A diabetes chronic complication early warning method, device, medium and computer equipment

Abstract

The application provides a diabetes chronic complication early warning method, device, medium and computer equipment, the early warning method comprises: obtaining the body data of a diabetes patient; filtering the body data by using a LASSO logistic regression model to generate a data training set; training a multi-layer perception model by using the data training set to obtain a diabetes complication early warning model; obtaining the body data to be detected, inputting the filtered body data into the diabetes complication early warning model for prediction to generate an early warning result; by using the method of the application, the collected body data is reduced in dimension by using the LASSO logistic regression model, the training data of the model can be effectively reduced, the computing power requirement of the training and reasoning process is reduced, the model can be conveniently deployed to a local computing end, so that the diabetes patient can realize home detection, and the prediction method provided in the application can enable the diabetes patient to realize non-invasive detection, and compared with conventional blood detection, has the advantages of non-invasiveness and small operation difficulty.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical technology, and more specifically, to a method, device, medium, and computer equipment for early warning of chronic complications of diabetes. Background Technology

[0002] Currently, the early diagnosis and prediction of chronic complications of diabetes largely rely on modern medical theories, using various biomarkers and imaging examinations for detection. Traditional models for predicting diabetic complications mainly focus on intervention and prevention through invasive blood tests, blood glucose monitoring, and blood pressure monitoring. However, while blood glucose monitoring and glycated hemoglobin (HbA1c) testing are important for diabetes management, these methods typically require frequent blood tests, which are invasive, expensive, and inconvenient for long-term home self-monitoring.

[0003] To address the aforementioned shortcomings, a non-invasive method for early warning of diabetic complications is proposed, which is relatively easy to perform and allows diabetic patients to conduct tests at home. This is of great importance to diabetic patients. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a method, device, medium and computer equipment for early warning of chronic complications of diabetes, so as to overcome the disadvantages of existing diabetes complication detection which requires blood tests, is invasive, has high testing costs and cannot be performed at home.

[0005] The above-mentioned technical objective of the present invention is achieved through the following technical solution:

[0006] Firstly, a method for early warning of chronic complications of diabetes includes:

[0007] S1. Obtain the body data of diabetic patients; the body data includes at least: the meridian conduction data of twelve pairs of Jing-Well points and the meridian conduction data of twelve pairs of Yuan-Source points;

[0008] S2. The body data is filtered using the LASSO logistic regression model to generate a data training set; the multilayer sensing model is trained using the data training set to obtain a diabetes complication early warning model.

[0009] S3. Obtain the physical data of the person to be tested, and after screening, input it into the diabetes complication early warning model for prediction;

[0010] S4. Generate an early warning result, specifically the probability of the occurrence of chronic complications of diabetes.

[0011] In one embodiment, obtaining the physical data of a diabetic patient specifically includes:

[0012] Obtain first meridian conductance data from diabetic patients without complications;

[0013] Acquire second meridian conductance data from patients with diabetic peripheral neuropathy.

[0014] In one embodiment, the first meridian conductivity data and the second meridian conductivity data respectively include:

[0015] Conductivity data for twelve pairs of wells;

[0016] Electrical conductivity data of twelve pairs of acupoints;

[0017] The ratio of the sum of electrical conductivity data of the twelve primary acupoints on the left side to the sum of electrical conductivity data of the twelve primary acupoints on the right side;

[0018] The ratio of the sum of the electrical conductivity data of the twelve wells on the left to the sum of the electrical conductivity data of the twelve wells on the right;

[0019] The ratio of the sum of electrical conductivity data of the six pairs of primary acupoints on the hand to the sum of electrical conductivity data of the six pairs of primary acupoints on the foot;

[0020] The ratio of the total electrical conductivity data of the six pairs of well points on the hand to the total electrical conductivity data of the six pairs of well points on the foot;

[0021] The ratio of the sum of electrical conductivity data of the six pairs of primary acupoints of the Yin meridians to the sum of electrical conductivity data of the six pairs of primary acupoints of the Yang meridians;

[0022] The ratio of the total electrical conductivity data of the six pairs of well points on the Yin meridians to the total electrical conductivity data of the six pairs of well points on the Yang meridians.

[0023] In one embodiment, the step of using a LASSO logistic regression model to filter the body data and generate a data training set specifically includes:

[0024] Constructing the feature matrix ;

[0025] in, Indicates the number of patients. Indicates the number of features; Indicates the first The first patient's The numerical value of the feature item;

[0026] The feature matrix was determined using the LASSO logistic regression model. The importance of each feature in the model is determined, and the feature coefficients of each feature are adjusted according to their importance. The feature coefficients include zero coefficients and non-zero coefficients.

[0027] A data training set is constructed using the feature terms whose feature coefficients are non-zero.

[0028] In one embodiment, the feature matrix is ​​determined using the LASSO logistic regression model. The importance of each feature is determined, and the feature coefficients of each feature are adjusted according to their importance. These feature coefficients include zero coefficients and non-zero coefficients, specifically including:

[0029] Construct a logistic regression model with LASSO regularization:

[0030] ;

[0031] Determine the objective function of the LASSO logistic regression model:

[0032] ;

[0033] Define the set of candidate parameters The candidate parameter set includes Candidate parameters , ;

[0034] For each of the candidate parameters Each candidate parameter is obtained by performing 10-fold cross-validation using the body data. Average cross-validation performance metrics ;

[0035] Select the optimal parameters ,include:

[0036] ;

[0037] Using the optimal parameters The LASSO logistic regression model was retrained based on all the aforementioned body data, and the coefficients of each feature were adjusted and compressed into zero and non-zero coefficients based on the training results, including: .

[0038] In one embodiment, for each of the candidate parameters Each candidate parameter is obtained by performing 10-fold cross-validation using the body data. Average cross-validation performance metrics Specifically, it includes:

[0039] The body data Randomly divided into 10 subsets The difference in the number of samples contained in any two subsets is less than a predetermined threshold.

[0040] For each of the candidate parameters Select a subset in turn The remaining nine subsets will be used as the training set, serving as the validation set.

[0041] After training the LASSO logistic regression model using the training set, the classification performance index of the trained LASSO logistic regression model is calculated using the validation set corresponding to the training set, thus obtaining the candidate parameters. The corresponding ten performance metrics:

[0042] ;

[0043] Calculate each of the candidate parameters The average cross-validation performance metrics include:

[0044] .

[0045] In one embodiment, features with non-zero coefficients specifically include: electrical conductivity data of the left Lidui acupoint, electrical conductivity data of the right Zhongchong acupoint, electrical conductivity data of the left Chongyang acupoint, the ratio of the sum of electrical conductivity data of the twelve primary acupoints on the left to the sum of electrical conductivity data of the twelve primary acupoints on the right, electrical conductivity data of the right Yongquan acupoint, electrical conductivity data of the left Hegu acupoint, electrical conductivity data of the left Taichong acupoint, electrical conductivity data of the left Daling acupoint, the ratio of the sum of electrical conductivity data of the six pairs of primary acupoints of the Yin meridians to the sum of electrical conductivity data of the six pairs of primary acupoints of the Yang meridians, electrical conductivity data of the left Dadun acupoint, and electrical conductivity data of the left Yangchi acupoint.

[0046] A device for early warning of chronic complications of diabetes, comprising:

[0047] The acquisition unit is used to acquire the body data of diabetic patients; the body data includes at least: the meridian conduction data of twelve pairs of well points and the meridian conduction data of twelve pairs of original points;

[0048] The training unit is used to filter the body data using the LASSO logistic regression model to generate a data training set; and to train the multilayer perception model using the data training set to obtain a diabetes complication early warning model.

[0049] The prediction unit is used to acquire the physical data of the person to be tested, and after screening, input it into the diabetes complication early warning model for prediction.

[0050] The early warning unit is used to generate an early warning result, specifically the probability of the occurrence of chronic complications of diabetes.

[0051] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0052] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described above.

[0053] In summary, the present invention has the following beneficial effects: a method, device, medium, and computer equipment for early warning of chronic complications of diabetes. The early warning method includes: acquiring the body data of a diabetic patient; the body data includes at least: meridian conduction data of twelve pairs of well points and meridian conduction data of twelve pairs of original points; using a LASSO logistic regression model to filter the body data and generate a data training set; using the data training set to train the multilayer sensing model to obtain a diabetes complication early warning model; acquiring the body data of the person to be tested, filtering it, and inputting it into the diabetes complication early warning model for prediction; generating an early warning result. The specific results are: the probability of the occurrence of chronic complications of diabetes; using the method of this invention, the collected body data is reduced in dimensionality through the LASSO logistic regression model, which can effectively reduce the training data of the model and reduce the computing power requirements for the training and inference process, so that the model can be easily deployed to the local computing end, so that diabetic patients can carry out home testing. Moreover, the prediction method provided in this application can enable diabetic patients to carry out non-invasive testing. It only requires measuring the meridian conductivity values ​​of the twelve Jing-Well points and twelve Yuan-Source points of diabetic patients to determine the risk of complications of diabetic patients. Compared with conventional blood tests, it has the advantages of being non-invasive and easy to operate. Attached Figure Description

[0054] Figure 1 This is a flowchart of a method for early warning of chronic complications of diabetes according to the present invention;

[0055] Figure 2 This is a structural diagram of the early warning device for chronic complications of diabetes in an embodiment of the present invention;

[0056] Figure 3 This is an internal structural diagram of the computer device in an embodiment of the present invention;

[0057] Figure 4 This is a schematic diagram of the cross-validation results of the LASSO regression in this invention;

[0058] Figure 5 This is a diagram showing the LASSO regression coefficient path of this invention;

[0059] In the diagram: 1. Acquisition unit; 2. Training unit; 3. Prediction unit; 4. Early warning unit. Detailed Implementation

[0060] To make the objectives, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Several embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein.

[0061] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, the simultaneous existence of A and B, or the existence of B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0062] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0063] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0064] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part 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 to cause a device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0065] The above description is merely a specific embodiment of this application. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application. The protection scope of this application should be determined by the protection scope of the claims.

[0066] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0067] Example 1

[0068] To facilitate understanding of the technical solution of this application, the relevant technologies in this application will be described first.

[0069] Diabetic complications are multi-organ damage caused by long-term high blood sugar, mainly divided into two categories: acute and chronic complications. Chronic complications often present with insidious symptoms in their early stages, leading patients to ignore even minor symptoms and resulting in irreversible damage. Common chronic complications of diabetes include diabetic retinopathy, diabetic nephropathy, and diabetic neuropathy. Prevention and treatment of chronic diabetic complications requires blood sugar control, multi-factor management, and regular screening. During regular screening, it is generally recommended that diabetic patients undergo follow-up screenings at a predetermined interval (e.g., every six months or three months) to ensure early detection and intervention of complications, avoiding delays in treatment and more serious consequences. However, the above-mentioned regular screening process still has the following drawbacks: during the time between screenings, patients cannot clearly understand their own status, cannot monitor their condition at home, and multiple blood draws can cause trauma to the patient's body, affecting their quality of life.

[0070] In response to this situation, some studies have utilized non-invasive bioelectrical signal measurement methods to predict chronic complications of diabetes. The Digital Meridian System (DMS) measures the resistance values ​​of twelve acupoints on each of the twelve meridians on the left and right sides of the body, converting the resistance into a meridian energy index to reflect the state of the meridians and assist in traditional Chinese medicine diagnosis and treatment. However, existing DMS meridian systems generally suffer from insufficient sensitivity, significant susceptibility to environmental factors in measurement results, and a lack of specificity for the early diagnosis of diabetic complications, making it difficult to directly provide early warnings and assessments of diabetic complications.

[0071] In Traditional Chinese Medicine (TCM), existing research has indicated that the "collateral disease theory" has significant guiding significance for the diagnosis and treatment of diabetes, suggesting that chronic complications of diabetes are mainly caused by "internal obstruction of the collaterals." Based on this theory, TCM can alleviate the risk of diabetic complications to some extent by regulating and unblocking the body's meridians. However, traditional TCM meridian detection methods are limited by the complexity and time-consuming nature of manual operation, and lack precise quantitative measurement methods, making it difficult to achieve rapid, objective, and continuous monitoring.

[0072] To address the above problems, this invention provides a method for early warning of chronic complications of diabetes, such as... Figure 1 As shown, it includes:

[0073] S1. Obtain the body data of diabetic patients; the body data includes at least: the meridian conduction data of twelve pairs of Jing-Well points and the meridian conduction data of twelve pairs of Yuan-Source points;

[0074] The physical examinations of diabetic patients were collected from the diabetes chronic disease management clinic and the endocrinology inpatient department. To ensure the accuracy of the test data, diabetic patients needed to be screened, specifically including the following conditions: 1. The diabetic patient was 18 years of age or older; 2. Informed consent was signed; 3. The diabetic patient met the diagnostic criteria for type 2 diabetes, which are one of the following four conditions: ① Glycated hemoglobin ≥ 6.5%; ② Fasting blood glucose ≥ 7.0 mmol / L; ③ 2-hour blood glucose ≥ 11.1 mmol / L during an oral glucose tolerance test; ④ In patients with typical hyperglycemia or hyperglycemic crisis symptoms, random blood glucose ≥ 11.1 mmol / L; 4. The diabetic patient met the criteria for chronic complications, specifically including at least one of the following three complications: diabetic nephropathy, diabetic retinopathy, and diabetic peripheral neuropathy.

[0075] To ensure the accuracy of the measurement data, it is necessary to exclude patients who cannot be measured, mainly those who cannot use the electrical conduction meridian assessment method. Specifically, this includes: ① those with implanted pacemakers; ② those with implanted metal parts; ③ those with limb defects; ④ those with local skin damage at the measurement site; ⑤ those with local skin scarring at the measurement site; ⑥ those with comprehension difficulties and unable to communicate; and ⑦ those who are planning to become pregnant, are pregnant, or are breastfeeding.

[0076] For the measurement of the patient's conductance data, the traditional Chinese medicine meridian detector produced and sold by Sichuan Chongzhou Jinhai Medical Device Co., Ltd., model JH / DMS-K100, medical device registration number: Sichuan Medical Device Approval No. 20172270231, is used to measure the meridian conductance data of diabetic patients. The instrument detection environment and operation are in accordance with the standard of the Chinese Association of Chinese Medicine, "Technical Operation Specification for Preventive Treatment of Disease in Traditional Chinese Medicine - Conductance Method Acupoint Evaluation" (T / CACM 1089-2018) issued in September 2018. And according to the meridian calculation method in the standard, the meridian values of each acupoint of the Jing-well points are calculated.

[0077] In the human body, the Jing-well points are one of the five transporting points (Jing-well, Ying-spring, Shu-stream, Jing-channel, He-sea), located at the extremities of the four limbs (fingers or toes). There are six Jing-well points on the hand, namely Shao Shang (Lung Meridian), Shang Yang (Large Intestine Meridian), Zhong Chong (Pericardium Meridian), Guan Chong (Triple Burner Meridian), Shao Chong (Heart Meridian), Shao Ze (Small Intestine Meridian); there are six Jing-well points on the foot, namely: Yin Bai (Spleen Meridian), Li Dui (Stomach Meridian), Da Dun (Liver Meridian), Zu Qiao Yin (Gallbladder Meridian), Zhi Yin (Bladder Meridian), Yong Quan (Kidney Meridian). The distribution law of the Jing-well points corresponds to the zang-fu organs, and it is widely used in clinical practice. It is an important means for traditional Chinese medicine to "prevent disease" and for first aid. Yuan-primary points are the acupoints where the original qi of the zang-fu organs is infused, passes through, and stays in the extremities of the twelve meridians. There are six Yuan-primary points on the hand, namely: Tai Yuan (Lung Meridian of Hand-Taiyin), He Gu (Large Intestine Meridian of Hand-Yangming), Shen Men (Heart Meridian of Hand-Shaoyin), Da Ling (Pericardium Meridian of Hand-Jueyin), Wan Gu (Small Intestine Meridian of Hand-Taiyang), Yang Chi (Triple Burner Meridian of Hand-Shaoyang); there are six Yuan-primary points on the foot, namely: Chong Yang (Stomach Meridian of Foot-Yangming), Tai Bai (Spleen Meridian of Foot-Taiyin), Tai Xi (Kidney Meridian of Foot-Shaoyin), Jing Gu (Bladder Meridian of Foot-Taiyang), Qiu Xu (Gallbladder Meridian of Foot-Shaoyang), Tai Chong (Liver Meridian of Foot-Jueyin). By measuring the Jing-well points and Yuan-primary points on the left hand, right hand, left foot, and right foot, the meridian conductance data of twelve pairs of Jing-well points and twelve pairs of Yuan-primary points can be obtained.

[0078] Based on the above data acquisition method, a total of 120 cases were collected in this embodiment, including 91 cases of diabetes without complications and 29 cases of diabetes with complications. In terms of gender distribution, there were 57 females in the non-complication group, accounting for 62.64%, and 16 females in the complication group, accounting for 55.17%. There was no statistically significant difference in the gender composition ratio between the groups ( The median ages were 59.00 and 62.00 years, respectively (P=0.41), indicating a similar age distribution between the two groups and ruling out the influence of age on the results. The median BMIs were 23.44 and 23.31, respectively (P=0.54), indicating similar obesity levels between the two groups, and BMI did not significantly affect the risk of complications. No statistical significance was found between the two groups. In the analysis of blood glucose indicators, there were no statistically significant differences between the groups in the lowest (P=0.19), highest (P=0.09) and lowest (P=0.95), highest (P=0.46) fasting blood glucose values. The results suggest that the complication group and the non-complication group are homogeneous in terms of demographic characteristics and baseline blood glucose indicators.

[0079] Table 1 Comparison of gender composition ratio between the two groups of patients

[0080]

[0081] Table 2 Comparison of basic information between the two groups of patients

[0082]

[0083] In Table 2, N represents the number of samples; The values ​​represent the mean and standard deviation; M represents the median, applicable to non-normally distributed data, representing the middle position; Q75-25 represents the interquartile range, showing the range and dispersion of the data distribution; Z-value is a standardized indicator used in statistical tests; and P-value represents the probability of observing the current data or more extreme data under the null hypothesis (that there is no difference between the two groups of data).

[0084] By measuring the electrical conductivity values ​​of acupoints in patients with known diabetic complications and labeling each value, the data can be used as training data for model prediction. Based on the clinical diagnostic criteria for chronic diabetic complications, subjects were categorized into four groups: 1. Non-complication group; 2. Diabetic retinopathy group; 3. Diabetic nephropathy group; 4. Diabetic peripheral neuropathy group. Patient data includes multiple components, such as basic demographic information including subject ID, age, gender, duration of diabetes, and diagnosis date. Furthermore, based on the research content of this application, the data also includes 48 electrical conductivity values ​​from twelve pairs of acupoints and twelve pairs of primary acupoints for each patient, as well as the ratios between acupoint electrical conductivity values. Since the collected data from patients with diabetic complications primarily consisted of those with diabetic peripheral neuropathy, with fewer patients in the other two groups (diabetic retinopathy and diabetic nephropathy), this embodiment focuses solely on analyzing the electrical conductance data of patients with diabetic peripheral neuropathy. Specifically, acquiring the body data of diabetic patients includes: acquiring the first meridian electrical conductance data of diabetic patients without complications; and acquiring the second meridian electrical conductance data of patients with diabetic peripheral neuropathy. Specifically, the first and second meridian electrical conductance data in this embodiment respectively include the following data:

[0085] The electrical conductivity data of twelve pairs of wells, totaling 24 data points, are represented in a matrix as follows: ;

[0086] The electrical conductivity data of the twelve pairs of primary acupoints, totaling 24 data points, are represented in the matrix as follows: ;

[0087] The ratio of the sum of the electrical conductivity data of the twelve primary acupoints on the left to the sum of the electrical conductivity data of the twelve primary acupoints on the right is represented in the matrix as follows: ;

[0088] The ratio of the sum of the conductivity data of the twelve wells on the left to the sum of the conductivity data of the twelve wells on the right is represented in the matrix as follows: ;

[0089] The ratio of the sum of electrical conductivity data of the six pairs of primary acupoints on the hand to the sum of electrical conductivity data of the six pairs of primary acupoints on the foot is represented in the matrix as follows: ;

[0090] The ratio of the sum of electrical conductivity data of the six pairs of well points on the hand to the sum of electrical conductivity data of the six pairs of well points on the foot is represented in the matrix as follows: ;

[0091] The ratio of the sum of the electrical conductivity data of the six pairs of primary acupoints of the Yin meridians to the sum of the electrical conductivity data of the six pairs of primary acupoints of the Yang meridians is represented in the matrix as follows: ;

[0092] The ratio of the sum of electrical conductivity data of the six pairs of Jing-Well points on Yin meridians to the sum of electrical conductivity data of the six pairs of Jing-Well points on Yang meridians is represented in the matrix as follows: ;

[0093] The data collection standards mentioned above can be found in the standard "Technical Operation Specifications for Disease Prevention and Treatment in Traditional Chinese Medicine - Electroconduction Method for Acupoint Evaluation" (T / CACM 1089-2018) issued by the China Association of Traditional Chinese Medicine in September 2018.

[0094] Based on the above data, construct the feature matrix. ;

[0095] The horizontal feature is the number of patients; in this embodiment, the number of patients... The vertical features are feature categories. In this embodiment, the number of feature categories is... .

[0096] In the actual measurement process, it is also necessary to classify the patient's data according to the patient's clinical criteria, and add labels to the patient's data according to each category in order to facilitate model training. The data mainly includes: obtaining the first meridian conductance data of diabetic patients without complications; and obtaining the second meridian conductance data of patients with diabetic peripheral neuropathy.

[0097] S2. The body data is filtered using the LASSO logistic regression model to generate a data training set; the multilayer sensing model is trained using the data training set to obtain a diabetes complication early warning model.

[0098] Specifically, the LASSO logistic regression model can adjust the coefficients of each feature based on their importance. It adjusts the coefficients of unimportant features to zero and retains the coefficients of important features by making them non-zero. The specific process is as follows:

[0099] First, construct a logistic regression model that includes LASSO regularization:

[0100] ;

[0101] in, Represents the feature vector of a given sample When, the sample belongs to category The conditional probability, This represents the intercept term, which is the term used when all... When both are 0, it is the baseline value of the logarithmic probability of the event occurring. Let be the regression coefficient, representing the th . The influence weight of each feature on the target variable is such that the larger the absolute value, the higher the importance of the feature. Through the regularization effect of the LASSO logistic regression model, the coefficients of unimportant features can be compressed to 0, thus achieving feature selection.

[0102] Then, determine the objective function of the LASSO logistic regression model:

[0103] ;

[0104] Define the set of candidate parameters The candidate parameter set includes Candidate parameters , The candidate parameters are predefined and are usually taken as values ​​with a uniform logarithmic distribution. In this embodiment, they are specifically: ;

[0105] For each of the candidate parameters Each candidate parameter is obtained by performing 10-fold cross-validation using the body data. Average cross-validation performance metrics ;

[0106] Tenfold cross-validation mainly includes the following steps:

[0107] All body data D are randomly divided into ten subsets of similar size. That is, the difference between the number of samples contained in any two subsets is less than a predetermined threshold, for example, the predetermined threshold can be 1 or 2;

[0108] For each of the candidate parameters Select a subset in turn The remaining nine subsets were used as the training set, serving as the validation set; that is, the LASSO logistic regression model uses each candidate parameter... In all cases, ten training iterations are performed, and ten validation iterations are required after training to determine the classification accuracy of the model.

[0109] After training the LASSO logistic regression model using the training set, the classification performance index of the trained LASSO logistic regression model is calculated using the validation set corresponding to the training set, thus obtaining the candidate parameters. The corresponding ten performance metrics:

[0110] ;

[0111] Calculate each of the candidate parameters The average cross-validation performance metrics include:

[0112] ;

[0113] Select the optimal parameters ,include:

[0114] ;

[0115] Using the optimal parameters The LASSO logistic regression model was retrained based on all the aforementioned body data, and the coefficients of each feature were adjusted and compressed into zero and non-zero coefficients based on the training results, including:

[0116] .

[0117] Constructing a data training set using feature terms with non-zero coefficients specifically includes: .

[0118] The features with non-zero coefficients selected based on the LASSO logistic regression model are shown in Table 3 below. There are a total of 11 features, namely: the electrical conductivity data of the left Lidui acupoint, the electrical conductivity data of the right Zhongchong acupoint, the electrical conductivity data of the left Chongyang acupoint, the ratio of the sum of the electrical conductivity data of the twelve primary acupoints on the left to the sum of the electrical conductivity data of the twelve primary acupoints on the right, the electrical conductivity data of the right Yongquan acupoint, the electrical conductivity data of the left Hegu acupoint, the electrical conductivity data of the left Taichong acupoint, the electrical conductivity data of the left Daling acupoint, the ratio of the sum of the electrical conductivity data of the six pairs of primary acupoints of the Yin meridians to the sum of the electrical conductivity data of the six pairs of primary acupoints of the Yang meridians, the electrical conductivity data of the left Dadun acupoint, and the electrical conductivity data of the left Yangchi acupoint.

[0119] Table 3 Number of variables selected in LASSO regression

[0120]

[0121] like Figure 4 As shown, this embodiment also provides cross-validation results based on LASSO regression, wherein, as Figure 4 The image shows the results of the 10-fold cross-validation, where the X-axis variable is: , representing the regularization parameter The logarithmic value; the Y-axis variable is: Binomial Deviance, which measures the goodness of fit of the binary classification model; the red dotted line represents different... The average cross-validation error at the λ value; the gray error line represents the standard error of the cross-validation error; the top number (53,53,53,52…0) represents the number of non-zero coefficients (i.e., selected variables) retained in the model at the corresponding λ value.

[0122] exist Figure 4 In the middle, from left to right, with As the selected variables increase, the model complexity decreases, but the error initially increases slowly and then rises rapidly. The two vertical dashed lines in the diagram represent this; the left dashed line represents... To produce the minimum cross-validation error Value; the dashed line on the right indicates The simplest model within a minimum error range of one standard error. Values. Each gray line in the graph represents the coefficient of a variable as... The trajectory of change, with most variable coefficients following Increased but compressed to zero, several key variables are at a relatively large level. The coefficient remains non-zero even at certain values; it can be positive (positive correlation) or negative (negative correlation). Through... Figure 5 Choose the optimal one Value, usually or ,pass Figure 5 Identification at the optimal The important variables to be retained are used to determine the number of variables to be included in the final model.

[0123] Through the above process, the data can be dimensionality reduced to obtain a training dataset. The number of features in the training dataset is reduced from 55 to 11, effectively reducing the computational resource requirements during subsequent model training. After dimensionality reduction, the data is input into a pre-built multilayer perceptron model for training, enabling the model to learn and perceive the implicit features of the data. Specifically, the multilayer perceptron model includes: an input layer, where the number of nodes is equal to the number of standard normal distributions in the training dataset; the main task of the input layer is to pass data samples (feature vectors) into the model for processing by subsequent layers. The number of nodes in the input layer is equal to the number of standard normal distributions in the training dataset, i.e., the feature dimension of the data. Due to the dimensionality reduction, the number of nodes in the input layer is relatively small, reducing computational complexity. There are two hidden layers, which extract features from the data received by the input layer and pass the extracted data to the output layer; the hidden layers extract and transform features from the input data, learning implicit features and complex relationships in the data through a combination of neurons and nonlinear activation functions. In this application, the two hidden layers have 128 and 64 nodes respectively. Each neuron in the hidden layers learns and adjusts the feature representation through weights and bias parameters. The output layer, with one node, outputs a binary classification result for early warning of chronic complications of diabetes. Specifically, the output layer uses the Sigmoid activation function. The output layer is used to generate the model's final prediction result.

[0124] In this embodiment, the model training process employs a gradient descent optimization algorithm, and a binary cross-entropy loss function is used to measure the difference between the model's predicted probability distribution and the true label. Cross-entropy loss directly measures how close the model's predicted probability is to the true label, and even with an imbalanced ratio of positive and negative samples, it still provides an accurate error metric with a clear optimization direction. In the binary classification problem of this application, cross-entropy avoids the gradient vanishing problem, making neural network learning more stable. The gradient descent method minimizes the loss function, and by adaptively adjusting the learning rate, it provides fast and stable convergence, reduces training time, and exhibits strong robustness to noise and gradient fluctuations.

[0125] S3. Obtain the physical data of the person to be tested and input it into the diabetes complication early warning model for prediction;

[0126] After the model is trained, in real life, diabetic patients can achieve non-invasive, home-based testing by measuring the meridian conductivity data of the twelve Jing-Well points and twelve Yuan-Source points on their hands and feet. Between each regular monitoring session, patients can use this method and the model trained by this method to predict their own condition, providing an effective means of detecting chronic complications for diabetic patients.

[0127] S4. Generate an early warning result, specifically the probability of the occurrence of chronic complications of diabetes.

[0128] Specifically, in this application, the output layer does not directly output the binary classification result, but outputs the probability of the classification result, which is the probability that the patient will develop chronic complications. When the probability of the patient developing chronic complications is high, the patient will be advised to go to the hospital for examination in order to provide more accurate test results for diabetic patients.

[0129] Table 4. Confusion Matrix Results of Neural Network Model

[0130]

[0131] Based on the training process described above, as shown in Table 4, the overall accuracy, sensitivity, and specificity of the diabetes complication early warning model on the training data were 0.97, 0.86, and 1.00. The confusion matrix results showed that the model correctly identified 91 negative samples and 25 positive samples, while generating 4 errors that incorrectly predicted positive samples as negative. Notably, the model demonstrated good accuracy in predicting negative samples. Although the sensitivity for positive samples was slightly lower (86.21%), its overall balance (F1-Score = 89.29%) still indicates that the model has high clinical practical value.

[0132] Table 5. Confusion matrix results from external validation

[0133]

[0134] As shown in Table 5 above, to evaluate the application value of the constructed model in a real clinical setting, this study collected 30 blinded external validation samples. These 30 patients did not participate in model training or preliminary validation during data collection, ensuring the independence of the external data. After unblinding, it was found that 8 of these 30 diabetic patients had chronic complications. The confusion matrix results showed that among those predicted "no complications," 20 were actually without complications, and 2 were actually with complications; among those predicted "complications," 6 were actually with complications, and 2 were actually without complications. The sensitivity was 75%, specificity 90.9%, accuracy 86.7%, and F1 was approximately 75%.

[0135] Furthermore, by using the aforementioned LASSO logistic regression model to reduce the dimensionality of the data, and then using the dimensionality-reduced data to train the multilayer perceptron model, the training speed can be improved and resource requirements reduced. Simultaneously, during home-based detection, user data can be continuously collected, and the training process of the LASSO logistic regression model can be repeated to further adjust the coefficients of each feature, thereby improving the model's classification accuracy.

[0136] Example 2

[0137] Please see Figure 2 A device for early warning of chronic complications of diabetes, the device comprising:

[0138] Acquisition unit 1 is used to acquire the body data of diabetic patients; the body data includes at least: the meridian conduction data of twelve pairs of well points and the meridian conduction data of twelve pairs of original points;

[0139] Training unit 2 is used to filter the body data using the LASSO logistic regression model to generate a data training set; and to train the multilayer perception model using the data training set to obtain a diabetes complication early warning model.

[0140] Prediction unit 3 is used to acquire the physical data of the person to be tested, and after screening, input it into the diabetes complication early warning model for prediction;

[0141] Early warning unit 4 is used to generate early warning results, specifically the probability of occurrence of chronic complications of diabetes.

[0142] Specific limitations regarding the early warning device for chronic complications of diabetes can be found in the limitations of the early warning method for chronic complications of diabetes mentioned above, and will not be repeated here. Each module in the aforementioned early warning device for chronic complications of diabetes can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0143] Those skilled in the art will understand that Figure 2 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the present application. Specific early warning devices for chronic complications of diabetes may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0144] Example 3

[0145] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for early warning of chronic complications of diabetes as described in Example 1.

[0146] Example 4

[0147] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. When executed by the processor, the computer program implements a method for early warning of chronic complications of diabetes.

[0148] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0149] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps: including:

[0150] S1. Obtain the body data of diabetic patients; the body data includes at least: the meridian conduction data of twelve pairs of Jing-Well points and the meridian conduction data of twelve pairs of Yuan-Source points;

[0151] S2. The body data is filtered using the LASSO logistic regression model to generate a data training set; the multilayer sensing model is trained using the data training set to obtain a diabetes complication early warning model.

[0152] S3. Obtain the physical data of the person to be tested, and after screening, input it into the diabetes complication early warning model for prediction;

[0153] S4. Generate an early warning result, specifically the probability of the occurrence of chronic complications of diabetes.

[0154] In one embodiment, obtaining the physical data of a diabetic patient specifically includes:

[0155] Obtain first meridian conductance data from diabetic patients without complications;

[0156] Acquire second meridian conductance data from patients with diabetic peripheral neuropathy.

[0157] In one embodiment, the first meridian conductivity data and the second meridian conductivity data respectively include:

[0158] Conductivity data for twelve pairs of wells;

[0159] Electrical conductivity data of twelve pairs of acupoints;

[0160] The ratio of the sum of electrical conductivity data of the twelve primary acupoints on the left side to the sum of electrical conductivity data of the twelve primary acupoints on the right side;

[0161] The ratio of the sum of the electrical conductivity data of the twelve wells on the left to the sum of the electrical conductivity data of the twelve wells on the right;

[0162] The ratio of the sum of electrical conductivity data of the six pairs of primary acupoints on the hand to the sum of electrical conductivity data of the six pairs of primary acupoints on the foot;

[0163] The ratio of the total electrical conductivity data of the six pairs of well points on the hand to the total electrical conductivity data of the six pairs of well points on the foot;

[0164] The ratio of the sum of electrical conductivity data of the six pairs of primary acupoints of the Yin meridians to the sum of electrical conductivity data of the six pairs of primary acupoints of the Yang meridians;

[0165] The ratio of the total electrical conductivity data of the six pairs of well points on the Yin meridians to the total electrical conductivity data of the six pairs of well points on the Yang meridians;

[0166] The ratio of the sum of electrical conductivity data of the twelve pairs of primary acupoints to the sum of electrical conductivity data of the twelve pairs of well acupoints.

[0167] In one embodiment, the step of using a LASSO logistic regression model to filter the body data and generate a data training set specifically includes:

[0168] Constructing the feature matrix ;

[0169] in, Indicates the number of patients. Indicates the number of features; Indicates the first The first patient's The numerical value of the feature item;

[0170] The feature matrix was determined using the LASSO logistic regression model. The importance of each feature in the model is determined, and the feature coefficients of each feature are adjusted according to their importance. The feature coefficients include zero coefficients and non-zero coefficients.

[0171] A data training set is constructed using the feature terms whose feature coefficients are non-zero.

[0172] In one embodiment, the feature matrix is ​​determined using the LASSO logistic regression model. The importance of each feature is determined, and the feature coefficients of each feature are adjusted according to their importance. These feature coefficients include zero coefficients and non-zero coefficients, specifically including:

[0173] Construct a logistic regression model with LASSO regularization:

[0174] ;

[0175] Determine the objective function of the LASSO logistic regression model:

[0176] ;

[0177] Define the set of candidate parameters The candidate parameter set includes Candidate parameters , ;

[0178] For each of the candidate parameters Each candidate parameter is obtained by performing 10-fold cross-validation using the body data. Average cross-validation performance metrics ;

[0179] Select the optimal parameters ,include:

[0180] ;

[0181] Using the optimal parameters The LASSO logistic regression model was retrained based on all the aforementioned body data, including:

[0182] ;

[0183] The coefficients of various features in the body data are adjusted using the trained LASSO logistic regression model to obtain zero and non-zero coefficients.

[0184] In one embodiment, for each of the candidate parameters Each candidate parameter is obtained by performing 10-fold cross-validation using the body data. Average cross-validation performance metrics Specifically, it includes:

[0185] The body data Randomly divided into 10 subsets The difference in the number of samples contained in any two subsets is less than a predetermined threshold.

[0186] For each of the candidate parameters Select a subset in turn The remaining nine subsets will be used as the training set, serving as the validation set.

[0187] After training the LASSO logistic regression model using the training set, the classification performance index of the trained LASSO logistic regression model is calculated using the validation set corresponding to the training set, thus obtaining the candidate parameters. The corresponding ten performance metrics:

[0188] ;

[0189] Calculate each of the candidate parameters The average cross-validation performance metrics include:

[0190] .

[0191] In one embodiment, features with non-zero coefficients specifically include: electrical conductivity data of the left Lidui acupoint, electrical conductivity data of the right Zhongchong acupoint, electrical conductivity data of the left Chongyang acupoint, the ratio of the sum of electrical conductivity data of the twelve primary acupoints on the left to the sum of electrical conductivity data of the twelve primary acupoints on the right, electrical conductivity data of the right Yongquan acupoint, electrical conductivity data of the left Hegu acupoint, electrical conductivity data of the left Taichong acupoint, electrical conductivity data of the left Daling acupoint, the ratio of the sum of electrical conductivity data of the six pairs of primary acupoints of the Yin meridians to the sum of electrical conductivity data of the six pairs of primary acupoints of the Yang meridians, electrical conductivity data of the left Dadun acupoint, and electrical conductivity data of the left Yangchi acupoint.

[0192] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0193] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0194] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for early warning of chronic complications of diabetes, characterized in that, include: Acquire physical data from diabetic patients; said physical data includes: first meridian conductance data of diabetic patients without complications, and second meridian conductance data of patients with diabetic peripheral neuropathy. The first meridian conductance data and the second meridian conductance data respectively include: Conductivity data for twelve pairs of wells; Electrical conductivity data of twelve pairs of acupoints; The ratio of the sum of electrical conductivity data of the twelve primary acupoints on the left to the sum of electrical conductivity data of the twelve primary acupoints on the right; The ratio of the sum of the electrical conductivity data of the twelve wells on the left to the sum of the electrical conductivity data of the twelve wells on the right; The ratio of the sum of electrical conductivity data of the six pairs of primary acupoints on the hand to the sum of electrical conductivity data of the six pairs of primary acupoints on the foot; The ratio of the total electrical conductivity data of the six pairs of well points on the hand to the total electrical conductivity data of the six pairs of well points on the foot; The ratio of the sum of electrical conductivity data of the six pairs of primary acupoints of the Yin meridians to the sum of electrical conductivity data of the six pairs of primary acupoints of the Yang meridians; The ratio of the total electrical conductivity data of the six pairs of well points on the Yin meridians to the total electrical conductivity data of the six pairs of well points on the Yang meridians; The body data was filtered using a LASSO logistic regression model to generate a training set. The training set included: electrical conductivity data of the left Lidui acupoint, electrical conductivity data of the right Zhongchong acupoint, electrical conductivity data of the left Chongyang acupoint, the ratio of the sum of electrical conductivity data of the twelve primary acupoints on the left to the sum of electrical conductivity data of the twelve primary acupoints on the right, electrical conductivity data of the right Yongquan acupoint, electrical conductivity data of the left Hegu acupoint, electrical conductivity data of the left Taichong acupoint, electrical conductivity data of the left Daling acupoint, the ratio of the sum of electrical conductivity data of the six pairs of primary acupoints of the Yin meridians to the sum of electrical conductivity data of the six pairs of primary acupoints of the Yang meridians, electrical conductivity data of the left Dadun acupoint, and electrical conductivity data of the left Yangchi acupoint. A diabetes complication early warning model is obtained by training a pre-constructed multilayer sensing model using the aforementioned data training set. The physical data of the individuals to be tested are obtained, filtered, and then input into the diabetes complication early warning model for prediction. Generate an early warning result, specifically the probability of diabetic peripheral neuropathy.

2. The method for early warning of chronic complications of diabetes according to claim 1, characterized in that, The step of using the LASSO logistic regression model to filter the body data and generate a data training set specifically includes: Constructing the feature matrix ; in, Indicates the number of patients. Indicates the number of features; Indicates the first The first patient's The numerical value of the feature item; The feature matrix was determined using the LASSO logistic regression model. The importance of each feature in the model is determined, and the feature coefficients of each feature are adjusted according to their importance. The feature coefficients include zero coefficients and non-zero coefficients. A data training set is constructed using the feature terms whose feature coefficients are non-zero.

3. The method for early warning of chronic complications of diabetes according to claim 2, characterized in that, The feature matrix is ​​determined using the LASSO logistic regression model. The importance of each feature is determined, and the feature coefficients of each feature are adjusted according to their importance. These feature coefficients include zero coefficients and non-zero coefficients, specifically including: Construct a logistic regression model with LASSO regularization: ； in, Represents the feature vector of a given sample When, the conditional probability that the sample belongs to category 1, Represents the intercept term. These are the regression coefficients; Determine the objective function of the LASSO logistic regression model: ； in, For sample categories; For regularization parameters; Define the set of candidate parameters The candidate parameter set includes Candidate parameters , ; For each of the candidate parameters Each candidate parameter is obtained by performing 10-fold cross-validation using the body data. Average cross-validation performance metrics ; Select the optimal parameters ,include: ； Using the optimal parameters The LASSO logistic regression model was retrained based on all the aforementioned body data, and the coefficients of each feature were adjusted and compressed into zero and non-zero coefficients based on the training results, including: .

4. The method for early warning of chronic complications of diabetes according to claim 3, characterized in that, For each of the candidate parameters Each candidate parameter is obtained by performing 10-fold cross-validation using the body data. Average cross-validation performance metrics Specifically, it includes: The body data Randomly divided into 10 subsets The difference in the number of samples contained in any two subsets is less than a predetermined threshold. For each of the candidate parameters Select a subset in turn The remaining nine subsets will be used as the training set, serving as the validation set. After training the LASSO logistic regression model using the training set, the classification performance index of the trained LASSO logistic regression model is calculated using the validation set corresponding to the training set, thus obtaining the candidate parameters. The corresponding ten performance metrics: ； Calculate each of the candidate parameters The average cross-validation performance metrics include: 。 5. A device for early warning of chronic complications of diabetes, characterized in that, The device includes: The acquisition unit is used to acquire physical data of diabetic patients; the physical data includes: first meridian conductance data of diabetic patients without complications, and second meridian conductance data of patients with diabetic peripheral neuropathy. The first meridian conductance data and the second meridian conductance data respectively include: Conductivity data for twelve pairs of wells; Electrical conductivity data of twelve pairs of acupoints; The ratio of the sum of electrical conductivity data of the twelve primary acupoints on the left to the sum of electrical conductivity data of the twelve primary acupoints on the right; The ratio of the sum of the electrical conductivity data of the twelve wells on the left to the sum of the electrical conductivity data of the twelve wells on the right; The ratio of the sum of electrical conductivity data of the six pairs of primary acupoints on the hand to the sum of electrical conductivity data of the six pairs of primary acupoints on the foot; The ratio of the total electrical conductivity data of the six pairs of well points on the hand to the total electrical conductivity data of the six pairs of well points on the foot; The ratio of the sum of electrical conductivity data of the six pairs of primary acupoints of the Yin meridians to the sum of electrical conductivity data of the six pairs of primary acupoints of the Yang meridians; The ratio of the total electrical conductivity data of the six pairs of well points on the Yin meridians to the total electrical conductivity data of the six pairs of well points on the Yang meridians; The training unit is used to filter the body data using a LASSO logistic regression model to generate a data training set. The data training set includes: electrical conductivity data of the left Lidui acupoint, the right Zhongchong acupoint, the left Chongyang acupoint, the ratio of the sum of electrical conductivity data of the twelve primary acupoints on the left to the sum of electrical conductivity data of the twelve primary acupoints on the right, the right Yongquan acupoint, the left Hegu acupoint, the left Taichong acupoint, the left Daling acupoint, the ratio of the sum of electrical conductivity data of the six pairs of primary acupoints of the Yin meridians to the sum of electrical conductivity data of the six pairs of primary acupoints of the Yang meridians, the left Dadun acupoint, and the left Yangchi acupoint. The pre-constructed multilayer sensing model is trained using this data training set to obtain a diabetes complication early warning model. The prediction unit is used to acquire the physical data of the person to be tested, and after screening, input it into the diabetes complication early warning model for prediction. The early warning unit is used to generate an early warning result, specifically the probability of diabetic peripheral neuropathy.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for early warning of chronic complications of diabetes as described in any one of claims 1-4.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method for early warning of chronic complications of diabetes as described in any one of claims 1-4.

Images