A diabetes early warning and risk assessment method and system

Abstract

The present invention relates to the field of disease prevention technology, and discloses a diabetes early warning and risk assessment method and system, including: collecting clinical test data of patients, generating complication baseline data based on the clinical test data, and obtaining the patient's real-time blood glucose data through a portable device; constructing a cross-clock early warning model based on a Bayesian reasoning network, using the complication baseline data as a slow clock and the real-time blood glucose data as a fast clock, integrating a three-layer gating mechanism to adjust the observation weight, and updating and dynamically outputting the risk level through logarithmic Bayesian cumulative update; when the risk level exceeds a preset threshold, or the uncertainty of the cross-clock early warning model exceeds an upper limit, triggering a review schedule, and outputting the risk level and review date. The present invention has high clinical feasibility and scalability, is particularly suitable for individualized risk management of diabetes, and can significantly improve the efficiency of early screening for complications and the quality of blood glucose control.

Description

Technical Field

[0001] The present invention relates to the technical field of disease prevention, and in particular to a diabetes early warning and risk assessment method and system. Background Art

[0002] Diabetes is a metabolic disease characterized by chronic hyperglycemia. As the disease progresses, it often leads to a variety of serious complications, including diabetic retinopathy, diabetic nephropathy, peripheral neuropathy, and cardiovascular and cerebrovascular disease. These complications often lack obvious symptoms in the early stages, and once obvious clinical manifestations appear, they are usually in the middle or late stages. Treatment is difficult and intervention costs are high, making them the main cause of disability and death in diabetes.

[0003] Although continuous glucose monitoring (CGM) technology has become widely used in daily blood sugar control in recent years, it still has significant limitations in the early identification and prediction of complications. Currently, interventions often rely on periodic testing at hospitals, which fails to dynamically reflect the complex relationship between a patient's blood sugar fluctuations and the evolution of various complications. It also fails to rationally arrange follow-up plans based on the patient's individual risk status, resulting in both wasted resources and delayed treatment.

[0004] In response to the above problems, the present invention proposes a diabetes complication early warning and follow-up scheduling method based on cross-clock reasoning, which realizes the fusion modeling of clinical test data and real-time blood glucose data, can dynamically evaluate the development risk of multiple complications, and intelligently recommend review time, thereby improving early screening rate and resource utilization efficiency. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is that there are still significant limitations in the early identification and prediction of diabetic complications.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: a diabetes early warning and risk assessment method, comprising: collecting clinical test data of a patient, generating complication baseline data based on the clinical test data, and obtaining the patient's real-time blood glucose data through a portable device;

[0008] A cross-clock early warning model is constructed based on a Bayesian inference network, using complication baseline data as a slow clock and real-time blood glucose data as a fast clock. A three-layer gating mechanism is integrated to adjust the observation weights, and the risk level is dynamically output through logarithmic Bayesian cumulative updates.

[0009] When the risk level exceeds the preset threshold, or the uncertainty of the cross-clock warning model exceeds the upper limit, the review scheduling is triggered and the risk level and review date are output.

[0010] As a preferred embodiment of the diabetes early warning and risk assessment method described in the present invention, the clinical detection data includes retinopathy risk data, nephropathy risk data, neuropathy risk data and cardiovascular risk data.

[0011] As a preferred embodiment of the diabetes early warning and risk assessment method of the present invention, the real-time blood glucose data includes collecting continuous blood glucose values ​​of the patient within a fixed time window by a portable device and marking them with timestamps, and calculating blood glucose fluctuation characteristics within the time window;

[0012] Obtain the patient's behavioral information records, align the behavioral information records with blood sugar fluctuation characteristics, and identify event-related fluctuation intervals;

[0013] A blood glucose threshold with a fixed time window is set based on the event-related fluctuation interval, and blood glucose detection and early warning are achieved by comparing the blood glucose fluctuation characteristics and the blood glucose threshold.

[0014] As a preferred embodiment of the diabetes early warning and risk assessment method described in the present invention, the cross-clock early warning model includes: the cross-clock early warning model adopts a dual-time axis modeling framework, each type of complication corresponds to a slow clock hidden state trajectory, and each hidden state trajectory is driven and updated by two types of inputs:

[0015] Blood glucose fluctuation characteristics from the fast clock;

[0016] baseline data on complications from the slow clock;

[0017] A three-layer gating mechanism is established between the fast clock and the slow clock to generate gating weights for each fixed time window, regulating the response intensity of different complications to the fast clock data;

[0018] Log-cumulative Bayesian updating was used to update the status based on blood glucose fluctuation characteristics and complication baseline data;

[0019] Based on the mutual influence of complications, risk transduction jump nodes between complications are established, and the risk transmission relationship between complications is calculated.

[0020] As a preferred embodiment of the diabetes early warning and risk assessment method of the present invention, the three-layer gating mechanism includes three layers: individual sensitivity coefficient, intraday rhythm coefficient and real-time event coefficient;

[0021] The individual sensitivity coefficient is a fixed value set according to clinical test data. The individual sensitivity coefficient is positively correlated with the degree of complication development and is updated each time clinical test data is collected.

[0022] The intraday rhythm coefficient includes retrieving real-time blood glucose data from the past month, calculating a quantile median smooth curve on a 24-hour scale, and using this as the intraday baseline. The intraday rhythm coefficient is positively correlated with the degree to which the current window blood glucose data deviates from the intraday baseline.

[0023] The real-time event coefficient is determined based on the complication preset template library, and the calculation of the gating weight is expressed as:

[0024] ;

[0025] in, Represents a time window gating weight within; represents the individual sensitivity coefficient; represents the intraday rhythm coefficient; Indicates the real-time event coefficient.

[0026] As a preferred embodiment of the diabetes early warning and risk assessment method of the present invention, the method of updating and dynamically outputting the risk level through logarithmic Bayesian cumulative updating includes initializing the corresponding state prior distribution of the complication baseline data and calculating the gating weight within the time window;

[0027] Based on the blood glucose fluctuation characteristics within the time window, the observation likelihood is fitted under the complication baseline data, and Bayesian update is performed through logarithmic accumulation, which is expressed as:

[0028] ;

[0029] in, Indicates at a point in time , after Complications after a fast clock time window The posterior probability of the updated risk state; The risk level indicating the risk status; Indicates complications At the time point The initial posterior probability of ; Indicates the Complications under a fast clock time window The gating weight of Indicates the Blood glucose fluctuation characteristics in a fast clock time window; Indicates logarithmic calculation;

[0030] Obtain the posterior probability distribution through Softmax normalization ;

[0031] The risk transmission relationship between complications is calculated through the risk transduction jump node, and the risk jump factor is calculated for each pair of related complications:

[0032] ;

[0033] in, Indicates source complications Target complications At the time point The jump factor of The calculation function representing the jump factor; represents the gate weight; Indicates source complications Blood sugar fluctuation characteristics; Indicates target complications Blood sugar fluctuation characteristics;

[0034] like Greater than the preset threshold ,Will As a gain term, adjust the target complication The posterior state distribution of :

[0035] ;

[0036] in, Indicates complications after jump factor adjustment The posterior probability distribution of ; represents the normalization function; Indicates unregulated complications The posterior probability distribution of ; Indicates the jump gain coefficient;

[0037] The posterior probability distribution of complications is mapped to risk levels. When the risk level exceeds the preset threshold, a review schedule is triggered.

[0038] As a preferred embodiment of the diabetes early warning and risk assessment method described in the present invention, the review scheduling includes forming a set of candidate review dates in a 7-day step within the next 2-26 weeks based on the current date. For each candidate review date, the information gain and resource cost are calculated, a benefit index is generated, and the date corresponding to the highest benefit index is output as the review date.

[0039] A diabetes early warning and risk assessment system using any of the methods described in the present invention, wherein: a collection module collects clinical test data of a patient, generates complication baseline data based on the clinical test data, and obtains real-time blood glucose data of the patient via a portable device;

[0040] The prediction module builds a cross-clock early warning model based on a Bayesian inference network, using complication baseline data as a slow clock and real-time blood glucose data as a fast clock. It integrates a three-layer gating mechanism to adjust observation weights, and dynamically outputs risk levels through logarithmic Bayesian cumulative updates.

[0041] The scheduling module triggers review scheduling and outputs the risk level and review date when the risk level exceeds the preset threshold or the uncertainty of the cross-clock warning model exceeds the upper limit.

[0042] Beneficial effects of the present invention: The present invention realizes deep fusion modeling of slow clocks and fast clocks by constructing a cross-clock Bayesian inference network, combines a multi-layer gating mechanism to dynamically adjust the observation weights, and continuously corrects the risk status of various complications driven by minute-level data. The introduced risk transduction jump mechanism can effectively model the risk transmission between comorbid symptoms and improve the linkage recognition ability. At the same time, based on the risk-benefit and review cost, the review time is intelligently recommended to avoid the coexistence of resource waste and missed detection. The present invention has high clinical feasibility and scalability, and is particularly suitable for personalized risk management of diabetes, and can significantly improve the efficiency of early screening of complications and the quality of blood sugar control. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive efforts.

[0044] Figure 1 This is an overall flow chart of a diabetes early warning and risk assessment method provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0045] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0046] Example 1, with reference to Figure 1 , as one embodiment of the present invention, provides a diabetes early warning and risk assessment method, comprising:

[0047] S1: Collect the patient's clinical test data, generate complication baseline data based on the clinical test data, and obtain the patient's real-time blood glucose data through a portable device.

[0048] In the management of diabetic patients, long-term abnormal blood sugar levels not only trigger acute metabolic disorders but, more importantly, can gradually induce a variety of chronic complications, including retinopathy, renal impairment, neuropathy, and cardiovascular disease, without obvious clinical symptoms. To address the need for early warning of these serious complications, this invention establishes a dual-dimensional monitoring mechanism combining a professional baseline test with real-time blood sugar trends, enabling personalized high-risk identification and early warning.

[0049] Specifically, based on the four types of high-risk complications currently recognized in clinical practice, the following four types of structured medical data are collected as individual complication risk baselines:

[0050] Fundus photography and AI-powered automated scoring are used to assess the risk of diabetic retinopathy (DR). Fundus photography and AI-powered automated scoring have been widely used clinically, enabling high-resolution lesion identification. Numerous studies have confirmed that the development and progression of DR are closely associated with chronically elevated blood sugar levels and variability.

[0051] The urine albumin-to-creatinine ratio and estimated glomerular filtration rate are used to assess the risk of diabetic kidney disease (DKD). These measurements have been standardized globally and are core indicators of the clinical grading of diabetic kidney disease. Sustained hyperglycemia has been shown to impair glomerular filtration function and is a key driver of DKD progression.

[0052] Nerve conduction velocity, or quantitative sensory scoring, is used to assess the risk of diabetic peripheral neuropathy (DPN). This indicator is the internationally accepted diagnostic standard for neuropathy, and chronic hyperglycemia and blood sugar fluctuations are positively correlated with nerve conduction dysfunction.

[0053] Coronary artery calcium scores, levels, or cardiac ultrasound indicators are used to assess the risk of cardiovascular complications in patients with diabetes. These biomarkers and imaging parameters are widely used in predicting the risk of coronary artery disease and heart failure. The role of blood glucose fluctuations in promoting endothelial function, inflammatory responses, and atherosclerosis has been well established clinically.

[0054] It should be noted that this invention focuses on complications that are difficult to detect with portable devices. Detection that can be accomplished with portable devices is not within the scope of this invention. For example, for diabetic foot or plantar ulcers, mature wearable products such as smart insoles and smart socks already enable continuous at-home monitoring. Furthermore, conditions such as autonomic dysfunction, sleep apnea syndrome, and diabetes-related cognitive impairment currently lack standardized early detection protocols or predictive models, and are not the focus of this invention's modeling.

[0055] Furthermore, the system first obtains the basic information of the patient to initialize the individual feature vector, and then obtains four types of clinical test data, namely, retinal lesion risk data, kidney disease risk data, neuropathy risk data and cardiovascular risk data, through hospitals, etc. The specific form of the clinical test data and its acquisition are as described above. There are currently mature technologies, so the present invention will not describe it in detail and directly obtain the test results.

[0056] Map clinical test data to complication baseline data. Clinical test data is specific data collected clinically. According to clinical consensus, the specific data is mapped to the risk level of the corresponding complication, for example, it is mapped to four levels: low, medium, high and very high. The larger the value, the higher the risk of the corresponding complication. This risk level is the complication baseline data.

[0057] It should be noted that since the clinical detection data of these complications can only be obtained through professional hospital testing, the complication baseline data constructed by the present invention is data for long-term use. Based on the complication baseline data, the probability of complications in the future is estimated according to the degree and time of the impact of blood sugar fluctuations on complications, and the detection and early warning of these complications are realized. When the risk is high, an alarm is issued and the patient is prompted to go to the hospital for testing.

[0058] Furthermore, the patient's blood sugar is collected in real time through a wearable CGM (continuous glucose monitoring) device, with a sampling interval of no more than 5 minutes. The obtained blood sugar data is uploaded to the patient's mobile device. Each record includes the blood sugar value, collection timestamp, abnormal signal mark and device calibration status.

[0059] Secondly, patients use the APP to record behavioral events closely related to blood sugar fluctuations, including meal time and type, exercise time and type, medication use time, sleep start and end time, etc. The system can also be connected to smart wearable devices (such as watches, exercise rings) to realize automatic recognition and synchronous labeling of some behaviors.

[0060] After uploading, the real-time blood glucose data undergoes data cleaning and denoising. The system calculates key dynamic characteristic indicators based on a 5-15 minute sliding window, including blood glucose variability, high blood glucose exposure time, low blood glucose exposure time, average blood glucose, blood glucose standard deviation and maximum change rate, etc., to construct blood glucose fluctuation characteristics.

[0061] The system further aligns behavioral events with blood glucose trend sequences, automatically annotating event-related blood glucose fluctuation intervals through a sliding time window mechanism for subsequent inference model input. This behavioral annotation data, along with the blood glucose feature matrix, is fed into a cross-clock Bayesian inference network. This network, combined with the slow-clock complication risk baseline, dynamically updates the probability of complication progression.

[0062] When the system detects that the dynamic pattern of blood sugar presents a high-risk structure (such as blood sugar continues to rise / fall beyond the threshold and the fluctuation amplitude is abnormal), or deviates significantly from the historical baseline, it will automatically issue risk warning signals of different levels and link the intervention module to generate personalized treatment suggestions.

[0063] S2: A cross-clock early warning model is constructed based on the Bayesian inference network, with complication baseline data as the slow clock and real-time blood glucose data as the fast clock. A three-layer gating mechanism is integrated to adjust the observation weight, and the risk level is updated and dynamically output through logarithmic Bayesian accumulation.

[0064] Furthermore, to achieve real-time risk monitoring and personalized dynamic prediction of typical diabetic complications, the present invention constructs a cross-clock early warning model. This model uses a dual-time axis modeling framework. Each type of complication corresponds to a slow-clock latent state trajectory, and each latent state trajectory is driven by two types of inputs:

[0065] Blood glucose fluctuation characteristics of the fast clock;

[0066] Baseline data on complications of slow clocks.

[0067] It should be noted that the blood sugar levels of some diabetic patients do not fluctuate much, but remain in a relatively high range for a long time, which can easily induce chronic complications such as retinopathy and glomerulosclerosis.

[0068] To detect the pressure of complications caused by elevated blood sugar levels, the slow clock input features, in addition to complication baseline data, further incorporate a hyperglycemia exposure time metric calculated using a continuous glucose monitoring system, such as the cumulative time with blood sugar levels above 10 mmol / L over the past 90 days (TAR-90d). This metric measures the burden of chronic hyperglycemia and serves as an important input factor for complication risk accumulation. It is mapped, along with the baseline vector, into the initial prior distribution of the complication status inference model.

[0069] This indicator influences the probability of transitioning from complication states during the modeling process. When TAR-90d is above the upper quartile of the normal population distribution, the rate of transition from early-stage to mid- to late-stage states is enhanced. By introducing this long-term indicator, the present invention can effectively identify the risk of delayed complications associated with "relatively stable but chronically elevated blood sugar levels."

[0070] Furthermore, fluctuations in continuous glucose monitoring (fast clock) data affect different complications to varying degrees, and this impact changes dynamically with individual constitutions and scenarios. Indiscriminately using all glucose fluctuations to update latent complication states would amplify noise and diminish the value of the semi-annual clinical baseline (slow clock data). To this end, this paper proposes a three-tiered, interpretable gating framework that dynamically adjusts the importance of different glucose windows for different complications using hierarchical coefficients. Higher gating values ​​indicate a greater impact of the current glucose window on the risk assessment of a particular complication.

[0071] Specifically, the three-layer gating mechanism includes individual sensitivity coefficients, intraday rhythm coefficients, and real-time event coefficients. Each coefficient is assigned a fixed value by mapping a numerical range. This not only regulates the impact of fast clocks on slow clocks, but also allows for rapid calculations without consuming computing power.

[0072] The individual sensitivity coefficient is a fixed value set based on clinical test data. It is positively correlated with the degree of complication development and is updated each time clinical test data is collected. For example, based on baseline complication data, a fixed individual sensitivity coefficient is set based on risk level, with low risk corresponding to 0.1 and medium risk corresponding to 0.3.

[0073] The intraday rhythm coefficient involves retrieving real-time blood glucose data from the past month and calculating a quantile median smooth curve on a 24-hour scale. This curve serves as the intraday baseline. The intraday rhythm coefficient is positively correlated with the degree to which the current window's blood glucose data deviates from the intraday baseline. For example, if a safety range is set, if the current window's blood glucose falls within ±10% of the baseline, the intraday rhythm coefficient is 0.5; if it exceeds the baseline ±10%, the intraday rhythm coefficient is 1.

[0074] The real-time event coefficient is determined based on the preset complication template library. For example, the DR high-risk template is blood glucose continuously >10mmol / L for ≥2 hours at night. The high-risk template and threshold for each complication are set based on professional knowledge. The high-risk threshold can also be adjusted according to the patient's condition. When the high-risk template is met, the real-time event coefficient is 1.

[0075] The system calculates the gating weight in real time within each time window, which is expressed as:

[0076] ;

[0077] in, Represents a time window gating weight within; represents the individual sensitivity coefficient; represents the intraday rhythm coefficient; Indicates the real-time event coefficient. It is used as the scaling factor of the fast clock observation likelihood and input into the cross-clock early warning model to complete the hidden state posterior update.

[0078] Furthermore, to achieve dynamic coupling between minute-by-minute blood sugar fluctuations (fast clock data) and semi-annual clinical indicators (slow clock data), a Bayesian update structure is employed to maintain continuous probabilistic reasoning of latent states. This structure enables the system to continuously refine its estimated probabilities of various complication progression stages as noninvasive data streams continue to feed in, thereby enabling continuous risk quantification and stratification.

[0079] To avoid model misjudgment or oversensitivity caused by high-frequency observations, the present invention designs a three-layer gating function that dynamically assigns complication-related weighting factors to the observation data of each fast clock window. Considering that the gating coefficient may be significantly less than 1, to prevent the loss of numerical precision in the probability product, the present invention adopts the Bayesian posterior accumulation formula in the logarithmic domain, which is expressed as:

[0080] ;

[0081] in, Indicates at a point in time , after Complications after a fast clock time window The posterior probability of the updated risk state; The risk level indicating the risk status; Indicates complications At the time point The initial posterior probability of ; Indicates the Complications under a fast clock time window The gating weight of Indicates the Blood glucose fluctuation characteristics in a fast clock time window; Denotes logarithmic calculation. Finally, the posterior probability distribution is obtained through Softmax normalization.

[0082] It's important to note that the gating coefficients calculated by the three-tier gating mechanism are generally between 0.3 and 1. Decimal multiplication factors can lead to precision loss in multi-window accumulation, while logarithmic accumulation maintains stability. Compared to hard-judgment state transitions or fixed classifications, the use of continuous probability distributions preserves both the trend and ambiguity of disease progression, facilitating dynamic alerting and follow-up scheduling.

[0083] Furthermore, numerous studies have shown that diabetic complications such as diabetic keratinization (DKD), cardiovascular disease (CV), and diabetic pulmonary neuropathy (DPN) have significant overlap in their pathological mechanisms. Furthermore, for multi-task Bayesian networks, introducing intermediate correlated variables can improve the accuracy of the primary task in the presence of correlated targets. Therefore, this paper proposes a risk transduction jump node structure that explicitly introduces cross-information transduction pathways between complication risks, enabling risk sharing and early warning across high-risk pathways.

[0084] Specifically, a risk transduction jump node is added between each pair of potentially related complication latent state nodes. The node does not represent an independent stage of disease onset, but rather represents the intensity and direction of the impact of the source risk on the target risk. The risk jump factor is calculated for each pair of related complications:

[0085] ;

[0086] in, Indicates source complications Target complications At the time point The jump factor of The calculation function representing the jump factor; represents the gate weight; Indicates source complications Blood sugar fluctuation characteristics; Indicates target complications Blood sugar fluctuation characteristics.

[0087] It should be noted that Rather than a fixed formula, a discriminatory calculation model should be designed based on the medical correlations and data characteristics between different complication pairs. Taking DKD→CV as an example, this type of relationship has a large number of pathological bases in medical research. The jump factor can be constructed based on the following two dimensions:

[0088] First, whether the fast clock gating weight is enhanced synchronously.

[0089] like and A synchronous increase indicates that the source symptoms are weighted up, which may have an implicit induction on the target symptoms. The similarity score can be obtained by calculating the Pearson correlation or DTW dynamic matching of the gated coefficients in the sliding window.

[0090] Second, whether the fast clock observation characteristics match the known cross-high-risk event template.

[0091] For example: persistent high blood sugar at night + large fluctuation in the morning → is associated with the simultaneous worsening of DKD + CV. Combined with the preset template library of real-time event coefficients, if a cross-high risk is matched, Just take a higher value.

[0092] like Greater than the preset threshold ,Will As a gain term, adjust the target complication The posterior state distribution of :

[0093] ;

[0094] in, Indicates complications after jump factor adjustment The posterior probability distribution of ; represents the normalization function; Indicates unregulated complications The posterior probability distribution of ; Indicates the jump gain coefficient.

[0095] The posterior probability distribution of complications is mapped to risk levels. When the risk level exceeds the preset threshold, a review schedule is triggered.

[0096] S3: When the risk level exceeds the preset threshold, or the uncertainty of the cross-clock warning model exceeds the upper limit, the review scheduling is triggered and the risk level and review date are output.

[0097] Furthermore, the review scheduling is only activated when the system determines that the user is currently at high risk or there is significant uncertainty in the inference model. Specifically, if the high-risk probability of any complication in the cross-clock early warning model exceeds the set threshold, or the system detects a persistently high entropy value in the continuous sliding window, the system will deem the current state to require additional slow clock information calibration. In this case, the review scheduling is automatically activated, evaluating the optimal review time in the next few weeks and making a decision based on the comprehensive risk-benefit and treatment cost.

[0098] Review scheduling combines the information entropy reduction potential of the comprehensive risk curve and the consumption of medical resources to dynamically determine the optimal date for the next in-hospital review.

[0099] First, the system caches the current entropy value of each complication and generates a set of candidate review dates on a weekly basis within the 2–26 week range. Then, based on historical follow-up data, it predicts the information gain of performing an examination on the candidate dates. It also incorporates parameters such as hospital outpatient workload and patient convenience to calculate a comprehensive benefit index. The date with the highest benefit index is selected as the recommended review date, and alternative dates within the adjacent ±1 week are also output for physicians to adjust based on their actual outpatient schedules.

[0100] Through the above mechanism, the system can keep the uncertainty of complication risk at a controllable level while maintaining an acceptable burden on patients, effectively reduce the probability of missed complications due to insufficient follow-up, and avoid unnecessary high-frequency examinations.

[0101] Example 2: In an exemplary embodiment, a diabetes early warning and risk assessment system is also provided, including an acquisition module for collecting clinical test data of patients, generating complication baseline data based on the clinical test data, and obtaining real-time blood glucose data of patients through a portable device.

[0102] The prediction module builds a cross-clock early warning model based on the Bayesian inference network, uses the complication baseline data as the slow clock and the real-time blood glucose data as the fast clock, integrates the three-layer gating mechanism to adjust the observation weight, and updates and dynamically outputs the risk level through logarithmic Bayesian accumulation.

[0103] The scheduling module triggers review scheduling and outputs the risk level and review date when the risk level exceeds the preset threshold or the uncertainty of the cross-clock warning model exceeds the upper limit.

[0104] If the above functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0105] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0106] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, and then editing, interpreting, or processing in another suitable manner as necessary, and then storing it in a computer memory.

[0107] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0108] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A diabetes early warning and risk assessment method, characterized in that: include: Collect the patient's clinical test data, generate complication baseline data based on the clinical test data, and obtain the patient's real-time blood glucose data through portable devices; A cross-clock early warning model is constructed based on a Bayesian inference network, using complication baseline data as a slow clock and real-time blood glucose data as a fast clock. A three-layer gating mechanism is integrated to adjust the observation weights, and the risk level is dynamically output through logarithmic Bayesian cumulative updates. When the risk level exceeds the preset threshold, or the uncertainty of the cross-clock warning model exceeds the upper limit, the review schedule is triggered and the risk level and review date are output; The clinical test data include retinopathy risk data, nephropathy risk data, neuropathy risk data and cardiovascular risk data; The real-time blood glucose data includes collecting continuous blood glucose values ​​of the patient within a fixed time window through a portable device and marking them with a timestamp, and calculating the blood glucose fluctuation characteristics within the time window; Obtain the patient's behavioral information records, align the behavioral information records with blood sugar fluctuation characteristics, and identify event-related fluctuation intervals; Set a fixed time window blood glucose threshold based on event-related fluctuation intervals, and compare blood glucose fluctuation characteristics with blood glucose thresholds to achieve blood glucose detection and early warning; The cross-clock early warning model includes a dual-time axis modeling framework. Each type of complication corresponds to a slow clock hidden state trajectory, and each hidden state trajectory is updated by two types of input drive: Blood glucose fluctuation characteristics from the fast clock; baseline data on complications from the slow clock; A three-layer gating mechanism is established between the fast clock and the slow clock to generate gating weights for each fixed time window, regulating the response intensity of different complications to the fast clock data; Log-cumulative Bayesian updating was used to update the status based on blood glucose fluctuation characteristics and complication baseline data; Based on the mutual influence of complications, risk transduction jump nodes between complications are established, and the risk transmission relationship between complications is calculated; The three-layer gating mechanism includes three layers: individual sensitivity coefficient, intraday rhythm coefficient and real-time event coefficient; The individual sensitivity coefficient is a fixed value set according to clinical test data. The individual sensitivity coefficient is positively correlated with the degree of complication development and is updated each time clinical test data is collected. The intraday rhythm coefficient includes retrieving real-time blood glucose data from the past month, calculating a quantile median smooth curve on a 24-hour scale, and using this as the intraday baseline. The intraday rhythm coefficient is positively correlated with the degree to which the current window blood glucose data deviates from the intraday baseline. The real-time event coefficient is determined based on the complication preset template library, and the calculation of the gating weight is expressed as: ； in, Represents a time window gating weight within; represents the individual sensitivity coefficient; represents the intraday rhythm coefficient; Indicates the real-time event coefficient.

2. The diabetes early warning and risk assessment method according to claim 1, wherein: The cumulative update and dynamic output of the risk level by logarithmic Bayesian includes initializing the corresponding state prior distribution of the complication baseline data and calculating the gating weight within the time window; Based on the blood glucose fluctuation characteristics within the time window, the observation likelihood is fitted under the complication baseline data, and Bayesian update is performed through logarithmic accumulation, which is expressed as: ； in, Indicates at a point in time , after Complications after a fast clock time window The posterior probability of the updated risk state; The risk level indicating the risk status; Indicates complications At the time point The initial posterior probability of ; Indicates the Complications under a fast clock time window The gating weight of Indicates the Blood glucose fluctuation characteristics in a fast clock time window; Indicates logarithmic calculation; Obtain the posterior probability distribution through Softmax normalization ; The risk transmission relationship between complications is calculated through the risk transduction jump node, and the risk jump factor is calculated for each pair of related complications: ； in, Indicates source complications Target complications At the time point The jump factor of The calculation function representing the jump factor; represents the gate weight; Indicates source complications Blood sugar fluctuation characteristics; Indicates target complications Blood sugar fluctuation characteristics; like Greater than the preset threshold ,Will As a gain term, adjust the target complication The posterior state distribution of : ； in, Indicates complications after jump factor adjustment The posterior probability distribution of ; represents the normalization function; Indicates unregulated complications The posterior probability distribution of ; Indicates the jump gain coefficient; The posterior probability distribution of complications is mapped to risk levels. When the risk level exceeds the preset threshold, a review schedule is triggered.

3. The diabetes early warning and risk assessment method according to claim 2, wherein: The review scheduling includes forming a set of candidate review dates in 7-day steps within the next 2-26 weeks based on the current date, calculating information gain and resource cost for each candidate review date, generating a benefit index, and outputting the date corresponding to the highest benefit index as the review date.

4. A diabetes early warning and risk assessment system, applied to a diabetes early warning and risk assessment method according to any one of claims 1 to 3, characterized in that: include, The acquisition module collects the patient's clinical test data, generates complication baseline data based on the clinical test data, and obtains the patient's real-time blood glucose data through a portable device; The prediction module builds a cross-clock early warning model based on a Bayesian inference network, using complication baseline data as a slow clock and real-time blood glucose data as a fast clock. It integrates a three-layer gating mechanism to adjust observation weights, and dynamically outputs risk levels through logarithmic Bayesian cumulative updates. The scheduling module triggers review scheduling and outputs the risk level and review date when the risk level exceeds the preset threshold or the uncertainty of the cross-clock warning model exceeds the upper limit.

Images