Diabetic neuropathy risk assessment method, device, equipment and storage medium

Abstract

This invention provides a method, device, equipment, and storage medium for assessing the risk of diabetic neuropathy, relating to the field of medical data processing technology. The method includes: acquiring physiological data and treatment behavior data of a target user related to neuropathy; calculating treatment adherence indicators for the target user based on the treatment behavior data; inputting the treatment adherence indicators and key physiological data of neuropathy into a trained neuropathy early warning model to obtain the neuropathy risk probability of the target user; wherein, the neuropathy early warning model is constructed based on the relationship between neuropathy characteristics in the fundus and foot and the severity of neuropathy. This invention can help identify high-risk individuals early, adjust intervention plans in a timely manner, reduce the probability of serious complications, and balance the scientific rigor and practicality of the assessment.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, and in particular to a method, apparatus, device, and storage medium for assessing the risk of diabetic neuropathy. Background Technology

[0002] Diabetic neuropathy is one of the most common chronic complications of diabetes, affecting both central and peripheral nerves. Early symptoms are often subtle, but late-stage complications can lead to serious consequences such as limb ulcers and blindness. Early identification and risk assessment are crucial for slowing disease progression and improving patient prognosis. Public data shows that the incidence of diabetic neuropathy is as high as 50% in patients with a history of diabetes for more than 10 years, and this incidence is increasing with the duration and severity of the disease.

[0003] Traditional neuropathy assessments rely heavily on regular hospital specialist examinations, such as evaluating peripheral nerve function through neurophysiological testing or diagnosing retinal diseases through ophthalmologists interpreting fundus images. This model has several significant limitations: First, it is inherently intermittent and lagging, failing to capture early changes or risk fluctuations between visits, leading to delays in intervention. Second, the data dimensions used in the assessment are often limited, primarily relying on immediate physiological indicators (such as blood glucose) or isolated clinical symptoms, lacking continuous and quantitative consideration of long-term influencing factors such as the patient's daily self-management behaviors, resulting in an incomplete risk assessment. Finally, diagnosis and risk stratification largely depend on the examining physician's clinical experience and subjective judgment; differences between assessors can affect the objectivity and consistency of the assessment results. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and storage medium for assessing the risk of diabetic neuropathy, addressing the issues of low comprehensiveness and reliability in the early assessment of the risk of diabetic neuropathy.

[0005] In a first aspect, embodiments of the present invention provide a method for assessing the risk of diabetic neuropathy, comprising: Acquire physiological data on neuropathic diseases and treatment behavior data of target users; Based on treatment behavior data, calculate treatment adherence indicators for target users; Treatment adherence indicators and key physiological data of neuropathy are input into a trained neuropathy early warning model to obtain the neuropathy risk probability of the target user; the neuropathy early warning model is constructed based on the relationship between the characteristics of neuropathy in the fundus and foot and the severity of neuropathy.

[0006] In one possible implementation, the neuropathy early warning model includes a fundus neuropathy sub-model, a foot neuropathy sub-model, and a fusion assessment model. Treatment adherence indicators and key physiological data of neuropathy are input into the trained neuropathy early warning model to obtain the neuropathy risk probability of the target user, including: By inputting treatment adherence indicators and key physiological data of neuropathy into the fundus neuropathy sub-model, the risk probability of fundus neuropathy for the target user can be obtained. By inputting treatment adherence indicators and key physiological data of neuropathy into the foot neuropathy sub-model, the risk probability of foot neuropathy for the target user can be obtained. By inputting the risk probabilities of fundus neuropathy and foot neuropathy into the fusion assessment model, the comprehensive neuropathy risk probability of the target user is obtained.

[0007] In one possible implementation, the sub-model of fundus neuropathy is as follows:

[0008] in, This represents the probability of retinal nerve damage. The ratio of urine albumin to creatinine. The grade of lesions in fundus imaging. As an indicator of treatment adherence, , , Preset weights.

[0009] In one possible implementation, the foot neuropathy sub-model is as follows:

[0010] in, The probability of foot nerve damage. As a vibration perception index, The ratio of urine albumin to creatinine. As an indicator of treatment adherence, , , Preset weights.

[0011] In one possible implementation, before inputting treatment adherence indicators and key physiological data of neuropathy into a trained neuropathy early warning model to obtain the neuropathy risk probability of the target user, the following steps are also included: Obtain the first training sample set; wherein each training sample in the first training sample set includes a user's treatment compliance index, key physiological data of neuropathy, and the user's fundus neuropathy risk probability label and / or foot neuropathy risk probability label; Based on the first training sample set, the weights of the fundus neuropathy sub-model and the foot neuropathy sub-model are iteratively optimized.

[0012] In one possible implementation, before inputting the risk probabilities of fundus neuropathy and foot neuropathy into the fusion assessment model, the following is also included: Obtain a second training sample set; wherein each training sample in the second training sample set includes a user's risk probability of fundus neuropathy, risk probability of foot neuropathy, and the corresponding comprehensive neuropathy severity label for that user; The initial fusion evaluation model is trained based on the second training sample set to obtain the trained fusion evaluation model.

[0013] In one possible implementation, the formula for calculating treatment adherence indicators is:

[0014] in, As an indicator of treatment adherence, As an indicator of medication adherence, For standardized blood glucose monitoring index, To improve the fitness achievement rate, , , Preset weights.

[0015] Secondly, embodiments of the present invention provide a diabetic neuropathy risk assessment device, comprising: The acquisition module is used to acquire the neuropathological physiological data and treatment behavior data of the target user; The calculation module is used to calculate the treatment adherence index of the target user based on treatment behavior data; The assessment module is used to input treatment adherence indicators and key physiological data of neuropathy into a trained neuropathy early warning model to obtain the neuropathy risk probability of the target user; the neuropathy early warning model is constructed based on the relationship between the characteristics of neuropathy in the fundus and foot and the severity of neuropathy.

[0016] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.

[0017] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect or any possible implementation thereof.

[0018] The diabetic neuropathy risk assessment method, device, equipment, and storage medium provided in this invention integrate physiological data of neuropathy with treatment behavior data, overcoming the shortcomings of single-dimensional assessment in existing technologies. The introduction of treatment compliance indicators quantifies the quality of intervention implementation, accurately reflects its impact on disease progression, and improves the comprehensiveness of the assessment. The neuropathy early warning model constructed based on the relationship between fundus and foot neuropathy characteristics and disease severity can uncover the correlation patterns of lesions in multiple sites, which is more in line with clinical practice than single-site assessment, making the risk probability output more accurate and providing more reliable quantitative references for clinical practice. This helps to identify high-risk groups early, adjust intervention plans in a timely manner, reduce the probability of serious complications, and balance the scientific nature and practicality of the assessment. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the implementation of the diabetic neuropathy risk assessment method provided in this embodiment of the invention. Figure 2 This is a schematic diagram of the structure of the diabetic neuropathy risk assessment device provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0020] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0021] See Figure 1 The flowchart illustrating the implementation of the diabetic neuropathy risk assessment method provided in this embodiment of the invention is described in detail below: Step 101: Obtain the neuropathological physiological data and treatment behavior data of the target user.

[0022] In this embodiment, the physiological data of neuropathy covers core indicators reflecting the function of nerves and related organs, which can be obtained through clinical testing, medical imaging, and measurement using specialized equipment, such as the urine albumin-to-creatinine ratio, fundus imaging data, and vibration perception threshold. Treatment behavior data records key behaviors during the user's disease intervention process, including insulin injection records, blood glucose monitoring logs, and exercise execution records. This data can be simultaneously retrieved from the hospital's electronic medical record system, smart medical device terminals, and the user's health management platform. Some data is combined with user-reported information and verified by medical personnel to ensure data authenticity and completeness. Physiological data directly reflects the underlying state of the user's neuropathy, while treatment behavior data reflects the implementation of intervention measures. The combination of the two provides a comprehensive basis for risk assessment.

[0023] Step 102: Calculate the treatment adherence index of the target user based on treatment behavior data.

[0024] In this embodiment, the treatment adherence index comprehensively considers the user's adherence to medication, disease monitoring, and lifestyle interventions, and can be obtained using a multi-dimensional weighted calculation method. Specifically, medication adherence is reflected by the ratio of actual medication use to the number of times medication is administered; disease monitoring adherence is reflected by the ratio of actual monitoring days to the recommended monitoring days; and the lifestyle intervention dimension focuses on exercise achievement, i.e., the number of days the recommended exercise intensity and duration are actually reached. By assigning appropriate weights to each dimension, the multi-dimensional data is integrated into a single quantitative index. The higher the index value, the better the user's treatment adherence; conversely, a lower value indicates that the intervention measures are not being implemented effectively, which may exacerbate the risk of neuropathy progression.

[0025] Step 103: Input the treatment compliance indicators and key physiological data of neuropathy into the trained neuropathy early warning model to obtain the neuropathy risk probability of the target user; wherein, the neuropathy early warning model is constructed based on the relationship between the characteristics of neuropathy in the fundus and foot and the severity of neuropathy.

[0026] In this embodiment, "neuropathy risk" is a warning signal used to indicate that a target user's health status may have deviated from the normal range or is evolving towards a disease state. Its core meaning is not a medical diagnosis of diabetic neuropathy, but rather, based on multi-dimensional dynamic data, identifying situations that strongly recommend transfer to more rigorous clinical monitoring or require initiation of further specialist diagnostic procedures. The direct use of this risk assessment result is to provide objective and prioritized decision-making basis for tiered follow-up management. For example, when the risk probability assessed by the system exceeds a preset threshold, its purpose is to automatically prompt medical staff to recommend authoritative diagnostic methods such as nerve conduction velocity measurement and detailed ophthalmological examination, or to review and strengthen the user's individualized treatment and management plan, thereby promoting early intervention for high-risk patients.

[0027] The neuropathy early warning model is built upon clinical big data. By analyzing the correlation between fundus and foot neuropathy characteristics and the overall severity of neuropathy in a large number of diabetic patients, core correlation patterns are identified and transformed into model parameters. During model training, cases of neuropathy with different severities are used as samples to learn the mapping relationship between physiological characteristics, treatment adherence, and neuropathy risk. Iterative optimization ensures predictive accuracy. After processing, the input data outputs a risk probability value in the range of 0 to 1. Higher values ​​indicate a greater likelihood of the user developing neuropathy or experiencing a worsening of neuropathy in the short term, providing a quantitative reference for clinical intervention.

[0028] For the clinical big data samples used to train the neuropathy early warning model, this scheme sets clear inclusion and exclusion criteria. The core purpose is to ensure the authenticity, relevance and completeness of the dataset, provide high-quality data support for model training, ensure that the model can accurately capture the characteristic patterns of diabetic neuropathy, and improve the reliability and generalization of the early warning. The inclusion criteria were set primarily based on the core needs of model training: Users diagnosed with diabetes for at least one year were explicitly included because diabetic peripheral neuropathy often develops gradually during disease progression; users with short disease durations have a low incidence of neuropathy, making it difficult to provide effective training samples, thus ensuring the samples have a clear basis for potential disease risk. A complete follow-up record of at least twelve consecutive months, including fundus imaging, neurophysiological examination, laboratory biochemical tests, and self-management behavior logs, was required because the development of neuropathy is a long-term dynamic process, and complete multi-dimensional follow-up data can comprehensively reflect the evolution of nerve function, providing sufficient evidence for the model to capture early characteristics of the disease. The age was limited to between eighteen and eighty years old because users in this age group have relatively stable physical functions, the occurrence and development of neuropathy are typical, and they can cooperate in completing the entire measurement and follow-up process, avoiding sample heterogeneity caused by being too young or too old, and reducing model training errors.

[0029] The exclusion criteria were set primarily to eliminate interfering factors, ensure sample homogeneity, and prevent irrelevant variables from affecting the model's discriminative performance. Excluding users with neuropathy caused by alcohol, genetic factors, autoimmune diseases, or drug toxicity ensures that all neuropathy in the samples is related to diabetes, preventing non-diabetic factors from interfering with the model's learning of diabetic neuropathy characteristics and ensuring the model's warnings are targeted. Excluding users with severe liver or kidney dysfunction, malignant tumors, or other serious diseases that limit life expectancy is because such users have severely impaired bodily functions, may exhibit abnormal indicators unrelated to neuropathy, and are difficult to follow up long-term, leading to data distortion or missing data and affecting dataset quality. Excluding samples with a missing rate of over 30% for key data fields is because excessive missing key data leads to incomplete feature extraction, failing to accurately reflect the user's neurological function status and thus reducing the accuracy of model training.

[0030] Regarding sample size, the training set of this scheme should have no less than 2,000 samples and the independent test set should have no less than 500 samples. Furthermore, resampling or weighting should be used to ensure that the sample distribution of each risk level is relatively balanced. The basis for this is that a sufficient sample size can improve the generalization ability of the model and reduce random errors, while a balanced sample distribution can avoid model bias caused by class imbalance, ensuring that the model can accurately identify neuropathies of different risk levels, and further guaranteeing the model's early warning effect and clinical applicability.

[0031] This invention integrates physiological data of neuropathy with treatment behavior data, overcoming the shortcomings of single-dimensional assessment in existing technologies. The introduction of treatment compliance indicators quantifies the quality of intervention implementation, accurately reflects its impact on disease progression, and improves the comprehensiveness of the assessment. The neuropathy early warning model constructed based on the relationship between fundus and foot nerve lesion characteristics and lesion severity can uncover the correlation patterns of lesions in multiple sites. Compared with single-site assessment, it is more in line with clinical practice, making the risk probability output more accurate and providing more reliable quantitative references for clinical practice. This helps to identify high-risk groups early, adjust intervention plans in a timely manner, reduce the probability of serious complications, and balance the scientific nature and practicality of the assessment.

[0032] In one possible implementation, the neuropathy early warning model includes a fundus neuropathy sub-model, a foot neuropathy sub-model, and a fusion assessment model. Treatment adherence indicators and key physiological data of neuropathy are input into the trained neuropathy early warning model to obtain the neuropathy risk probability of the target user, including: By inputting treatment adherence indicators and key physiological data of neuropathy into the fundus neuropathy sub-model, the risk probability of fundus neuropathy for the target user can be obtained. By inputting treatment adherence indicators and key physiological data of neuropathy into the foot neuropathy sub-model, the risk probability of foot neuropathy for the target user can be obtained. By inputting the risk probabilities of fundus neuropathy and foot neuropathy into the fusion assessment model, the comprehensive neuropathy risk probability of the target user is obtained.

[0033] In this embodiment, the neuropathy early warning model includes a fundus neuropathy sub-model, a foot neuropathy sub-model, and a fusion assessment model. Accurate risk prediction is achieved through modular assessment and comprehensive fusion. The entire assessment process is as follows: First, user treatment adherence indicators and key physiological data on neuropathy are input in parallel into two specialized assessment sub-models. The fundus neuropathy sub-model focuses on analyzing risk factors associated with diabetic retinopathy; it processes inputs such as the urinary albumin-to-creatinine ratio and fundus imaging grade, and outputs a risk probability value specifically for fundus neuropathy.

[0034] Meanwhile, the foot neuropathy sub-model focuses on assessing the risk of distal symmetrical polyneuropathy. It emphasizes the analysis of input data reflecting peripheral nerve function, such as vibration threshold, and outputs a risk probability value for foot neuropathy.

[0035] The specific risk probabilities output by these two sub-models are not directly used as the final conclusion, but are instead fed together into a third-layer fusion assessment model. The role of this fusion model is to learn and comprehensively weigh the risk signals from both the fundus and foot dimensions. By performing higher-level information integration and judgment on these two probability values, it ultimately generates a comprehensive neuropathic risk probability that fully reflects the user's overall neuropathic condition. This overcomes the limitations of single-site assessment and serves as the final neuropathic risk in this application, indicating that the target user's health status may have deviated from the normal range or is evolving into a disease state.

[0036] In one possible implementation, the sub-model of fundus neuropathy is as follows:

[0037] in, This represents the probability of retinal nerve damage. The ratio of urine albumin to creatinine. The grade of lesions in fundus imaging. As an indicator of treatment adherence, , , Preset weights.

[0038] In this embodiment, the urine albumin-to-creatinine ratio is a key indicator for assessing renal microvascular damage. Since renal and retinal microvascular lesions often develop synchronously during the progression of diabetes, this indicator is indirectly related to the degree of retinal neuropathy. OCT (Optical Coherence Tomography) is used to grade retinal lesions. First, optical coherence tomography (OCT) images of the user's fundus are acquired. Based on lesion characteristics such as retinal microaneurysms, hemorrhages, and exudations, they are classified into grades 0 to 4 according to clinical standards. The core criteria for this classification include the number and distribution of microaneurysms, the extent of intraretinal hemorrhage and exudations, beaded venous changes, and the presence or absence of neovascularization, which can be achieved using an image recognition model. The learning process of the image recognition model is based on thousands of images consistently annotated by multiple experts, ensuring complete reproducibility of its output. The interpretation results of the same image at different times are consistent, thus replacing the subjective step of manual image interpretation. Studies have confirmed that the grade of retinal lesions (grades 0 to 4) has an approximately linear increasing relationship with the risk of neuropathy. Based on this, the linear weighting of the grade variable in clinical data has statistical basis. Treatment adherence indicators reflect the long-term impact of glycemic control behaviors on the microvascular environment. Preset weights can be determined based on statistical analysis of clinical cases, reflecting ACR, DR, and [other indicators]. The impact on the risk of fundus neuropathy is assessed by mapping the linear combination results to probability values ​​through a logistic regression function, thereby achieving a quantitative assessment of the risk of fundus neuropathy.

[0039] In one possible implementation, the foot neuropathy sub-model is as follows:

[0040] in, The probability of foot nerve damage. As a vibration perception index, The ratio of urine albumin to creatinine. As an indicator of treatment adherence, , , Preset weights.

[0041] In this embodiment, the sub-model for assessing the risk of peripheral neuropathy in the foot is constructed on a similar principle to the fundus sub-model, but focuses on different input features and is also implemented using a logistic regression model. In the formula, the vibration perception index is calculated based on the vibration perception threshold. The vibration perception threshold is a value measured at locations such as the big toe using standardized quantitative vibration perception testing equipment. It objectively and quantitatively reflects the functional state of large-diameter sensory nerve fibers, and its elevation is a typical early manifestation of diabetic peripheral neuropathy. Specifically, since the development of neuropathy is essentially a cumulative damage process, its early warning signals often manifest first in the accelerated trend of functional decline rather than a single abnormal value. Therefore, introducing time dimension information can significantly improve the sensitivity of identifying subclinical neuropathy. Thus, vibration perception threshold data can be collected repeatedly at a preset cycle, such as once a month, to form a time series of vibration perception thresholds. Subsequently, variation features are extracted from the sequence, including the absolute change between two adjacent measurements, the slope of the linear change based on historical data fitting, and the degree of variation characterized by the standard deviation of the sequence. These time-series features are then normalized and linearly weighted with the current measurements and fused to obtain the vibration perception index, achieving feature dimensionality reduction without sacrificing discrimination effectiveness.

[0042] Because the current measured value, rate of change, and variability of the vibration perception threshold have different dimensions and numerical ranges, they need to be normalized before being input into the logistic regression model. This embodiment uses the Z-score standardization method, which involves subtracting the mean from the training set value from the original value for each feature, and then dividing by the standard deviation of the training set, so that the mean of each feature is zero and the standard deviation is one. Standardized features eliminate the influence of dimensions, making the weight coefficients of each feature comparable during model training, and also accelerating the convergence speed of gradient descent.

[0043] ACR is also introduced here because endonephritis caused by microvascular lesions is one of the important pathological mechanisms of neuropathy. As a behavioral indicator, its impact also encompasses the direct toxic effects of metabolic control on neurotissue. The pre-defined weighting coefficients in the model quantify the relative importance of vibrational sensation, microvascular damage markers, and behavioral management in the risk of foot neuropathy; these coefficients can also be derived from clinical data.

[0044] In one possible implementation, before inputting treatment adherence indicators and key physiological data of neuropathy into a trained neuropathy early warning model to obtain the neuropathy risk probability of the target user, the following steps are also included: Obtain the first training sample set; wherein each training sample in the first training sample set includes a user's treatment compliance index, key physiological data of neuropathy, and the user's fundus neuropathy risk probability label and / or foot neuropathy risk probability label; Based on the first training sample set, the weights of the fundus neuropathy sub-model and the foot neuropathy sub-model are iteratively optimized.

[0045] In this embodiment, to ensure the accurate predictive ability of the aforementioned fundus or foot neuropathy sub-models, they need to be adequately trained before being put into use. This training process begins with constructing a high-quality first training sample set. Each sample in this set corresponds to a historical patient, including the patient's treatment adherence indicators, a complete set of key physiological data on neuropathy, and crucial training labels.

[0046] The labels here are not the overall results that the model directly predicts, but rather independent risk levels of fundus neuropathy and foot neuropathy determined based on the patient's comprehensive clinical examination results and the speed and extent of disease progression (e.g., according to the International Retinopathy Classification for the fundus; and according to nerve conduction studies for the foot). These risk level labels serve as the learning targets for the model, enabling the trained sub-model to accurately identify the probability of neuropathy risk in the corresponding location based on the input variables.

[0047] It's important to clarify that the fundus neuropathy risk label and the DR grade are not the same concept. The DR grade directly describes the severity of retinal microaneurysms, hemorrhages, exudates, etc., and is only a single objective measurement. The fundus neuropathy risk label, however, is a comprehensive clinical risk level, implicitly considering the risk of natural disease progression, the degree of visual threat, and the urgency of intervention. Therefore, the determination of the fundus neuropathy risk label heavily relies on the DR grade, but it is not the only factor. Its determination process involves a refined and dynamic mapping from lesion morphology to clinical risk. For example, two users may have the same DR grade (both moderate non-proliferative), but patient A has a persistently elevated ACR and poor compliance, while patient B has a normal ACR and good compliance. Clinical experts might subconsciously assess A as "medium-high risk" and B as "medium risk." Through training, the model can capture these subtle differences from the data; even with the same DR value, the risk may vary depending on ACR and compliance. The different values ​​output different risk probabilities. This simulates the comprehensive thinking of an expert, thus enabling the determination of fundus neuropathy risk, which is beneficial for accurately identifying the comprehensive neuropathy risk value.

[0048] During the training phase, user physiological data and compliance indicators are used as input, and the corresponding risk level labels are used as the expected output. Specifically, gradient descent can be used to iteratively optimize the weight parameters in the sub-model formula. In each iteration, the error between the model's current predicted value and the true label is calculated, and the weights are adjusted in the direction of reducing the error until the model's predictive performance on the training set reaches a stable and optimal state, thus obtaining a set of optimal weights that can well capture the correlation between features and risk.

[0049] In one possible implementation, before inputting the risk probabilities of fundus neuropathy and foot neuropathy into the fusion assessment model, the following is also included: Obtain a second training sample set; wherein each training sample in the second training sample set includes a user's risk probability of fundus neuropathy, risk probability of foot neuropathy, and the corresponding comprehensive neuropathy severity label for that user; The initial fusion evaluation model is trained based on the second training sample set to obtain the trained fusion evaluation model.

[0050] In this embodiment, after obtaining the risk probabilities for the fundus and foot separately through sub-models, the fusion assessment model needs to be specifically trained to provide a more global, single risk assessment conclusion. This training relies on constructing a second training sample set of different nature. Each sample in this set no longer contains the original physiological or behavioral data, but rather the risk probabilities of fundus neuropathy and foot neuropathy for a historical patient, processed by the trained fundus and foot sub-models or obtained through expert evaluation. Simultaneously, each sample must be equipped with a comprehensive neuropathy severity label determined by a senior clinical expert based on the patient's overall condition (considering multiple aspects such as fundus, nerve, and kidney conditions), for example, categorized as none, mild, moderate, and severe. This label is the ultimate target that the fusion model needs to learn and predict.

[0051] During training, two probability values ​​are used as input features, and the expert's overall severity assessment is used as the output target. Supervised machine learning is then applied to the initial fusion assessment model. By learning from a large number of such samples, the model eventually learns how to accurately infer the user's overall neuropathic risk level based on the numerical combination and pattern of two specific risk probabilities, thus achieving an improvement from local assessment to comprehensive judgment.

[0052] In one possible implementation, the formula for calculating treatment adherence indicators is:

[0053] in, As an indicator of treatment adherence, As an indicator of medication adherence, For standardized blood glucose monitoring index, To improve the fitness achievement rate, , , Preset weights.

[0054] In this embodiment, regarding treatment adherence indicators The specific calculation can be achieved by integrating multi-source behavioral data using a linear weighted formula. The calculation of medication adherence indicators must be compatible with users on different treatment plans and ensure the comparability of indicator values. Specifically, firstly, the medication plan type is determined based on the target user's electronic prescription or medical order, including oral medication alone, insulin injection alone, or oral medication combined with insulin. For oral medication, the total number of actual doses and the total number of doses required are recorded within a statistical period; for insulin, the total number of actual injections and the total number of injections required are recorded. For those using combined medications, the oral medication adherence rate and insulin adherence rate are calculated separately, and then a weighted average is calculated according to the weights of the two drugs in the overall treatment plan. These weights are pre-set by clinical experts based on the contribution of each drug to glycemic control. The final medication adherence indicator is the ratio of actual doses to required doses. This ratio always ranges between zero and one, thus enabling horizontal comparison of adherence among users on different treatment pathways. This ensures that regardless of the medication form used, the medication adherence indicator reflects the consistency of patients' adherence to medical orders, and the normalized ratio eliminates the influence of differences in absolute frequency.

[0055] The Blood Glucose Monitoring Standardization Index (GQSI) is used to replace simple statistics of monitoring days, providing a more accurate reflection of the quality of monitoring behavior. The index is calculated in two steps: First, based on the user's medical orders, a set of key monitoring times for each day is determined, such as fasting, two hours after breakfast, two hours after lunch, two hours after dinner, and bedtime. Second, for each day within the statistical period, the system calculates the degree of match between the actual monitoring time and the recommended monitoring time. If the deviation between the actual monitoring time and the recommended time is within a preset tolerance range (±30 minutes), the monitoring at that time is considered qualified. The proportion of qualified monitoring times to the total number of recommended monitoring times is the standardized monitoring score for that day. The average of the standardized monitoring scores for all days within the period yields the final GQSI. The theoretical basis for this index is that non-standard monitoring times lead to incomparable blood glucose data, thus losing their clinical value in guiding treatment adjustments. Therefore, the effectiveness of monitoring behavior depends not only on frequency but also on the accuracy of the monitoring time. The standardization index can penalize both insufficient monitoring frequency and deviation from the recommended monitoring time, making it more clinically relevant than simply counting the number of days. The exercise achievement rate is calculated by dividing the number of days with achieved exercise goals within the statistical period by the total number of days in the statistical period.

[0056] The pre-defined weighting coefficients in the formula assign different levels of importance to medication, monitoring, and exercise behaviors in the overall adherence assessment. These weighting coefficients can be determined using multiple linear regression or ridge regression analysis: using the three components (IIF / EIF, BMD, ESD) in the treatment adherence index calculation formula as independent variables and the decrease in HbA1c over the past six months as the dependent variable, multiple linear regression or ridge regression analysis is performed to quantify the contribution of each behavior to the clinical outcome. The obtained regression coefficients are then standardized and can be used as... , , The objective estimate.

[0057] This formula standardizes and integrates three different types and scales of behavioral data into a single, comparable comprehensive indicator, thus providing a key quantitative input for subsequent risk assessment models.

[0058] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0059] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0060] Figure 2 A schematic diagram of the structure of the diabetic neuropathy risk assessment device provided in an embodiment of the present invention is shown. For ease of explanation, only the parts related to the embodiment of the present invention are shown, and are described in detail below: like Figure 2 As shown, the diabetic neuropathy risk assessment device 2 includes: Module 21 is used to acquire the neuropathological physiological data and treatment behavior data of the target user; Calculation module 22 is used to calculate the treatment compliance index of the target user based on treatment behavior data; The assessment module 23 is used to input treatment compliance indicators and key physiological data of neuropathy into a trained neuropathy early warning model to obtain the neuropathy risk probability of the target user; wherein, the neuropathy early warning model is constructed based on the relationship between the characteristics of neuropathy in the fundus and foot and the severity of neuropathy.

[0061] In one possible implementation, the neuropathy early warning model includes a fundus neuropathy sub-model, a foot neuropathy sub-model, and a fusion assessment model; the assessment module 23 is specifically used for: By inputting treatment adherence indicators and key physiological data of neuropathy into the fundus neuropathy sub-model, the risk probability of fundus neuropathy for the target user can be obtained. By inputting treatment adherence indicators and key physiological data of neuropathy into the foot neuropathy sub-model, the risk probability of foot neuropathy for the target user can be obtained. By inputting the risk probabilities of fundus neuropathy and foot neuropathy into the fusion assessment model, the comprehensive neuropathy risk probability of the target user is obtained.

[0062] In one possible implementation, the sub-model of fundus neuropathy is as follows:

[0063] in, This represents the probability of retinal nerve damage. The ratio of urine albumin to creatinine. The grade of lesions in fundus imaging. As an indicator of treatment adherence, , , Preset weights.

[0064] In one possible implementation, the foot neuropathy sub-model is as follows:

[0065] in, The probability of foot nerve damage. As a vibration perception index, The ratio of urine albumin to creatinine. As an indicator of treatment adherence, , , Preset weights.

[0066] In one possible implementation, the evaluation module 23 is also used for: Before inputting treatment adherence indicators and key physiological data of neuropathy into the trained neuropathy early warning model to obtain the neuropathy risk probability of the target user, a first training sample set is obtained; wherein, each training sample in the first training sample set includes a user's treatment adherence indicators, key physiological data of neuropathy, and the user's fundus neuropathy risk probability label and / or foot neuropathy risk probability label. Based on the first training sample set, the weights of the fundus neuropathy sub-model and the foot neuropathy sub-model are iteratively optimized.

[0067] In one possible implementation, the evaluation module 23 is also used for: Before inputting the risk probabilities of fundus neuropathy and foot neuropathy into the fusion assessment model, a second training sample set is obtained; wherein, each training sample in the second training sample set includes the risk probabilities of fundus neuropathy and foot neuropathy for a user, as well as the comprehensive neuropathy severity label for that user; The initial fusion evaluation model is trained based on the second training sample set to obtain the trained fusion evaluation model.

[0068] In one possible implementation, the formula for calculating treatment adherence indicators is:

[0069] in, As an indicator of treatment adherence, As an indicator of medication adherence, For standardized blood glucose monitoring index, To improve the fitness achievement rate, , , Preset weights.

[0070] This invention integrates physiological data of neuropathy with treatment behavior data, overcoming the shortcomings of single-dimensional assessment in existing technologies. The introduction of treatment compliance indicators quantifies the quality of intervention implementation, accurately reflects its impact on disease progression, and improves the comprehensiveness of the assessment. The neuropathy early warning model constructed based on the relationship between fundus and foot nerve lesion characteristics and lesion severity can uncover the correlation patterns of lesions in multiple sites. Compared with single-site assessment, it is more in line with clinical practice, making the risk probability output more accurate and providing more reliable quantitative references for clinical practice. This helps to identify high-risk groups early, adjust intervention plans in a timely manner, reduce the probability of serious complications, and balance the scientific nature and practicality of the assessment.

[0071] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. For example... Figure 3 As shown, the electronic device 3 of this embodiment includes a processor 30 and a memory 31. The memory 31 stores a computer program 32. When the processor 30 executes the computer program 32, it implements the steps in the various method embodiments described above. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each module / unit in the various device embodiments described above.

[0072] For example, computer program 32 may be divided into one or more modules / units, which are stored in memory 31 and executed by processor 30 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 32 in electronic device 3.

[0073] Electronic device 3 may include, but is not limited to, processor 30 and memory 31. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 3 and does not constitute a limitation on electronic device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 3 may also include input / output devices, network access devices, buses, etc.

[0074] The processor 30 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0075] The memory 31 can be an internal storage unit of the electronic device 3, such as a hard disk or memory of the electronic device 3. The memory 31 can also be an external storage device of the electronic device 3, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 3. Furthermore, the memory 31 can include both internal and external storage units of the electronic device 3. The memory 31 is used to store the computer program 32 and other programs and data required by the electronic device 3. The memory 31 can also be used to temporarily store data that has been output or will be output.

[0076] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.

[0077] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the methods described in the above-described method embodiments.

[0078] This invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the methods described in the above-described method embodiments.

[0079] Computer programs include computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. Computer-readable media can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0080] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0081] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for assessing the risk of diabetic neuropathy, characterized in that, include: Acquire physiological data on neuropathic diseases and treatment behavior data of target users; Based on the treatment behavior data, the treatment compliance index of the target user is calculated; The treatment compliance indicators and the key physiological data of the neuropathy are input into the trained neuropathy early warning model to obtain the neuropathy risk probability of the target user; wherein, the neuropathy early warning model is constructed based on the relationship between the characteristics of neuropathy in the fundus and foot and the severity of neuropathy.

2. The method for assessing the risk of diabetic neuropathy according to claim 1, characterized in that, The neuropathy early warning model includes a fundus neuropathy sub-model, a foot neuropathy sub-model, and a fusion assessment model; the step of inputting the treatment compliance indicators and the key physiological data of the neuropathy into the trained neuropathy early warning model to obtain the neuropathy risk probability of the target user includes: The treatment compliance indicators and the key physiological data of the neuropathy are input into the fundus neuropathy sub-model to obtain the fundus neuropathy risk probability of the target user; The treatment compliance indicators and the key physiological data of the neuropathy are input into the foot neuropathy sub-model to obtain the risk probability of foot neuropathy for the target user. The probability of risk of fundus neuropathy and the probability of risk of risk of foot neuropathy are input into the fusion assessment model to obtain the comprehensive neuropathy risk probability of the target user.

3. The method for assessing the risk of diabetic neuropathy according to claim 2, characterized in that, The sub-model of fundus neuropathy is as follows: in, This represents the probability of retinal nerve damage. The ratio of urine albumin to creatinine. The grade of lesions in fundus imaging. As an indicator of treatment adherence, , , Preset weights.

4. The method for assessing the risk of diabetic neuropathy according to claim 2, characterized in that, The foot nerve lesion sub-model is as follows: in, The probability of foot nerve damage. As a vibration perception index, The ratio of urine albumin to creatinine. As an indicator of treatment adherence, , , Preset weights.

5. The method for assessing the risk of diabetic neuropathy according to claim 3 or 4, characterized in that, Before inputting the treatment adherence indicators and key physiological data of neuropathy into the trained neuropathy early warning model to obtain the neuropathy risk probability of the target user, the method further includes: Obtain a first training sample set; wherein each training sample in the first training sample set includes a user's treatment compliance index, key physiological data of neuropathy, and the user's fundus neuropathy risk probability label and / or foot neuropathy risk probability label; Based on the first training sample set, the weights of the fundus neuropathy sub-model and the foot neuropathy sub-model are iteratively optimized.

6. The method for assessing the risk of diabetic neuropathy according to claim 2, characterized in that, Before inputting the risk probabilities of fundus neuropathy and foot neuropathy into the fusion assessment model, the method further includes: Obtain a second training sample set; wherein each training sample in the second training sample set includes a user's risk probability of fundus neuropathy, risk probability of foot neuropathy, and the comprehensive neuropathy severity label corresponding to the user; The initial fusion evaluation model is trained based on the second training sample set to obtain the trained fusion evaluation model.

7. The method for assessing the risk of diabetic neuropathy according to any one of claims 1 to 5, characterized in that, The formula for calculating treatment adherence indicators is: in, As an indicator of treatment adherence, As an indicator of medication adherence, For standardized blood glucose monitoring index, To improve the fitness achievement rate, , , Preset weights.

8. A device for assessing the risk of diabetic neuropathy, characterized in that, include: The acquisition module is used to acquire the neuropathological physiological data and treatment behavior data of the target user; The calculation module is used to calculate the treatment compliance index of the target user based on the treatment behavior data; An assessment module is used to input the treatment compliance indicators and the key physiological data of the neuropathy into a trained neuropathy early warning model to obtain the neuropathy risk probability of the target user; wherein, the neuropathy early warning model is constructed based on the relationship between the characteristics of neuropathy in the fundus and foot and the severity of neuropathy.

9. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images