A method and system for managing health data of diabetic patients

Abstract

This invention discloses a method and system for managing health data of diabetic patients, relating to the field of data management technology. A health data management system for diabetic patients includes a health data monitoring module and a health data management module. This invention not only considers blood glucose fluctuations but also incorporates multi-dimensional signals such as molecular markers, sleep cycles, and bioelectrical feedback. It establishes a patient molecular physiological interaction index through mapping along a unified time axis and correcting a causal inference model. By capturing cross-channel synergistic anomalies, it effectively reduces misjudgments caused by occasional fluctuations. By constructing individual metabolic adaptation domains for each patient and metabolic adaptation domains for similar patients, and employing a dual-gating judgment mode, it avoids excessive warnings in health data management caused by a single threshold standard, ensuring personalized applicability based on individual patient differences.

Description

Technical Field

[0001] This invention relates to the field of data management technology, and in particular to a method and system for managing health data of diabetic patients. Background Technology

[0002] Existing methods for managing the health of diabetic patients mainly rely on single means such as blood glucose monitoring, medication management, and lifestyle interventions. Although these methods can monitor the health status of patients to some extent, they usually have the following shortcomings: First, they rely too much on blood glucose levels as the sole indicator and fail to fully consider the potential impact of other physiological signals on blood glucose fluctuations; second, most existing methods fail to consider the lag effect of physiological signals, resulting in a relatively slow judgment and early warning of the disease.

[0003] Therefore, there is a need for a diabetes health data management method based on multi-dimensional signal fusion and causal inference, which can significantly improve the accuracy of diabetes risk assessment and health management. Summary of the Invention

[0004] This invention aims to provide a method and system for managing health data of diabetic patients, which can not only improve the accuracy of health risk assessment, but also achieve personalized intervention, significantly improving the long-term management effect and quality of life of diabetic patients.

[0005] A method for managing health data of diabetic patients includes the following steps:

[0006] The patient's blood glucose fluctuations and molecular marker concentration fluctuations are acquired according to a set monitoring period and correlated to obtain the molecular regulation ratio parameter in the patient's body; the patient's sleep cycle fluctuation parameters and bioelectrical feedback fluctuation parameters are acquired in real time; and the patient's molecular physiological interaction index is constructed based on the molecular regulation ratio parameter, sleep cycle fluctuation parameter, and bioelectrical feedback fluctuation parameter.

[0007] Obtain the patient's raw parameter monitoring data; compare the raw parameter monitoring data with the patient's individual metabolic adaptation domain and the metabolic adaptation domain of similar patients to obtain abnormal results of the patient's health data; if the abnormal results of the patient's health data are abnormal, determine the abnormality of diabetes health based on the patient's molecular physiological interaction index and output the patient's diabetes health risk index; continue to implement the corresponding diabetes monitoring and management strategy based on the patient's diabetes health risk index; if the abnormal results of the patient's health data are normal, continue to manage the health data of the diabetic patient.

[0008] As a preferred embodiment of the present invention, the specific steps for constructing a patient molecular physiological interaction index based on molecular regulation ratio parameters, sleep cycle stability parameters, and bioelectrical feedback fluctuation parameters include:

[0009] Blood glucose fluctuations and molecular marker concentration fluctuations were correlated by time series to obtain several time segments; these time segments included pre- and post-meal segments, sleep cycle segments, pre- and post-meal segments, and pre- and post-medication segments; the amplitude of changes in blood glucose fluctuations and molecular marker concentration fluctuations in each time segment was extracted to obtain the corresponding molecular regulation ratio parameters.

[0010] The rhythmic consistency of sleep cycle fluctuation parameters was calculated to obtain the sleep time regulation factor; pattern recognition was performed on the bioelectric feedback fluctuation parameters to extract abnormal bioelectric fluctuation features under different neural states; molecular regulation ratio parameters, sleep time regulation factor and abnormal bioelectric fluctuation features were synchronously mapped on a unified time axis to obtain a multidimensional blood glucose-molecular interaction matrix.

[0011] Based on the degree of synchronization deviation of different dimensions of features in the multidimensional glucose-molecular interaction matrix, a patient molecular physiological interaction index is generated.

[0012] As a preferred embodiment of the present invention, the specific steps for determining the original parameter monitoring data and the individual metabolic adaptation domain of the patient and the metabolic adaptation domain of similar patients include:

[0013] Based on the patient's parameter monitoring data over a recent period, the data is stratified according to diurnal time phases and physiological states to obtain several time-state segments. Key event anchors and patient circadian rhythms are obtained based on the original parameter monitoring data. Using the key event anchors as boundaries, parameter stability intervals and parameter fluctuation profiles are calculated for each time-state segment, serving as the key segment adaptation domain. Data quality weights are calculated for each key segment adaptation domain. Each key segment adaptation domain is then sorted according to the patient's circadian rhythm and data quality weights to obtain time-series-based individual metabolic adaptation domains for each patient.

[0014] Based on a publicly available database of diabetic patient parameters; matching patients with their corresponding fingerprint features to output a behavioral-physiological twin patient group; and constructing a metabolic adaptation domain for similar patients based on the behavioral-physiological twin patient group.

[0015] Based on the original parameter monitoring data, the abnormal results of the patient's health data are obtained by judging the individual metabolic adaptation domain of the patient and the metabolic adaptation domain of similar patients.

[0016] As a preferred embodiment of the present invention, the specific steps for judging diabetic health abnormalities by analyzing the patient's molecular physiological interaction index include:

[0017] The fluctuation lag strength is identified based on the patient molecular physiological interaction index and causal inference model; the patient molecular physiological interaction index is corrected according to the fluctuation lag strength to obtain the corrected patient molecular physiological interaction index; the corrected patient molecular physiological interaction index is used to judge diabetic health abnormalities using the abnormal coupling effect judgment mechanism to obtain the patient's diabetic health risk index.

[0018] Specifically, the modified patient molecular physiological interaction index is divided into N types of modified diabetes abnormality indicators; the contribution of each type of modified diabetes abnormality indicator is calculated; if the N types of modified diabetes abnormality indicators are superimposed in the same time window and the contribution of each type exceeds the preset threshold, the patient's diabetes health risk indicator is determined to be a composite diabetes risk; if only some of the modified diabetes abnormality indicators are superimposed in the same time window or any one of the contribution is lower than the preset threshold, the patient's diabetes health risk indicator is determined to be a diabetes observation risk.

[0019] As a preferred embodiment of the present invention, the specific steps for continuing to implement the corresponding diabetes monitoring and management strategy based on the patient's diabetes health risk indicators include:

[0020] When patients are at multiple risk of diabetes, a high-frequency data collection mode is implemented to shorten the length and frequency of monitoring cycles.

[0021] If a patient is at risk of diabetes observation, the monitoring period should be shortened by at least two consecutive monitoring periods. If the patient is not at risk of diabetes complex or diabetes observation in two consecutive monitoring periods, the monitoring period should be restored to the set period. Otherwise, the patient should be subjected to high-frequency data collection mode.

[0022] As a preferred technical solution of the present invention, the abnormal coupling effect determination mechanism is set based on the causal inference algorithm.

[0023] A health data management system for diabetic patients, comprising:

[0024] The health data monitoring module includes a data monitoring unit. The data monitoring unit is used to acquire the patient's blood glucose fluctuations and molecular marker concentration fluctuations according to a set monitoring cycle, perform correlation calculations, and obtain the molecular regulation ratio parameters in the patient's body. It also acquires the patient's sleep cycle fluctuation parameters and bioelectrical feedback fluctuation parameters in real time. Based on the molecular regulation ratio parameters, sleep cycle fluctuation parameters, and bioelectrical feedback fluctuation parameters, it constructs the patient's molecular physiological interaction index.

[0025] The health data management module includes a diabetes anomaly judgment unit and a diabetes health judgment unit. The diabetes anomaly judgment unit acquires the patient's raw parameter monitoring data; it then judges the raw parameter monitoring data against the patient's individual metabolic adaptation domain and the metabolic adaptation domain of similar patients to obtain the patient's health data anomaly result. The diabetes health judgment unit, if the patient's health data anomaly result is abnormal, performs a diabetes health anomaly judgment on the patient's molecular physiological interaction index and outputs the patient's diabetes health risk index; based on the patient's diabetes health risk index, it continues to execute the corresponding diabetes monitoring and management strategy; if the patient's health data anomaly result is normal, it continues to manage the diabetes patient's health data.

[0026] The present invention has the following advantages:

[0027] 1. This invention not only considers blood glucose fluctuations but also incorporates multi-dimensional signals such as molecular markers, sleep cycles, and bioelectrical feedback. By mapping a unified time axis and correcting the causal inference model, a patient molecular physiological interaction index is established. Compared with single blood glucose monitoring, this method can capture cross-channel synergistic anomalies, effectively reducing misjudgments caused by occasional fluctuations. By constructing individual metabolic adaptation domains for patients and metabolic adaptation domains for similar patients, a dual-gating judgment mode is adopted. Anomaly judgment is only initiated when both the individual and group domains show a consistent and significant deviation, avoiding excessive warnings caused by a single threshold standard and ensuring personalized applicability under individual differences.

[0028] 2. This invention identifies the delayed effects of molecular markers, sleep rhythms, and neural activity on blood glucose by constructing a causal inference model and automatically adjusts the time window and intensity, overcoming the shortcomings of traditional real-time monitoring that ignores physiological lag, making risk identification more consistent with pathological processes and human response patterns; by dividing the modified interaction index into multiple types of modified abnormal indicators and calculating their contribution, it achieves hierarchical output; it can not only capture signals of potential serious complications in a timely manner, but also avoid the over-interpretation of mild abnormalities, thus improving the clinical value of health data management. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the structure of a health data management system for diabetic patients used in an embodiment of the present invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0031] Example 1: A method for managing health data of diabetic patients, comprising the following steps:

[0032] The patient's blood glucose fluctuations and molecular marker concentration fluctuations are acquired according to a set monitoring period and correlated to obtain the molecular regulation ratio parameter in the patient's body; the patient's sleep cycle fluctuation parameters and bioelectrical feedback fluctuation parameters are acquired in real time; and the patient's molecular physiological interaction index is constructed based on the molecular regulation ratio parameter, sleep cycle fluctuation parameter, and bioelectrical feedback fluctuation parameter.

[0033] Specifically, for diabetic patients requiring health data management, continuous blood glucose fluctuation data and multiple molecular marker concentration fluctuation data are acquired according to a set monitoring cycle. Blood glucose fluctuation refers to the dynamic changes in blood glucose levels over a period of time, including blood glucose values ​​at different time points such as fasting and postprandial, as well as the rate of rise, magnitude of fall, and differences between peaks and troughs, reflecting the body's immediate response to various factors such as diet, exercise, and medication. Molecular marker concentration fluctuation refers to the changes in the concentration of specific molecules (such as FXNIP, XIAP, ZBP1, etc.) in vivo over time that are related to glucose metabolism, oxidative stress, or inflammatory responses. Elevations or decreases in the levels of these molecules often precede changes in blood glucose and can reveal potential abnormalities in metabolic regulation or cell signaling.

[0034] The specific steps for constructing a patient's molecular physiological interaction index based on molecular regulation ratio parameters, sleep cycle stability parameters, and bioelectrical feedback fluctuation parameters include:

[0035] Blood glucose fluctuations and molecular marker concentration fluctuations were correlated by time series to obtain several time segments; these time segments included pre- and post-meal segments, sleep cycle segments, pre- and post-meal segments, and pre- and post-medication segments; the amplitude of changes in blood glucose fluctuations and molecular marker concentration fluctuations in each time segment was extracted to obtain the corresponding molecular regulation ratio parameters.

[0036] Prioritize cleaning, artifact removal, and time alignment of input blood glucose fluctuations and molecular marker concentration fluctuations to eliminate the impact of sampling frequency differences, packet loss, and outliers; establish anchor points according to the time stamps of events such as meals, falling asleep and waking up, and medication, and simultaneously unfold the two types of sequences along the same time axis, and perform soft alignment in combination with possible physiological lags, so that the two types of signals can present linkage characteristics in a comparable manner during key physiological stages;

[0037] The time-phase segments include at least pre- and post-meal segments, sleep cycle segments, and pre- and post-medication segments. The pre- and post-meal segments are centered on the meal time, with fixed or adaptive observation windows set forward and backward, covering the pre-meal baseline, post-meal rise, peak, and fall phases. The sleep cycle segments are based on the sleep-onset to wakefulness axis, subdivided into deep sleep, light sleep, and REM sleep stages, with brief nighttime awakenings separately labeled to avoid diluting rhythmic information. The pre- and post-medication segments are set with the drug administration time as the boundary, combining the onset and duration of drug action with observation windows, and are summarized separately under different dosage forms or dosage conditions to ensure consistency between dosage and pharmacodynamic environment within the segment. Each type of time-phase segment retains metadata such as event anchor points, duration, data quality scores, and alignment methods to provide a basis for subsequent calculations and verifications.

[0038] The amplitude and key kinetic elements of blood glucose fluctuations and molecular marker concentration fluctuations were extracted from each time segment, including the slope of rise and fall, the difference between peak and trough values, the timing of peak appearance and the rate of fall, etc., and paired with two types of features within the same time window. At the same time, in order to reduce the interference of non-physiological factors, the food segment was normalized according to carbohydrate intake, the medication segment according to actual dose and dosage form, and the sleep segment according to the stability of the nighttime rhythm. Finally, the paired features were compressed into molecular regulation ratio parameters that can reflect the strength and direction of molecular changes in blood glucose response.

[0039] The rhythmic consistency of sleep cycle fluctuation parameters was calculated to obtain the sleep time regulation factor; pattern recognition was performed on the bioelectric feedback fluctuation parameters to extract abnormal bioelectric fluctuation features under different neural states; molecular regulation ratio parameters, sleep time regulation factor and abnormal bioelectric fluctuation features were synchronously mapped on a unified time axis to obtain a multidimensional blood glucose-molecular interaction matrix.

[0040] The system assesses the consistency of a patient's circadian rhythm based on continuous sleep stage sequences throughout the night and over multiple days (including sleep onset and wake-up times, the proportion of deep sleep, light sleep, and REM sleep stages, and the number and duration of nighttime awakenings). Specifically, based on the patient's work attributes or other sleep characteristics, the system avoids misjudging shift work, cross-time zone, or weekend work-rest changes as abnormalities. The system first identifies the patient's long-term circadian rhythm and biological clock phase, then measures the recent deviation from this phase, the stability of stage transitions, and the intensity of nighttime awakenings' disturbance to the stage structure. The system also models and analyzes the differences between weekdays and rest days separately, and outputs sleep time regulation factors to describe the modulating effect of sleep rhythm on the strength of metabolic responses.

[0041] Simultaneously, pattern recognition was performed on the bioelectrical feedback fluctuation parameters. The data sources for these parameters included signals such as intercardiac interval variation, skin conductance level and transient response, peripheral skin temperature and respiratory rhythm. The signals were first corrected for motion artifacts and baseline drift, and then segmented according to time phase and scenario (resting, postprandial, exercise recovery, and nighttime sleep). In each time phase segment, characteristic patterns such as persistently increased sympathetic activity, slow recovery, sudden high arousal that did not match the sleep stage, and nocturnal autonomic imbalance were identified. These patterns were then compared with the patient's previous data in the same scenario to screen out abnormal fluctuations that occurred continuously or in clusters. The final output is a bioelectrical abnormality fluctuation feature, which includes the abnormality type, duration, frequency of occurrence, and confidence level, used to label the potential perturbation of metabolism by neural activity. After extracting three types of multidimensional information—molecular, sleep, and bioelectrical—the data is synchronously mapped using a unified time axis. Specifically, each channel is resampled to the same time grid and aligned using events such as eating, medication, falling asleep, waking up, and the start and end of exercise as anchor points. If there are missing intervals, conservative imputation is used and the weights are reduced. Abnormal fragments retain their source and quality labels, thereby constructing a multidimensional glucose-molecular interaction matrix. In the matrix, rows correspond to time steps, and columns correspond to features such as molecular regulation ratios, sleep time regulation factors, and bioelectrical abnormality fluctuations. The matrix cells simultaneously store alignment methods, event labels, and data confidence levels to ensure the interpretability of subsequent judgments.

[0042] Based on the degree of synchronization deviation of different dimensions of features in the multidimensional glucose-molecular interaction matrix, a patient molecular physiological interaction index is generated. During the generation process, cross-channel co-directional or complementary co-directional deviations are prioritized, isolated abnormalities in single channels are filtered out, and a hysteresis threshold is introduced to avoid alarm jitter. Co-directional deviations with longer duration, larger amplitude, and more consistent evidence are given higher contributions. Signals from low-quality fragments are automatically downweighted. The output of the patient molecular physiological interaction index includes an overall level and a description of the contribution of each channel, along with a confidence range and a summary of the triggering event anchor. It can be directly used for the triggering and grading of abnormal coupling effects, and can also serve as a basis for adjusting the evaluation cycle and intervention timing in subsequent closed-loop optimization.

[0043] Obtain the patient's raw parameter monitoring data; compare the raw parameter monitoring data with the patient's individual metabolic adaptation domain and the metabolic adaptation domain of similar patients to obtain abnormal results of the patient's health data; if the abnormal results of the patient's health data are abnormal, determine the abnormality of diabetes health based on the patient's molecular physiological interaction index and output the patient's diabetes health risk index; continue to implement the corresponding diabetes monitoring and management strategy based on the patient's diabetes health risk index; if the abnormal results of the patient's health data are normal, continue to manage the health data of the diabetic patient.

[0044] The specific steps for determining the raw parameter monitoring data and the individual metabolic adaptation domain of a patient, as well as the metabolic adaptation domain of similar patients, include:

[0045] Based on the patient's parameter monitoring data over a recent period, the data is stratified according to diurnal time phases and physiological states to obtain several time-state segments. Key event anchors and patient circadian rhythms are obtained based on the original parameter monitoring data. Using the key event anchors as boundaries, parameter stability intervals and parameter fluctuation profiles are calculated for each time-state segment, serving as the key segment adaptation domain. Data quality weights are calculated for each key segment adaptation domain. Each key segment adaptation domain is then sorted according to the patient's circadian rhythm and data quality weights to obtain time-series-based individual metabolic adaptation domains for each patient.

[0046] After obtaining continuous raw parameter monitoring data from patients, multi-channel data is preprocessed and aligned based on a unified time axis, including blood glucose sequences, molecular marker sequences, sleep stage sequences, and bioelectrical signal sequences. To ensure the naturalness of segment boundaries, the entire time period is first divided into phases such as early morning, daytime, nighttime, and before and after falling asleep according to the circadian rhythm. Then, physiological state information is superimposed to refine the segmentation. For example, signal morphology changes in adjacent intervals are identified by events such as eating, exercise, falling asleep, waking up, and medication. At the same time, change point detection and inflection point search are combined to confirm the stability of the cutting points. For data spanning days or weeks, segments of the same scenario are merged into the same category, and minimum and maximum duration constraints are applied to segments that are too short or too long, so that each phase-state segment can cover the complete physiological process without fragmentation due to noise.

[0047] While forming time-state segments, key event anchors and patients' circadian rhythms are automatically extracted based on raw parameter monitoring data. Key event anchors are derived from cross-verification of multi-source records. For example, wearable monitoring devices provide sleep onset, wakefulness, and nighttime wakefulness times; dietary records and pre- and post-meal blood glucose fluctuations jointly determine the timing and type of meals; pillbox or mobile phone reminders, combined with blood drug effect timing, are used to locate the dosing time and dosage; exercise devices and geolocation indicate the start and end times and intensity of exercise. These events are concatenated in chronological order, and their repeatability and offset within a week and a month are calculated to extract patients' circadian rhythms and medication cycles, which are then used as time references in subsequent segment matching and comparison.

[0048] Using key event anchor points as boundaries, the parameter stability interval and parameter fluctuation profile of each time-state segment are calculated. Specifically, robust denoising and short-window smoothing are performed on the blood glucose and molecular marker sequences within the segment; the central level and upper and lower confidence bands of the segment are estimated; and dynamic elements such as the ascent slope, peak position, peak-to-valley difference, descent rate, and plateau duration are extracted to construct a complete fluctuation morphology description. Sleep and bioelectrical information are simultaneously incorporated to assess the stability of phase transitions and the stress level of the autonomic nervous system, in order to correct for the segment's modulation background of metabolic responses. The obtained parameter stability interval is used to define the range that the segment should exhibit under normal conditions, while the parameter fluctuation profile is used to describe the reasonable change path that the segment may exhibit when driven by events. Together, they constitute the critical segment adaptation domain of the segment. For each critical segment... The adaptation domain is used to calculate and assign corresponding data quality weights to reflect the credibility and contribution of the segment in subsequent judgments. The data quality weights are formed by a combination of factors: for example, sampling completeness measures the proportion and continuity of missing data, signal-to-noise ratio and artifact rate measure motion interference and sensor drift, cross-device consistency measures the degree of agreement between data from different sources during overlapping periods, time alignment credibility measures the synchronization accuracy of multiple channels in the segment, and event label reliability measures the consistency between the segment and the recorded event. In addition, the impact of external environment and unconventional situations (such as cross-time zone travel, night shift work, and acute illness) is considered, and relevant segments are automatically downgraded or marked as delayed to avoid contamination of the regular threshold domain by abnormal conditions. The final quality weight is normalized to a continuous quantity between zero and one and archived together with the segment adaptation domain.

[0049] After identifying the adaptation domains and quality weights of key segments, a time-series individual metabolic adaptation domain is constructed based on the patient's daily routine and medication rhythm. This is achieved by arranging segments within a day and week according to event sequence and biological priority. Priority rules include drug effects taking precedence over postprandial effects, postprandial effects taking precedence over general exercise, and the sleep stage applying monitoring background modulation to the entire process. When segments overlap or conflict, the segment with the higher quality weight takes precedence, or the segments are divided into two consecutive sub-segments in chronological order to retain their respective characteristics. Subsequently, continuous adaptation domain time bands are generated on the time axis, allowing for rapid retrieval of matching segment adaptation domains and their quality weights at any given time, and automatically inheriting the differences between weekdays and rest days on the time axis. The resulting time-series individual metabolic adaptation domain can serve as an immediate comparison of the original monitoring data and a first-level reference for subsequent abnormality gating, reflecting both long-term stable patterns and adaptively updating according to changes in the patient's recent condition.

[0050] Based on a publicly available database of diabetic patient parameters; matching patients' corresponding fingerprint features to output a behavioral-physiological twin patient group; constructing a metabolic adaptation domain for similar patients based on the behavioral-physiological twin patient group; the corresponding fingerprint features of patients are unique feature patterns after compressing and combining their multi-dimensional health data. The individualized fingerprint features of patients can reflect their metabolic characteristics and life rhythms, and can also be matched with the features of other patients in the database.

[0051] The sources of the diabetes patient parameter database mainly include various publicly available or ethically approved medical and research data resources; for example, long-term follow-up data from clinical trials and cohort studies, which often record patients' blood glucose levels, medication use, complication development, and lifestyle characteristics; based on the diabetes patient parameter database, the behavioral-physiological fingerprint characteristics of target patients (such as sleep structure ratio, postprandial blood glucose response delay pattern, medication adherence rhythm, and bioelectrical feedback pattern) are matched to screen out a group of behavioral-physiological twin patients who are highly similar in multimodal characteristics, which can be used as a reference for group data indicators;

[0052] Within the behavioral-physiological twin patient group, a secondary segmentation and adaptation domain reconstruction are performed according to the phase-state segmentation method. The stable intervals and fluctuation profiles of various segments are weighted and integrated, and deviations introduced by regional, seasonal or equipment differences are corrected. The resulting metabolic adaptation domains of the same patient group can reflect the overall pattern of the group and retain the key phase dependence, providing a dynamic and personalized group reference for the determination of abnormal deviations in target patients.

[0053] Based on the original parameter monitoring data, the judgment is made in the individual metabolic adaptation domain of the patient and the metabolic adaptation domain of similar patients to obtain abnormal results of the patient's health data;

[0054] Newly acquired raw parameter monitoring data undergoes data normalization. First, a first round of comparison is performed within the individual patient's metabolic adaptation domain. Specifically, key indicators for the current segment (including the rise and fall characteristics of blood glucose morphology, the intensity and direction of changes in molecular regulatory ratio parameters, the degree of metabolic modulation by sleep rhythm, and abnormal neural activity indicated by bioelectrical signals) are matched item by item with the corresponding stable interval and fluctuation profile of that segment. The direction, amplitude, and duration of deviation are calculated, and an individual domain deviation score and its confidence level are formed by combining the segment's data quality weights. Then, a second round of comparison is performed within the metabolic adaptation domains of similar patients. To eliminate environmental and equipment differences... First, the raw data is corrected for region, season, and caliber. Then, under the same time-phase and state, the distribution bands of patient indicators and twin populations, typical fluctuation curves, and typical response time histories are compared one by one to obtain the population domain bias score and its confidence level. The individual metabolic adaptation domain of the patient is set as the first gate, and the metabolic adaptation domain of the same type of patient is set as the second gate. Only when both gates are opened simultaneously in the same observation window and the deviation direction is consistent, the amplitude reaches their respective minimum effective thresholds, and the minimum duration requirement is met, is it considered a qualified trigger, that is, the abnormal result of the patient's health data is normal. If only one gate is triggered or the two gates are in opposite directions, it is marked as non-abnormal, that is, the abnormal result of the patient's health data is abnormal.

[0055] The specific steps for assessing diabetic health abnormalities using molecular physiological interaction indices in patients include:

[0056] The fluctuation lag strength is identified based on the patient molecular physiological interaction index and causal inference model; the patient molecular physiological interaction index is corrected according to the fluctuation lag strength to obtain the corrected patient molecular physiological interaction index; the corrected patient molecular physiological interaction index is used to judge diabetic health abnormalities using the abnormal coupling effect judgment mechanism to obtain the patient's diabetic health risk index.

[0057] Specifically, the modified patient molecular physiological interaction index is divided into N types of modified diabetes abnormality indicators; the contribution of each type of modified diabetes abnormality indicator is calculated; if N types of modified diabetes abnormality indicators are superimposed in the same time window and their contributions all exceed the preset threshold, the patient's diabetes health risk indicator is determined to be a composite diabetes risk; if only some of the modified diabetes abnormality indicators are superimposed in the same time window or any contribution is lower than the preset threshold, the patient's diabetes health risk indicator is determined to be a diabetes observation risk; the preset threshold is set manually by professional technicians.

[0058] The corrected patient molecular physiological interaction index was grouped into three categories: the first category was molecular, sleep, and neurological; the second category was postprandial, nighttime, exercise recovery, and post-medication; and the third category was rapid onset, slow accumulation, and repeated fluctuations. Under the same category, similar segments were aggregated into N categories of corrected diabetes abnormality indicators based on amplitude, duration, and frequency of occurrence. Then, the contribution of each category was calculated within the selected time window. First, the average amplitude and coverage duration of the abnormality were assessed. Then, the consistency with synchronous deviations from other channels, the strength of deviation relative to the same population, the physiological matching degree with event anchors (eating, medication, falling asleep, etc.), and the data quality and delay alignment reliability of the segment were added. The above factors were combined into a standardized score as the contribution. Within the same time window, the scores of each category were summarized and checked to see if they all exceeded the preset threshold. If all were met, it was judged as a composite risk. If only part were met or any contribution was insufficient, it was judged as an observation risk. At the same time, the score ranking of each category was retained for subsequent intervention priority.

[0059] The causal inference model is trained based on historical data. Using the patient's molecular physiological interaction index as input, the model is evaluated to determine the hysteresis intensity of fluctuations in the current monitoring period. Hysteresis intensity refers to the strength of the delayed effect of changes in a certain type of physiological signal (such as molecular marker levels, sleep rhythm parameters, or bioelectrical feedback characteristics) on blood glucose fluctuations. It not only indicates a causal relationship between the signal and blood glucose but also quantifies the degree of lag and magnitude of this relationship over time. For example, an increase in a molecular marker may only trigger a significant increase in blood glucose several hours later. If this effect is consistently stable across multiple time windows, the hysteresis intensity is considered high. Conversely, if the effect is weak or unstable across different windows, the intensity is low. Hysteresis intensity aims to avoid misjudgments based solely on instantaneous values ​​and more accurately reveal potential pathological mechanisms and individual differences.

[0060] The specific steps for training the causal inference model include: preprocessing long-term collected blood glucose sequences, molecular marker concentration curves, sleep quality records, and bioelectrical feedback signals, including time alignment, missing value imputation, and outlier removal; constructing candidate causal structure graphs, using different physiological variables as nodes, and setting possible causal paths based on event anchors, such as eating, medication, and falling asleep; using statistical learning methods to fit causal relationships on historical data and identify the chronological and lag relationships between signals; introducing cross-validation and counterfactual testing during training to ensure that the model can not only fit the data but also maintain explanatory power in unseen time periods; and outputting the causal strength and time delay distribution among variables, continuously updating parameters with new data to achieve adaptive optimization, thereby more accurately characterizing the physiological regulatory mechanisms of diabetic patients in different states.

[0061] The molecular physiological interaction index of patients is modified by both time and weight based on the intensity of fluctuation hysteresis. The specific steps are as follows: within the identified delay window, corresponding time offsets are applied to different channels to realign truly causally significant linkage segments; synchronous segments that do not conform to physiological time sequences or have low credibility are downweighted or removed; evidence segments within high-quality sampling intervals and consistent with event anchors are appropriately weighted; after obtaining the modified molecular physiological interaction index, an abnormal coupling effect judgment mechanism is invoked to assess diabetic health abnormalities. The abnormal coupling effect judgment mechanism is based on a causal inference algorithm and is used to identify abnormal linkages between multiple physiological signals (such as molecular markers, blood glucose, sleep rhythm, and neural activity) in the patient's body. When multiple signals are in a certain state... When anomalies occur simultaneously and influence each other under time delay, this mechanism can determine whether it is a single indicator fluctuation or a clinically significant composite risk. First, causal inference methods are used to identify the lag relationship between each signal, determining the temporal order and causal strength of different channels. Then, real-time monitoring data is compared with individual and population health reference domains to screen for deviations in multiple signals within the same time window that are consistent in direction, significant in amplitude, and of sufficient duration. If the conditions are met, an anomaly determination is triggered, and further classified into observational risk or composite risk. If the signals are inconsistent or insufficient to support a causal relationship, the observational state is maintained. The mechanism for determining anomaly coupling effects is based on causal inference algorithms, combined with historical data and prior physiological knowledge, to construct a causal graph of multiple signals and estimate their delay effects.

[0062] The specific steps for continuing to implement the corresponding diabetes monitoring and management strategy based on the patient's diabetes health risk indicators include:

[0063] When a patient is at multiple risk of diabetes, a high-frequency data collection mode is implemented to shorten the length and frequency of the monitoring cycle; the monitoring cycle is set manually by professional technicians.

[0064] If a patient is at risk of diabetes observation, the monitoring period should be shortened by at least two consecutive monitoring periods. If the patient is not at risk of diabetes complex or diabetes observation in two consecutive monitoring periods, the monitoring period should be restored to the set period. Otherwise, the patient should be subjected to high-frequency data collection mode.

[0065] Specifically, in one embodiment, patient A is identified as having a combined risk of diabetes during a continuous monitoring period. The system automatically activates a high-frequency data acquisition mode, adjusting the original daily blood glucose monitoring (i.e., the set monitoring cycle) to once every four hours, while simultaneously acquiring molecular markers and bioelectrical feedback signals in real time, and shortening the statistical interval for sleep cycle assessment to capture subtle metabolic abnormalities and lag effects. In this high-density mode, the patient's risk status is rapidly verified and updated, and professional technicians simultaneously receive early warning reports containing the chain of evidence for developing personalized intervention plans. In another embodiment, patient B is identified as having a watch risk of diabetes during a monitoring period. The system does not immediately switch to high-frequency acquisition, but instead shortens the assessment window from once a day to twice a day in the next monitoring period, performing more intensive statistical analysis of sleep rhythm fluctuations and blood glucose responses. If the patient's risk results do not trigger a combined risk or watch risk in the next two monitoring periods, the system automatically reverts to the regular monitoring cycle, i.e., daily blood glucose collection and nightly sleep parameter recording. If the risk signal persists or intensifies during this period, the system immediately upgrades to a high-frequency data acquisition mode to ensure that potential deterioration of the condition can be captured and managed in a timely manner.

[0066] Example 2, a health data management system for diabetic patients, see [link to example]. Figure 1 As shown, it includes:

[0067] The health data monitoring module includes a data monitoring unit. The data monitoring unit is used to acquire the patient's blood glucose fluctuations and molecular marker concentration fluctuations according to a set monitoring cycle, perform correlation calculations, and obtain the molecular regulation ratio parameters in the patient's body. It also acquires the patient's sleep cycle fluctuation parameters and bioelectrical feedback fluctuation parameters in real time. Based on the molecular regulation ratio parameters, sleep cycle fluctuation parameters, and bioelectrical feedback fluctuation parameters, it constructs the patient's molecular physiological interaction index.

[0068] The health data management module includes a diabetes anomaly judgment unit and a diabetes health judgment unit. The diabetes anomaly judgment unit acquires the patient's raw parameter monitoring data; it then judges the raw parameter monitoring data against the patient's individual metabolic adaptation domain and the metabolic adaptation domain of similar patients to obtain the patient's health data anomaly result. The diabetes health judgment unit, if the patient's health data anomaly result is abnormal, performs a diabetes health anomaly judgment on the patient's molecular physiological interaction index and outputs the patient's diabetes health risk index; based on the patient's diabetes health risk index, it continues to execute the corresponding diabetes monitoring and management strategy; if the patient's health data anomaly result is normal, it continues to manage the diabetes patient's health data.

[0069] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for managing health data of diabetic patients, characterized in that, Includes the following steps: The patient's blood glucose fluctuations and molecular marker concentration fluctuations are acquired according to a set monitoring period and correlated to obtain the molecular regulation ratio parameter in the patient's body; the patient's sleep cycle fluctuation parameters and bioelectrical feedback fluctuation parameters are acquired in real time; and the patient's molecular physiological interaction index is constructed based on the molecular regulation ratio parameter, sleep cycle fluctuation parameter, and bioelectrical feedback fluctuation parameter. Obtain the patient's raw parameter monitoring data; The raw parameter monitoring data, along with the individual metabolic adaptation domain of the patient and the metabolic adaptation domain of similar patients, are used to determine abnormal results in the patient's health data. If the patient's health data results are abnormal, the patient's molecular physiological interaction index is used to determine the abnormality of diabetes health and output the patient's diabetes health risk index. Continue to implement the corresponding diabetes monitoring and management strategy based on the patient's diabetes health risk indicators; If the patient's health data shows abnormal results but is normal, then continue with the management of the diabetic patient's health data. The specific steps for determining the raw parameter monitoring data and the individual metabolic adaptation domain of a patient, as well as the metabolic adaptation domain of similar patients, include: Based on the patient's parameter monitoring data over a recent period, the data is stratified according to diurnal time phases and physiological states to obtain several time-state segments. Key event anchors and patient circadian rhythms are obtained based on the original parameter monitoring data. Using the key event anchors as boundaries, parameter stability intervals and parameter fluctuation profiles are calculated for each time-state segment, serving as the key segment adaptation domain. Data quality weights are calculated for each key segment adaptation domain. Each key segment adaptation domain is then sorted according to the patient's circadian rhythm and data quality weights to obtain time-series-based individual metabolic adaptation domains for each patient. Based on a publicly available database of diabetic patient parameters; matching patients with their corresponding fingerprint features to output a behavioral-physiological twin patient group; and constructing a metabolic adaptation domain for similar patients based on the behavioral-physiological twin patient group. Based on the original parameter monitoring data, the judgment is made in the individual metabolic adaptation domain of the patient and the metabolic adaptation domain of similar patients to obtain abnormal results of the patient's health data; The specific steps for assessing diabetic health abnormalities using molecular physiological interaction indices in patients include: The fluctuation lag strength is identified based on the patient molecular physiological interaction index and causal inference model; the patient molecular physiological interaction index is corrected according to the fluctuation lag strength to obtain the corrected patient molecular physiological interaction index; the corrected patient molecular physiological interaction index is used to judge diabetic health abnormalities using the abnormal coupling effect judgment mechanism to obtain the patient's diabetic health risk index. Specifically, the modified patient molecular physiological interaction index is divided into N types of modified diabetes abnormality indicators; the contribution of each type of modified diabetes abnormality indicator is calculated; if the N types of modified diabetes abnormality indicators are superimposed in the same time window and the contribution of each type exceeds the preset threshold, the patient's diabetes health risk indicator is determined to be a composite diabetes risk; if only some of the modified diabetes abnormality indicators are superimposed in the same time window or any one of the contribution is lower than the preset threshold, the patient's diabetes health risk indicator is determined to be a diabetes observation risk.

2. The method for managing health data of diabetic patients according to claim 1, characterized in that, The specific steps for constructing a patient's molecular physiological interaction index based on molecular regulation ratio parameters, sleep cycle fluctuation parameters, and bioelectrical feedback fluctuation parameters include: Blood glucose fluctuations and molecular marker concentration fluctuations were correlated by time series to obtain several time segments; these time segments included pre- and post-meal segments, sleep cycle segments, pre- and post-meal segments, and pre- and post-medication segments; the amplitude of changes in blood glucose fluctuations and molecular marker concentration fluctuations in each time segment was extracted to obtain the corresponding molecular regulation ratio parameters. The rhythmic consistency of sleep cycle fluctuation parameters was calculated to obtain the sleep time regulation factor; pattern recognition was performed on the bioelectric feedback fluctuation parameters to extract abnormal bioelectric fluctuation features under different neural states; molecular regulation ratio parameters, sleep time regulation factor and abnormal bioelectric fluctuation features were synchronously mapped on a unified time axis to obtain a multidimensional blood glucose-molecular interaction matrix. Based on the degree of synchronization deviation of different dimensions of features in the multidimensional glucose-molecular interaction matrix, a patient molecular physiological interaction index is generated.

3. The method for managing health data of diabetic patients according to claim 2, characterized in that, The specific steps for continuing to implement the corresponding diabetes monitoring and management strategy based on the patient's diabetes health risk indicators include: When patients are at multiple risk of diabetes, a high-frequency data collection mode is implemented to shorten the length and frequency of monitoring cycles. If a patient is at risk of diabetes observation, the monitoring period should be shortened by at least two consecutive monitoring periods. If the patient is not at risk of diabetes complex or diabetes observation in two consecutive monitoring periods, the monitoring period should be restored to the set period. Otherwise, the patient should be subjected to high-frequency data collection mode.

4. The method for managing health data of diabetic patients according to claim 3, characterized in that, The abnormal coupling effect determination mechanism is based on a causal inference algorithm.

5. A health data management system for diabetic patients, characterized in that, The system employs a method for managing health data of diabetic patients as described in any one of claims 1-4, comprising: The health data monitoring module includes a data monitoring unit. The data monitoring unit is used to acquire the patient's blood glucose fluctuations and molecular marker concentration fluctuations according to a set monitoring cycle, perform correlation calculations, and obtain the molecular regulation ratio parameters in the patient's body. It also acquires the patient's sleep cycle fluctuation parameters and bioelectrical feedback fluctuation parameters in real time. Based on the molecular regulation ratio parameters, sleep cycle fluctuation parameters, and bioelectrical feedback fluctuation parameters, it constructs the patient's molecular physiological interaction index. The health data management module includes a diabetes anomaly judgment unit and a diabetes health judgment unit. The diabetes anomaly judgment unit acquires the patient's raw parameter monitoring data; it then judges the raw parameter monitoring data against the patient's individual metabolic adaptation domain and the metabolic adaptation domain of similar patients to obtain the patient's health data anomaly result. The diabetes health judgment unit, if the patient's health data anomaly result is abnormal, performs a diabetes health anomaly judgment on the patient's molecular physiological interaction index and outputs the patient's diabetes health risk index; based on the patient's diabetes health risk index, it continues to execute the corresponding diabetes monitoring and management strategy; if the patient's health data anomaly result is normal, it continues to manage the diabetes patient's health data.

Images