicator standardization, variable selection, risk scoring system construction, model training, and validation, emphasizing a comprehensive approach that combines data-driven methods with medical evidence.

[0003] The method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators refers to constructing a computational model for the probability of diabetic patients developing specific complications by combining quantitative detection results of metabolites in samples such as blood or urine with the patient's existing clinical diagnostic data. The main technical aspects involved include the collection and standardization of metabolomics data, screening of key metabolites and modeling the interaction between key metabolites and clinical indicators, and the construction of a diabetic complication risk scoring system based on statistical discriminant analysis. This is achieved by collecting metabolite content data through a high-throughput mass spectrometry platform, combining it with clinical indicators such as fasting blood glucose, glycated hemoglobin, and insulin levels, using logistic regression models and principal component analysis to perform feature reduction and risk factor identification on the samples, and using linear discriminant analysis to establish a complication risk assessment model to quantify individual risk.

[0004] Existing technologies rely on single-test data and statistical modeling methods, lacking a structured analysis mechanism for the evolution of indicators across time periods. In scenarios where metabolic information is unevenly distributed over time or records are incomplete, it is difficult to unify the expression of various types of data in the time dimension. This makes it difficult for different types of indicators to synergistically explain the trend of risk changes. For example, when there is a structural shift in the distribution of fatty acid chain lengths, it cannot be effectively matched with blood glucose trends, thereby reducing the sensitivity and continuity of overall risk assessment and affecting the true presentation of the process of complication risk changes. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators. The technical solution is as follows: A method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators includes the following steps: S1: Acquire continuous blood glucose, insulin concentration, fatty acid chain length, fasting blood glucose, and insulin resistance data of diabetic patients, arrange them in chronological order, mark missing time points to be filled in, organize them into a data set arranged in chronological order, and generate a clinical synchronous record sequence of diabetes metabolism. S2: Call the insulin concentration change content in the synchronous clinical record sequence of diabetes metabolism, determine the change direction of adjacent records in turn, mark the position after continuous increase to decrease or after continuous decrease to increase, check the integrity of the records before and after the marked position, and form a list of insulin direction mutation time. S3: Based on the insulin direction mutation time list, extract the indicator information of fasting blood glucose, insulin resistance, and glycated hemoglobin within the corresponding time period, screen the segments where the change direction of each indicator information is consistent and the duration is continuous, and generate a set of glucose metabolism trend linkage segments. S4: Call the fatty acid data corresponding to the time period of the sugar metabolism trend linkage segment set, classify the direction of chain length composition change, mark the structural offset region, exclude the discontinuous change segment, and collect the continuous part to form a sugar-lipid linkage change region label set.

[0006] As a further aspect of the present invention, the clinical synchronous recording sequence of diabetes metabolism includes blood glucose fluctuation nodes, insulin change trajectory, fatty acid chain length distribution characteristics, and time series synchronization identifiers; the insulin direction mutation time list includes mutation time points, change direction labels, record integrity markers, and continuity verification results; the glucose metabolism trend linkage segment set includes trend consistency segments, linkage duration range, indicator synergistic change patterns, and interruption removal markers; and the glucose-lipid linkage change region marker set includes chain length percentage change type, change segment time coverage range, fatty acid composition transfer trend, and continuous change markers.

[0007] As a further aspect of the present invention, the step of obtaining S1 is as follows: S101: Acquire continuous blood glucose records, insulin concentration change records, fatty acid chain length composition records, fasting blood glucose records, and insulin resistance records of diabetic patients during the follow-up phase, and perform synchronous retrieval based on the timestamps attached to each record. The records with the same timestamps are horizontally aggregated in the data frame structure, and the timestamp sequence in the aggregated data frame is called. All aggregated data frames are vertically sorted according to the chronological order to generate time-series arranged data frames. S102: Based on the time series data frame, for the missing time points in the continuous timestamp distribution interval, establish a complete target time series interval, call the sorted timestamps to traverse point by point, match each target time point with the sorting result, establish a corresponding missing flag bit for the unmatched time points, and insert the missing flag into the position of the corresponding time point in the data frame to generate a time series record set with missing flags. S103: Call the time series record set with missing identifiers, and vectorize and merge the various values ​​at the same time point according to the continuous blood glucose value, insulin concentration value, fatty acid chain length composition value, fasting blood glucose value, and insulin resistance value corresponding to each time point, and construct the corresponding synchronization status record vector sequence. Combine the synchronization vectors of all time points to generate the clinical synchronization record sequence of diabetes metabolism.

[0008] As a further aspect of the present invention, the step of obtaining S2 is as follows: S201: Call the insulin concentration change content in the synchronous clinical record sequence of diabetes metabolism, extract the insulin concentration values ​​corresponding to each consecutive time point in sequence, determine the direction of adjacent value pairs according to the time order, and assign a positive or negative direction label to each pair of values ​​by judging the increase or decrease trend of the values ​​of two adjacent time points, so as to obtain the insulin trend direction sequence. S202: Based on the insulin trend direction sequence, call each position in the direction identifier sequence, identify the turning point where the direction changes after the previous direction is continuous according to the continuity characteristics of the adjacent directions, extract the index of each turning point, and then map the index to the corresponding time point on the original time axis to obtain the insulin trend change time set. S203: Based on the set of insulin trend changes, perform a data integrity search on the insulin concentration values ​​at each time point in the synchronous clinical record sequence of diabetes metabolism, excluding time points with missing data on either side, and retaining only the content of time points with continuous values, to obtain a list of insulin direction mutation times.

[0009] As a further aspect of the present invention, the step of obtaining S3 is as follows: S301: Based on the insulin direction mutation time list, extract the continuous observation segment corresponding to each time point, and sequentially call the fasting blood glucose record, insulin resistance record, and glycated hemoglobin change trajectory within the segment. Arrange the three records in chronological order. For the three records at the same time point, determine their change trend direction numerically and label them with a unified trend direction type to obtain the trend direction sequence of the three indicators. S302: Call the trend direction sequence of the three indicators, identify the set of time points with the same trend direction, arrange multiple consecutive time points with a unified trend into a single segment, and search whether there is a deviation in trend direction between adjacent time points. If there is a deviation, end the collection process of the current segment to obtain a set of continuous time periods with consistent trend. S303: Based on the set of continuous time periods with consistent trends, retrieve the original indicator data again for the start and end time points included in each time period, exclude non-continuous time points caused by missing data, classify and extract the continuous time point sequences, and obtain the set of sugar metabolism trend linkage segments.

[0010] As a further aspect of the present invention, the step of obtaining S4 is as follows: S401: Based on the time range covered by the set of linked segments of sugar metabolism trend, call the fatty acid chain length composition record, retrieve the corresponding chain length composition data frame according to the start and end time of the segment, extract the medium chain proportion and long chain proportion sequence, perform a comparison judgment on the direction of proportion change of adjacent time points in the segment, write the direction pointing into the segment index field, obtain the direction pointing entry corresponding to each segment, and obtain the chain length composition direction segment index set. S402: Based on the chain length composition direction segment index set, perform migration determination on the direction of the proportion of medium chain and the proportion of long chain in each segment, identify the segment where the proportion of medium chain shifts to the proportion of long chain or the proportion of long chain shifts to the proportion of medium chain, write the segment index and migration type into the migration mapping table, and associate the migration mapping table with the segment time boundary to obtain the chain length migration segment mapping table. S403: For the chain length migration segment mapping table, retrieve the continuity status of the time axis of each segment, check the connection relationship between the interval of adjacent time points and the segment boundary, filter out segments with time jumps or direction breakage, retain segments with continuous time axis and consistent migration direction, write the retained segment index and migration type into the region index table, and obtain the glycolipid linkage change region marker set.

[0011] As a further aspect of the present invention, the method further includes: S5: Call the set of markers for the glucose-lipid linkage change regions, classify and organize the linkage frequency, duration and blood glucose direction of each region, classify the linkage feature segments of different combinations into levels, output the corresponding level according to the intensity of the performance, and form the result of the risk level of diabetic complications. The risk level results for diabetes complications include risk level category, regional characteristics, duration distribution, and blood glucose trend consistency score.

[0012] As a further aspect of the present invention, the step of obtaining S5 is as follows: S501: Call the set of regions marked by the linkage between glucose and lipid changes, extract the time range and sequence index of each region, and sequentially retrieve the number of times the region appears, the start and end time period covered, and the corresponding blood glucose direction identification information. Add the three features to the region index table in sequence, and classify them by comparing whether the blood glucose direction and the glucose and lipid direction are consistent, and obtain the glucose and lipid direction classification index table. S502: Based on the glycolipid orientation classification index table, according to the consistent pointing group index, the time-adjacent areas in each group are sequentially connected. The adjacent areas in each group are merged sequentially through time continuity, retaining the area segments that are immediately before and after the time, and outputting the merged time segment corresponding to the original group identifier to obtain the glycolipid linkage merged segment set. S503: For the aforementioned glucose and lipid linkage merged segment set, perform three feature comparison processing on the blood glucose direction value, time coverage range, and glucose and lipid direction identification content of each segment. According to the comparison results, find the corresponding classification level in the classification rule template, write the level name into the corresponding segment entry field, and obtain the result of the risk level of diabetic complications.

[0013] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, multiple metabolic indicators are organized through continuous time series to construct a synchronous relationship framework for blood glucose, insulin, and fatty acid chain length. Trend reversal nodes are used to characterize the rhythm of metabolic state changes. Combined with screening of segments with consistent directions of multiple indicators, a stable range of linked changes is formed. Chain length structure transfer features are introduced to further subdivide and classify the relationship between glucose and lipids, supporting the differentiation of risk levels according to time coverage and linkage features, and enhancing the adaptability and discriminativeness of complication risk assessment in dynamic changing scenarios. Attached Figure Description

[0014] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the acquisition process of S1 in this invention; Figure 3 This is a flowchart illustrating the acquisition process of S2 in this invention; Figure 4 This is a flowchart illustrating the acquisition process of S3 in this invention; Figure 5 This is a flowchart illustrating the acquisition process of S4 in this invention; Figure 6 This is a flowchart of the acquisition process for S5 of the present invention. Detailed Implementation

[0015] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0016] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0017] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0018] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0019] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0020] Please see Figure 1 This invention provides a technical solution: a method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators, comprising the following steps: S1: Acquire continuous blood glucose records, insulin concentration change records, fatty acid chain length composition records, fasting blood glucose records, and insulin resistance records during the follow-up phase of diabetic patients. Arrange the above records according to the collection time sequence, merge the corresponding contents at the same time point, mark and organize the missing time points, form a continuous time arrangement structure, and generate a clinical synchronous record sequence of diabetes metabolism. S2: Call the insulin concentration change content in the synchronous clinical record sequence of diabetes metabolism, judge the change direction between adjacent records segment by segment, mark the position of continuous rise and then fall or continuous fall and then rise, check the integrity of the records before and after the marked position, and sort out the time point when the change direction reverses to form a list of insulin direction mutation time. S3: Based on the list of insulin direction mutation times, call the fasting blood glucose records, insulin resistance records, and glycated hemoglobin change trajectory within the corresponding time range, make a consistency judgment on the change direction of each indicator within the time range, filter records with consistent direction and continuous duration, remove interrupted segments, organize them into continuous segments, and generate a set of glucose metabolism trend linkage segments. S4: Based on the time range covered by the set of linked segments of sugar metabolism trend, call the fatty acid chain length composition record, compare and judge the direction of chain length composition change in each time segment, mark the segments where the proportion of medium chain shifts to the proportion of long chain or the reverse shift, exclude the segments with discontinuous changes, sort out the continuous marking results, and form a set of labeled regions of sugar and lipid linkage change. S5: Call the set of markers for glucose and lipid-linked changes, classify the frequency of occurrence, continuous coverage, and consistency with the direction of blood glucose changes of each marker region, merge and organize regions with similar manifestations, classify them according to the combination of linked manifestations, output the corresponding risk attribution results, and generate the risk level results of diabetic complications.

[0021] The synchronized clinical records of diabetes metabolism include blood glucose fluctuation nodes, insulin change trajectories, fatty acid chain length distribution characteristics, and time series synchronization markers. The insulin direction mutation time list includes mutation time points, change direction labels, record integrity markers, and continuity verification results. The set of linked glucose metabolism trend segments includes trend consistency segments, linkage duration range, indicator synergistic change patterns, and interruption removal markers. The set of markers for linked glucose and lipid changes includes chain length percentage change types, change segment time coverage range, fatty acid composition shift trends, and continuous change markers. The results of diabetes complication risk levels include risk level categories, linked region performance characteristics, duration distribution, and blood glucose trend consistency scores.

[0022] Please see Figure 2 The steps to obtain S1 are as follows: S101: Acquire continuous blood glucose records, insulin concentration change records, fatty acid chain length composition records, fasting blood glucose records, and insulin resistance records of diabetic patients during the follow-up phase, and perform synchronous retrieval based on the timestamps attached to each record. The records with the same timestamps are horizontally aggregated in the data frame structure, and the timestamp sequence in the aggregated data frame is called. All aggregated data frames are vertically sorted according to the chronological order to generate time-series arranged data frames. In constructing the synchronous clinical recording sequence for diabetes metabolism, the data acquisition module was first initialized. It connected to the cloud server of the hospital's information system and the patient's personal wearable device via a dedicated medical network interface. Specific subjects were identified, and multi-threaded data acquisition was performed within a preset monitoring period. Parallel data were extracted from: subcutaneous interstitial fluid glucose levels uploaded by continuous glucose monitoring devices; venous blood insulin concentration data measured in the laboratory using chemiluminescence immunoassay; plasma fatty acid profiles containing carbon chain lengths from C8 to C22, measured using gas chromatography-mass spectrometry; daily morning fasting venous blood glucose records; and steady-state model insulin resistance index records. These five types of raw data were read into memory and standardized according to their respective sampling time fields, uniformly converting them into a standard timestamp format. Subsequently, horizontal aggregation logic is executed, selecting the continuous blood glucose record timestamps with the highest sampling frequency and strongest continuity as the primary key sequence. A time alignment tolerance threshold of 30 seconds is set. The remaining four data tables are traversed, and data rows whose timestamps fall within the primary key sequence deviation range are matched. For successfully matched records, their insulin concentration, fatty acid chain length composition, fasting blood glucose, and insulin resistance values ​​are written into the corresponding row of the data frame structure. For unmatched timestamps, the corresponding fields are temporarily stored as invalid values. After matching all data, a sorting algorithm is called to strictly sort the aggregated data frames in ascending order based on the numerical value of the timestamp sequence, generating time-series arranged data frames.

[0023] S102: Based on the time series arrangement of the data frame, for the missing time points in the continuous timestamp distribution interval, establish a complete target time series interval, call the sorted timestamps to traverse point by point, match each target time point with the sorting result, establish a corresponding missing flag bit for the unmatched time points, and insert the missing flag into the position of the corresponding time point in the data frame to generate a time series record set with missing flags. Based on time-series data frames, a rigorous timeline continuity review and repair process is performed to eliminate data interruptions that may occur due to packet loss in device communication or during sensor calibration, ensuring absolute equidistant continuity of the timeline. A standard sampling step size of 300 seconds is set. The traversal pointer is initialized and calculates the time difference between adjacent time points starting from the first position of the data frame. If the difference is strictly equal to the standard step size, continuity is determined; otherwise, a missing value is identified. For any missing interval, an interpolation completion mechanism is immediately triggered, inserting a new time point row at the missing position and setting the status flag of the newly inserted row to 1 to distinguish it from the original measured data. For data filling in the newly inserted row, linear trend filling logic is executed. Values ​​from valid times before and after the missing interval are read, and the slope of the connecting line is calculated. Theoretical values ​​are then used to fill in the missing values. If the missing interval spans a specific threshold (e.g., 30 minutes), only the time point is retained as a placeholder, and the numerical field is left blank to avoid misleading errors. This process ensures that the data frame has no physical breakpoints in the time dimension, generating a time-series record set with missing value indicators.

[0024] S103: Call the time series record set with missing markers, and vectorize and merge the various values ​​at the same time point according to the continuous blood glucose value, insulin concentration value, fatty acid chain length composition value, fasting blood glucose value, and insulin resistance value corresponding to each time point, and construct the corresponding synchronous status record vector sequence. Combine the synchronous vectors of all time points to generate the clinical synchronous record sequence of diabetes metabolism. The system calls upon a time-series record set with missing data markers, performs vectorization encapsulation of multidimensional data, and scans the record set row by row to extract continuous blood glucose values, insulin concentration values, fatty acid chain length vectors, fasting blood glucose values, and insulin resistance values ​​at any given time point. It checks the missing data markers in the current row; if the marker indicates that the data is a true measurement, the five types of values ​​are concatenated in a preset order to construct a synchronized state record vector containing full-dimensional features; if the marker indicates that key data is missing, it is marked as an invalid vector. All constructed valid synchronized state record vectors are stacked vertically in chronological order to build a high-dimensional matrix structure in memory, ensuring that each row strictly corresponds to a physical time point and each column strictly corresponds to a type of physiological indicator. This achieves spatiotemporal unification of multi-source heterogeneous clinical data, generating a synchronized clinical record sequence for diabetes metabolism.

[0025] Please see Figure 3 The steps to obtain S2 are as follows: S201: Call the insulin concentration change content in the synchronous clinical record sequence of diabetes metabolism, extract the insulin concentration values ​​corresponding to each consecutive time point in sequence, determine the direction of adjacent value pairs according to the time order, and assign a positive or negative direction label to each pair of values ​​by judging the increase or decrease trend of the values ​​of two adjacent time points, so as to obtain the insulin trend direction sequence. Insulin concentration feature columns were extracted from synchronous clinical records of diabetes metabolism and loaded as an independently processed sequence. A sliding window covering two adjacent time points was set to capture the direction of secretion fluctuations. The arithmetic difference between the value at the next time point and the value at the current time point was calculated point by point, and a fluctuation dead zone judgment logic was introduced, with a minimum significant change threshold set at 0.1 units. If the difference is greater than the positive threshold, insulin is determined to be in a positive increasing state and assigned a positive label; if the difference is less than the negative threshold, it is determined to be in a negative decreasing state and assigned a negative label; if the absolute value of the difference is within the threshold range, it is determined to be in a stationary state and assigned a zero label. This operation was performed sequentially on the entire sequence, transforming continuous numerical changes into a discrete stream of direction labels, resulting in an insulin trend direction sequence.

[0026] S202: Based on the insulin trend direction sequence, call each position in the direction identifier sequence, identify the turning point where the direction changes after the previous direction is continuous according to the continuity characteristics of the adjacent directions, extract the index of each turning point, and then map the index to the corresponding time point on the original time axis to obtain the insulin trend change time set. Based on the insulin trend direction sequence, a turning point feature recognition algorithm is executed, traversing the direction marker sequence and focusing on monitoring critical positions where the direction sign reverses. The recognition logic is set so that a strong turning point is determined if and only if the direction marker at the current position is inconsistent with the direction marker at the next position, and both are non-zero. The algorithm searches for peaks where the direction changes from positive to negative and troughs where the direction changes from negative to positive, extracts the array index values ​​of these turning point positions, and uses these indices to map back to the original timeline sequence to accurately pinpoint the physical time point of the turning point. The entire sequence is scanned, and all time points that meet the turning point conditions are extracted and summarized to obtain the insulin trend change time set.

[0027] S203: Based on the set of insulin trend changes, perform a data integrity search on the insulin concentration values ​​of each time point in the synchronous clinical record sequence of diabetes metabolism, excluding time points with missing data on either side, and retaining only the content of time points with continuous values ​​to obtain a list of insulin direction mutation times. Based on the time set of insulin trend changes, artifacts caused by data interpolation or missing data collection are eliminated. For each time point in the set, the original record set with missing data is back-queried. The row containing the turning point and the record rows of the adjacent time points are located, and the missing data bits of these three consecutive time points are checked. Only when all three show true measurement data is the turning point considered a physiologically significant mutation. If any neighboring point shows interpolated data, it is determined that the numerical change curve at that point is generated by algorithm simulation and the turning point feature is unreliable, and the time point is immediately removed from the set. Through data integrity verification, false mathematically fitted turning points are filtered out, resulting in a list of insulin direction mutation times.

[0028] Please see Figure 4 The steps to obtain S3 are as follows: S301: Based on the list of insulin direction mutation times, extract the continuous observation segment corresponding to each time point, and sequentially call the fasting blood glucose record, insulin resistance record, and glycated hemoglobin change trajectory within the segment. Arrange the three records in chronological order. For the three records at the same time point, determine their trend direction in terms of numerical values ​​and label them with a unified trend direction type to obtain the trend direction sequence of the three indicators. Based on the insulin direction mutation time list, multidimensional correlation analysis was performed on each confirmed valid mutation point. A certain observation window was extracted before and after the mutation point, and fasting blood glucose records, insulin resistance records, and the trajectory or short-term volatility of glycated hemoglobin (HbA1c), reflecting long-term blood glucose control, were extracted within the window. Linear regression slopes were calculated for these three indicators within the observation window to quantify local trends. If the slope of an indicator exceeded a preset positive threshold, it was marked as rising; if it was below a negative threshold, it was marked as falling; otherwise, it was marked as stable. The trend directions of the three indicators were summarized to generate a vector composed of three directional components for the mutation time point, resulting in a trend direction sequence for the three indicators.

[0029] S302: Call the trend direction sequence of three indicators, identify the set of time points with the same trend direction, arrange multiple consecutive time points with a unified trend into a single segment, and check whether there is a deviation in trend direction between adjacent time points. If there is a deviation, the collection process of the current segment ends, and a set of continuous time periods with consistent trend is obtained. The algorithm calls upon the trend direction sequences of three indicators, performs full consistency filtering and merging logic, scans point by point and retains only time points where the three indicators are completely consistent, and removes atypical state points with divergent directions. Among the retained time points, it identifies sets of consecutive adjacent points on the time axis with unified trend directions and merges them into independent continuous time periods. During the merging process, it monitors in real time whether there are abrupt changes in direction between adjacent points. If a direction reversal or consistency break occurs in the continuous sequence, the collection of the current segment is immediately truncated, and the collected part is encapsulated into a trend consistency unit, ultimately obtaining a set of trend-consistent continuous time periods.

[0030] S303: Based on the set of continuous time periods with consistent trends, retrieve the original indicator data again for the start and end time points included in each time period, exclude non-continuous time points caused by missing data, classify and extract the continuous time point sequences, and obtain the set of sugar metabolism trend linkage segments. For a set of continuous time periods with consistent trends, a secondary physical constraint verification is performed on the raw data covered by each time period. The raw indicator values ​​at all time points within the time period are retrieved, and abnormal data exceeding human physiological limits are screened out. If a time period contains extremely low or extremely high abnormal sensor readings, the data at that moment is deemed invalid, and the original time period is cut off at the invalidation point, splitting into unconnected sub-segments. The remaining valid continuous time periods after cleaning and segmentation are reorganized, removing fragments that are insufficient in length due to segmentation, and segments with high data reliability and multidimensional trend consistency are extracted to obtain a set of glucose metabolism trend-linked segments.

[0031] Please see Figure 5 The steps to obtain S4 are as follows: S401: Based on the time range covered by the linkage segment set of sugar metabolism trend, call the fatty acid chain length composition record, retrieve the corresponding chain length composition data frame according to the start and end time of the segment, extract the medium chain proportion and long chain proportion sequence, perform a comparison judgment on the direction of proportion change of adjacent time points within the segment, write the direction pointing into the segment index field, obtain the direction pointing entry corresponding to each segment, and obtain the chain length composition direction segment index set. Based on the set of linked segments related to glucose metabolism trends, the time range of each segment is locked, and lipid metabolism characteristics are analyzed by calling the fatty acid chain length composition records. The fatty acid profile is summarized into two aggregate indicators: the proportion of medium-chain fatty acids and the proportion of long-chain fatty acids. The proportion values ​​and net changes at the beginning and end time points of each segment are calculated. If the proportion of medium-chain fatty acids at the end of the segment is significantly increased compared to the beginning, it is recorded as a medium-chain increase; if the proportion of long-chain fatty acids is significantly decreased, it is recorded as a long-chain decrease. The direction of change in these two dimensions is written into the segment index attribute to obtain the chain length composition direction segment index set.

[0032] S402: Based on the chain length composition direction segment index set, the migration direction of the proportion of medium chain and the proportion of long chain in each segment is determined, and the segment of medium chain proportion turning into long chain proportion or long chain proportion turning into medium chain proportion is identified. The segment index and migration type are written into the migration mapping table, and the migration mapping table is associated with the segment time boundary to obtain the chain length migration segment mapping table. Based on the chain length composition direction segment index set, a fatty acid carbon chain migration mode determination logic is executed, focusing on identifying two specific metabolic modes: the first is long-chain to medium-chain migration, where the determination condition is that the proportion of long chains decreases, the proportion of medium chains increases, and the ratio of the sum and difference of the absolute values ​​of the changes in both is within a reasonable range to ensure that it becomes the main pathway; the second is medium-chain to long-chain migration, where the determination condition is that the proportion of medium chains decreases and the proportion of long chains increases. All segments are traversed, and segments that conform to the above modes are labeled with corresponding type tags. These tags are then bound and mapped to time boundaries to obtain a chain length migration segment mapping table.

[0033] S403: For the chain length migration segment mapping table, retrieve the continuity status of the time axis of each segment, check the connection relationship between the interval of adjacent time points and the segment boundary, filter out segments with time jumps or direction breakage, retain segments with continuous time axis and consistent migration direction, write the retained segment index and migration type into the region index table, and obtain the glycolipid linkage change region label set. For the chain length migration segment mapping table, the connection relationship of each segment on the time axis is examined to construct a complete metabolic event, and the time interval and migration type of adjacent segments are analyzed. If two segments are temporally continuous or have a very short interval and have completely consistent migration types, they are identified as the same continuous lipid metabolism process and retained; if there are significant time jumps or changes in migration types between adjacent segments, a metabolic state break is identified. Isolated fragmented segments with very short durations are screened out, and only steady-state segments with continuous time axis and consistent migration directions are retained. The segment index and type are written into the region index table to obtain a set of labeled regions for glycolipid-lipid linkage changes.

[0034] Please see Figure 6 The steps to obtain S5 are as follows: S501: Call the set of regions marked by the linkage between glucose and lipid changes, extract the time range and index of each region, and sequentially retrieve the number of times the region appears, the start and end time period covered, and the corresponding blood glucose direction identification information. Add the three features to the region index table in sequence, and classify them by comparing whether the blood glucose direction and the glucose and lipid direction are consistent, and obtain the glucose and lipid direction classification index table. The system retrieves a set of region markers for glucose-lipid linkage changes, extracts the temporal coverage, frequency of recurrence within historical periods, and glucose directional change indicators for each region, and appends these features to the region index table. Classification is then performed, comparing the biological synergy between glucose directional change and glucose-lipid migration direction. Definitions are made such as lipid migration towards easily oxidized medium-chain compounds accompanied by a decrease in glucose as a benign compensatory mode, and lipid migration towards medium-chain compounds while glucose still rises as a pathological resistance mode. Based on these rules, each region is assigned to its corresponding classification group, resulting in a glucose-lipid directional classification index table.

[0035] S502: Based on the glycolipid orientation classification index table, the time-adjacent regions in each group are sequentially connected according to the consistent orientation group index. The adjacent regions in each group are merged sequentially through time continuity, and the regions that are immediately before and after the time are retained. The merged time period is output in correspondence with the original group identifier to obtain the glycolipid linkage merged segment set. Based on the glycolipid orientation classification index table, regions belonging to the same pattern group are merged along the time axis. Regions within a group are scanned sequentially. If the time interval between two regions is less than a specific threshold and there are no intervening outliers, they are determined to be fluctuations of the same macrometabolic disorder. A merging operation is then performed to connect the time ranges of the two regions, forming a new segment with a longer coverage period, and the frequency of occurrence is accumulated. This process is repeated until all mergeable regions within the group have been processed, resulting in a set of glycolipid-linked merged segments.

[0036] S503: For the glucose and lipid linkage merged segment set, the three features of blood glucose direction value, time coverage range and glucose and lipid direction identification content of each segment are compared. According to the comparison results, the corresponding level is found in the classification rule template, the level name is written into the corresponding segment entry field, and the result of the risk level of diabetic complications is obtained. For the glucose-lipid linkage merged segment set, a pre-defined risk grading rule template is invoked to extract three key features for each merged segment: blood glucose direction value, glucose-lipid direction identifier, and duration. The risk level is determined based on the combination of features. For example, if blood glucose rises, lipids show a transition from medium-chain to long-chain lipids, and the duration exceeds a set high-risk threshold, it is matched as a high-risk level, indicating a severe risk of lipotoxicity; if blood glucose rises but lipids show a transition from long-chain to medium-chain lipids with a shorter duration, it is matched as a low-risk level. All segments are traversed, and the calculated level names are written into the segment entries to generate the risk level results for diabetic complications.

[0037] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators, characterized in that, Includes the following steps: S1: Acquire continuous blood glucose, insulin concentration, fatty acid chain length, fasting blood glucose, and insulin resistance data of diabetic patients, arrange them in chronological order, mark missing time points to be filled in, organize them into a data set arranged in chronological order, and generate a clinical synchronous record sequence of diabetes metabolism. S2: Call the insulin concentration change content in the synchronous clinical record sequence of diabetes metabolism, determine the change direction of adjacent records in turn, mark the position after continuous increase to decrease or after continuous decrease to increase, check the integrity of the records before and after the marked position, and form a list of insulin direction mutation time. S3: Based on the insulin direction mutation time list, extract the indicator information of fasting blood glucose, insulin resistance, and glycated hemoglobin within the corresponding time period, screen the segments where the change direction of each indicator information is consistent and the duration is continuous, and generate a set of glucose metabolism trend linkage segments. S4: Call the fatty acid data corresponding to the time period of the sugar metabolism trend linkage segment set, classify the direction of chain length composition change, mark the structural offset region, exclude the discontinuous change segment, and collect the continuous part to form a sugar-lipid linkage change region label set.

2. The method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators according to claim 1, characterized in that: The synchronized clinical recording sequence of diabetes metabolism includes blood glucose fluctuation nodes, insulin change trajectory, fatty acid chain length distribution characteristics, and time series synchronization markers. The insulin direction mutation time list includes mutation time points, change direction labels, record integrity markers, and continuity verification results. The set of glucose metabolism trend linkage segments includes trend consistency segments, linkage duration range, indicator synergistic change patterns, and interruption removal markers. The set of glucose and lipid linkage change region markers includes chain length percentage change type, change segment time coverage range, fatty acid composition transfer trend, and continuous change markers.

3. The method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators according to claim 1, characterized in that, The steps for obtaining S1 are as follows: S101: Acquire continuous blood glucose records, insulin concentration change records, fatty acid chain length composition records, fasting blood glucose records, and insulin resistance records of diabetic patients during the follow-up phase, and perform synchronous retrieval based on the timestamps attached to each record. The records with the same timestamps are horizontally aggregated in the data frame structure, and the timestamp sequence in the aggregated data frame is called. All aggregated data frames are vertically sorted according to the chronological order to generate time-series arranged data frames. S102: Based on the time series data frame, for the missing time points in the continuous timestamp distribution interval, establish a complete target time series interval, call the sorted timestamps to traverse point by point, match each target time point with the sorting result, establish a corresponding missing flag bit for the unmatched time points, and insert the missing flag into the position of the corresponding time point in the data frame to generate a time series record set with missing flags. S103: Call the time series record set with missing identifiers, and vectorize and merge the various values ​​at the same time point according to the continuous blood glucose value, insulin concentration value, fatty acid chain length composition value, fasting blood glucose value, and insulin resistance value corresponding to each time point, and construct the corresponding synchronization status record vector sequence. Combine the synchronization vectors of all time points to generate the clinical synchronization record sequence of diabetes metabolism.

4. The method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators according to claim 1, characterized in that, The steps for obtaining S2 are as follows: S201: Call the insulin concentration change content in the synchronous clinical record sequence of diabetes metabolism, extract the insulin concentration values ​​corresponding to each consecutive time point in sequence, determine the direction of adjacent value pairs according to the time order, and assign a positive or negative direction label to each pair of values ​​by judging the increase or decrease trend of the values ​​of two adjacent time points, so as to obtain the insulin trend direction sequence. S202: Based on the insulin trend direction sequence, call each position in the direction identifier sequence, identify the turning point where the direction changes after the previous direction is continuous according to the continuity characteristics of the adjacent directions, extract the index of each turning point, and then map the index to the corresponding time point on the original time axis to obtain the insulin trend change time set. S203: Based on the set of insulin trend changes, perform a data integrity search on the insulin concentration values ​​at each time point in the synchronous clinical record sequence of diabetes metabolism, excluding time points with missing data on either side, and retaining only the content of time points with continuous values, to obtain a list of insulin direction mutation times.

5. The method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators according to claim 1, characterized in that, The steps for obtaining S3 are as follows: S301: Based on the insulin direction mutation time list, extract the continuous observation segment corresponding to each time point, and sequentially call the fasting blood glucose record, insulin resistance record, and glycated hemoglobin change trajectory within the segment. Arrange the three records in chronological order. For the three records at the same time point, determine their change trend direction numerically and label them with a unified trend direction type to obtain the trend direction sequence of the three indicators. S302: Call the trend direction sequence of the three indicators, identify the set of time points with the same trend direction, arrange multiple consecutive time points with a unified trend into a single segment, and search whether there is a deviation in trend direction between adjacent time points. If there is a deviation, end the collection process of the current segment to obtain a set of continuous time periods with consistent trend. S303: Based on the set of continuous time periods with consistent trends, retrieve the original indicator data again for the start and end time points included in each time period, exclude non-continuous time points caused by missing data, classify and extract the continuous time point sequences, and obtain the set of sugar metabolism trend linkage segments.

6. The method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators according to claim 1, characterized in that, The steps for obtaining S4 are as follows: S401: Based on the time range covered by the set of linked segments of sugar metabolism trend, call the fatty acid chain length composition record, retrieve the corresponding chain length composition data frame according to the start and end time of the segment, extract the medium chain proportion and long chain proportion sequence, perform a comparison judgment on the direction of proportion change of adjacent time points in the segment, write the direction pointing into the segment index field, obtain the direction pointing entry corresponding to each segment, and obtain the chain length composition direction segment index set. S402: Based on the chain length composition direction segment index set, perform migration determination on the direction of the proportion of medium chain and the proportion of long chain in each segment, identify the segment where the proportion of medium chain shifts to the proportion of long chain or the proportion of long chain shifts to the proportion of medium chain, write the segment index and migration type into the migration mapping table, and associate the migration mapping table with the segment time boundary to obtain the chain length migration segment mapping table. S403: For the chain length migration segment mapping table, retrieve the continuity status of the time axis of each segment, check the connection relationship between the interval of adjacent time points and the segment boundary, filter out segments with time jumps or direction breakage, retain segments with continuous time axis and consistent migration direction, write the retained segment index and migration type into the region index table, and obtain the glycolipid linkage change region marker set.

7. The method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators according to claim 1, characterized in that, The method further includes: S5: Call the set of markers for the glucose-lipid linkage change regions, classify and organize the linkage frequency, duration and blood glucose direction of each region, classify the linkage feature segments of different combinations into levels, output the corresponding level according to the intensity of the performance, and form the result of the risk level of diabetic complications. The risk level results for diabetes complications include risk level category, regional characteristics, duration distribution, and blood glucose trend consistency score.

8. The method for assessing the risk of diabetic complications by integrating metabolomics and clinical indicators according to claim 7, characterized in that, The steps for obtaining S5 are as follows: S501: Call the set of regions marked by the linkage between glucose and lipid changes, extract the time range and sequence index of each region, and sequentially retrieve the number of times the region appears, the start and end time period covered, and the corresponding blood glucose direction identification information. Add the three features to the region index table in sequence, and classify them by comparing whether the blood glucose direction and the glucose and lipid direction are consistent, and obtain the glucose and lipid direction classification index table. S502: Based on the glycolipid orientation classification index table, according to the consistent pointing group index, the time-adjacent areas in each group are sequentially connected. The adjacent areas in each group are merged sequentially through time continuity, retaining the area segments that are immediately before and after the time, and outputting the merged time segment corresponding to the original group identifier to obtain the glycolipid linkage merged segment set. S503: For the aforementioned glucose and lipid linkage merged segment set, perform three feature comparison processing on the blood glucose direction value, time coverage range, and glucose and lipid direction identification content of each segment. According to the comparison results, find the corresponding classification level in the classification rule template, write the level name into the corresponding segment entry field, and obtain the result of the risk level of diabetic complications.

Images