An intelligent identification method and system for early-onset type 2 diabetes risk factors based on a deep neural network
Patent Information
- Application Number
- CN202611076500.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-20
- Publication Date
- 2026-10-23
AI Technical Summary
目前大多数研究针对中老年2型糖尿病的病因和防治开展，对年轻人早发2型糖尿病发病风险因素识别研究较少
[0014]经由上述的技术方案可知，与现有技术相比，本发明公开提供了一种基于深度神经网络的早发2型糖尿病发病风险因素智能识别方法及系统。相较于现有逻辑回归等传统分析手段，本发明融合深度神经网络实现多源异构医疗数据一体化处理，大幅提升早发2型糖尿病发病风险因素识别准确率与模型浮点运算利用率，模型准确率均超92%、AUC-ROC值均高于0.91，计算效率达到传统方法近10倍，可精准区分年轻人群中未成年、成年男女群体差异化核心致病风险因素，为年轻人群早发2型糖尿病实现早期风险筛查、针对性临床干预提供可靠量化依据，有效弥补现有技术处理大规模多类型临床数据时识别精度不足、运算效率低下的缺陷，为早发2型糖尿病的有效防控和早诊早治提供新思路。
Smart Images
Figure CN122967965A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the interdisciplinary field of artificial intelligence and the identification of risk factors for diabetes, and more specifically to an intelligent identification method and system for risk factors of early-onset type 2 diabetes based on deep neural networks. Background Technology
[0002] Early-onset type 2 diabetes refers to type 2 diabetes with onset before the age of 40. Compared with type 2 diabetes in middle-aged and elderly individuals, it is characterized by a longer disease course, a higher risk of complications, and more difficult blood sugar control. Early-onset type 2 diabetes is the result of the combined effects of multiple risk factors, such as obesity, hypertension, dyslipidemia, high-calorie diet, lack of exercise, sedentary lifestyle, and family history of diabetes. Currently, most research focuses on the etiology and prevention of type 2 diabetes in middle-aged and elderly individuals, with limited research on identifying risk factors for early-onset type 2 diabetes in young people. Furthermore, existing studies typically employ traditional analytical methods such as logistic regression and multiple regression, which suffer from insufficient accuracy and low computational efficiency when applied to large-scale, multi-type datasets.
[0003] Therefore, how to provide a method and system for intelligent identification of risk factors for early-onset type 2 diabetes based on deep neural networks that can effectively improve the accuracy of identifying risk factors for early-onset type 2 diabetes, facilitate the early identification of risk factors for early-onset type 2 diabetes, and enable targeted and timely intervention for early-onset type 2 diabetes, thus providing new ideas for the effective prevention, control, early diagnosis and treatment of early-onset type 2 diabetes, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0004] In view of this, the present invention provides a method and system for intelligent identification of risk factors for early-onset type 2 diabetes based on deep neural networks.
[0005] To achieve the above objectives, the present invention adopts the following technical solution: A method for intelligent identification of risk factors for early-onset type 2 diabetes based on deep neural networks includes: Step 1: Obtain data on multiple types of early-onset type 2 diabetes and perform data preprocessing; Step 2: Divide the preprocessed data into population groups based on age and gender, and construct a risk factor identification model for early-onset type 2 diabetes based on deep neural networks according to the data of different population groups. Step 3: Input data from different population groups into the corresponding early-onset type 2 diabetes risk factor identification model to obtain the risk factor identification results for different population groups.
[0006] Optionally, in step 1, data on multiple types of early-onset type 2 diabetes may include: questionnaires, clinical diagnosis and treatment, physical examinations, biological samples, medical images, and follow-up data.
[0007] Optionally, in step 1, data preprocessing includes: removing white noise, data standardization, and filling in missing values.
[0008] Optionally, white noise can be removed, specifically: We used singular value decomposition to remove white noise from various types of early-onset type 2 diabetes data, combining different data types.
[0009] Optional, data consistency, specifically: We perform unified data structure processing on different types of data in various types of early-onset type 2 diabetes.
[0010] Optional, fill in the missing values, specifically: For different types of data, missing values in the early-onset type 2 diabetes data were filled in using different data structures.
[0011] Optionally, in step 2, the preprocessed data is segmented into different population groups based on age and gender, specifically as follows: For individuals at risk of developing early-onset type 2 diabetes who are younger than the preset age threshold for young people, the population is divided into groups based on whether they are adults or not. The population, which was divided based on age, was then categorized by gender.
[0012] Optionally, in step 3, after inputting data from different population groups into the corresponding early-onset type 2 diabetes risk factor identification model and obtaining the risk factor identification results for different population groups, the following steps are also included: performing performance analysis on the early-onset type 2 diabetes risk factor identification model based on the model accuracy and area under the curve, and conducting a comprehensive evaluation of the early-onset type 2 diabetes risk factor identification model.
[0013] This invention also provides a deep neural network-based intelligent identification system for risk factors of early-onset type 2 diabetes, comprising: Multi-type data acquisition and preprocessing module: used to acquire multi-type early-onset type 2 diabetes data and perform data preprocessing; Population segmentation and model building module: This module is used to segment the preprocessed data into populations based on age and gender, and to build a deep neural network-based model for identifying risk factors for early-onset type 2 diabetes based on the data of different populations. Risk factor identification module: This module is used to input data from different population groups into the corresponding risk factor identification model for early-onset type 2 diabetes, and obtain the risk factor identification results for different population data.
[0014] As can be seen from the above technical solution, compared with the prior art, this invention discloses an intelligent identification method and system for risk factors of early-onset type 2 diabetes based on deep neural networks. Compared with existing traditional analysis methods such as logistic regression, this invention integrates deep neural networks to achieve integrated processing of multi-source heterogeneous medical data, significantly improving the accuracy of identifying risk factors for early-onset type 2 diabetes and the utilization rate of model floating-point operations. The model accuracy exceeds 92%, and the AUC-ROC value is higher than 0.91. The computational efficiency is nearly 10 times that of traditional methods. It can accurately distinguish the differentiated core pathogenic risk factors among minors and adult men and women in young populations, providing a reliable quantitative basis for early risk screening and targeted clinical intervention for early-onset type 2 diabetes in young populations. It effectively makes up for the shortcomings of insufficient identification accuracy and low computational efficiency of existing technologies when processing large-scale and multi-type clinical data, and provides a new approach for the effective prevention, control, early diagnosis and treatment of early-onset type 2 diabetes. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0016] Figure 1 This is a schematic diagram of the method flow provided by the present invention. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Example 1: Embodiment 1 of this invention discloses an intelligent identification method for risk factors of early-onset type 2 diabetes based on deep neural networks, such as... Figure 1 As shown, it includes: Step 1: Obtain data on multiple types of early-onset type 2 diabetes and perform data preprocessing.
[0019] Data on early-onset type 2 diabetes includes: questionnaires, clinical diagnosis and treatment, physical examinations, biological samples, medical imaging, follow-up data, etc.
[0020] After obtaining data on multiple types of early-onset type 2 diabetes, the data is classified, stored, and initially organized according to type. Misclassified data between datasets is checked and verified, and the data is then classified or merged again.
[0021] Data preprocessing includes: white noise removal, data standardization, and filling in missing values.
[0022] To remove white noise, specifically: We used singular value decomposition to remove white noise from various types of early-onset type 2 diabetes data, combining different data types.
[0023] Data consistency, specifically: We perform unified data structure processing on different types of data in various types of early-onset type 2 diabetes.
[0024] Fill in the missing values, specifically: For different types of data, missing values in the early-onset type 2 diabetes data were filled in using different data structures.
[0025] Step 2: Divide the preprocessed data into population groups based on age and gender, and construct a risk factor identification model for early-onset type 2 diabetes based on deep neural networks according to the data of different population groups.
[0026] The preprocessed data is segmented into different population groups based on age and gender, specifically as follows: The population at risk of early-onset type 2 diabetes who are younger than the preset young age threshold (40 years old in this embodiment of the invention) are divided into groups based on age and whether they are adults, including: younger than 18 years old and older than or equal to 18 years old and younger than 40 years old. The population, which was divided based on age, was then categorized by gender.
[0027] Step 3: Input data from different population groups into the corresponding early-onset type 2 diabetes risk factor identification model to obtain the risk factor identification results for different population groups.
[0028] After inputting data from different population groups into the corresponding risk factor identification model for early-onset type 2 diabetes and obtaining the risk factor identification results for different population groups, the process also includes: performing performance analysis on the risk factor identification model for early-onset type 2 diabetes based on model accuracy and area under the curve (AUC-ROC value), and conducting a comprehensive evaluation of the risk factor identification model for early-onset type 2 diabetes, as follows: Combining different populations at risk of early-onset type 2 diabetes, a risk factor identification model for early-onset type 2 diabetes was constructed using a deep neural network. Intelligent identification of risk factors for early-onset type 2 diabetes was then performed on different populations, yielding the risk factors for early-onset type 2 diabetes in different groups, as shown in Table 1. The results showed that body mass index (BMI) was the leading risk factor for early-onset type 2 diabetes across all populations. In adolescents, cholesterol was the second leading risk factor for early-onset type 2 diabetes in both men and women. In adults, early-onset type 2 diabetes was closely related to blood pressure, with systolic blood pressure being the second leading risk factor in men and diastolic blood pressure in women.
[0029] Table 1. Risk factors for early-onset type 2 diabetes in different populations
[0030] The performance of the intelligent identification model for early-onset type 2 diabetes risk factors for different populations is shown in Table 2. The accuracy of the models is greater than 92% and the area under the curve is greater than 0.91, which verifies and evaluates the performance and usability of the intelligent identification model for early-onset type 2 diabetes risk factors of the present invention.
[0031] Table 2 Performance of the intelligent identification model for risk factors of early-onset type 2 diabetes in different populations
[0032] Example 2: Embodiment 2 of this invention discloses a comparison between the deep reinforcement learning method of this invention and the existing logistic regression method, as follows: The performance of the intelligent identification model for early-onset type 2 diabetes risk factors in different populations was obtained using the logistic regression method, as shown in Table 3. Simultaneously, the risk factors for early-onset type 2 diabetes in different populations were obtained, as shown in Table 4. The comparison results show that the top three risk factors for early-onset type 2 diabetes obtained using the logistic regression method are basically consistent with the top three risk factors inferred using the deep neural network method. However, the accuracy of the early-onset type 2 diabetes risk factor identification model obtained using the logistic regression method is less than 80%, and the model AUC-ROC value is less than 0.80. This indicates that the accuracy of the early-onset type 2 diabetes risk factor identification model constructed using the logistic regression method is lower than that of the intelligent identification model for early-onset type 2 diabetes constructed using the deep neural network method in this invention, and the AUC-ROC value of the risk factor identification model constructed using the logistic regression method is lower than that of the intelligent identification model for early-onset type 2 diabetes risk factors constructed in this invention. Furthermore, the floating-point utilization (MFU) of the early-onset type 2 diabetes risk factor identification model constructed using the deep neural network method is compared with the MFU of the risk factor identification model constructed using the logistic regression method, as shown in Table 5. By employing optimization techniques such as increasing batch size, kernel optimization, adding model parameters, and improving parallel strategies, the computational load and density per unit time of the model are increased using deep neural network methods. This results in the MFU (Multi-Functional Unit) of the early-onset type 2 diabetes risk factor identification model based on deep neural networks being nearly 10 times more computationally efficient than the model constructed using logistic regression. Therefore, this further verifies the high accuracy and computational efficiency of the intelligent identification of early-onset type 2 diabetes risk factors using deep neural networks, demonstrating the effectiveness and efficiency of the intelligent identification model based on deep neural networks for early-onset type 2 diabetes risk factors.
[0033] Table 3 Performance of the logistic regression-based model for identifying risk factors for early-onset type 2 diabetes
[0034] Table 4. Risk factors for early-onset type 2 diabetes in different populations based on logistic regression.
[0035] Table 5 Comparison of floating-point utilization (MFU) of models for identifying risk factors for early-onset type 2 diabetes based on deep neural networks and logistic regression methods.
[0036] Example 3: Embodiment 3 of this invention discloses a deep neural network-based intelligent identification system for early-onset type 2 diabetes risk factors, comprising: Multi-type data acquisition and preprocessing module: used to acquire multi-type early-onset type 2 diabetes data and perform data preprocessing; Population segmentation and model building module: This module is used to segment the preprocessed data into populations based on age and gender, and to build a deep neural network-based model for identifying risk factors for early-onset type 2 diabetes based on the data of different populations. Risk factor identification module: This module is used to input data from different population groups into the corresponding risk factor identification model for early-onset type 2 diabetes, and obtain the risk factor identification results for different population data.
[0037] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.
[0038] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for intelligent identification of risk factors for early-onset type 2 diabetes based on deep neural networks, characterized in that, include: Step 1: Obtain data on multiple types of early-onset type 2 diabetes and perform data preprocessing; Step 2: Divide the preprocessed data into population groups based on age and gender, and construct a risk factor identification model for early-onset type 2 diabetes based on deep neural networks according to the data of different population groups. Step 3: Input data from different population groups into the corresponding early-onset type 2 diabetes risk factor identification model to obtain the risk factor identification results for different population groups.
2. The intelligent identification method for risk factors of early-onset type 2 diabetes based on deep neural networks according to claim 1, characterized in that, In step 1, the multi-type early-onset type 2 diabetes data includes: questionnaires, clinical diagnosis and treatment, physical examination, biological samples, medical images, and follow-up data.
3. The intelligent identification method for risk factors of early-onset type 2 diabetes based on deep neural networks according to claim 1, characterized in that, In step 1, the data preprocessing includes: removing white noise, data standardization, and filling in missing values.
4. The intelligent identification method for risk factors of early-onset type 2 diabetes based on deep neural networks according to claim 3, characterized in that, The removal of white noise specifically involves: We used singular value decomposition to remove white noise from various types of early-onset type 2 diabetes data, combining different data types.
5. The intelligent identification method for risk factors of early-onset type 2 diabetes based on deep neural networks according to claim 3, characterized in that, The data consistency refers to the following: We perform unified data structure processing on different types of data in various types of early-onset type 2 diabetes.
6. The intelligent identification method for risk factors of early-onset type 2 diabetes based on deep neural networks according to claim 3, characterized in that, The missing value to be filled is specifically: We used different data structures to fill in missing values in the data of early-onset type 2 diabetes.
7. The intelligent identification method for risk factors of early-onset type 2 diabetes based on deep neural networks according to claim 1, characterized in that, In step 2, the preprocessed data is segmented into different population groups based on age and gender, specifically as follows: For individuals at risk of developing early-onset type 2 diabetes who are younger than the preset age threshold for young people, the population is divided into groups based on whether they are adults or not. The population, which was divided based on age, was then categorized by gender.
8. The intelligent identification method for risk factors of early-onset type 2 diabetes based on deep neural networks according to claim 1, characterized in that, In step 3, after inputting data from different population groups into the corresponding early-onset type 2 diabetes risk factor identification model and obtaining the risk factor identification results for different population groups, the following steps are also included: performing performance analysis on the early-onset type 2 diabetes risk factor identification model based on model accuracy and area under the curve, and conducting a comprehensive evaluation of the early-onset type 2 diabetes risk factor identification model.
9. A system for intelligent identification of risk factors for early-onset type 2 diabetes based on deep neural networks, utilizing the intelligent identification method for early-onset type 2 diabetes based on deep neural networks as described in any one of claims 1-8, characterized in that, include: Multi-type data acquisition and preprocessing module: used to acquire multi-type early-onset type 2 diabetes data and perform data preprocessing; Population segmentation and model building module: This module is used to segment the preprocessed data into populations based on age and gender, and to build a deep neural network-based model for identifying risk factors for early-onset type 2 diabetes based on the data of different populations. Risk factor identification module: This module is used to input data from different population groups into the corresponding risk factor identification model for early-onset type 2 diabetes, and obtain the risk factor identification results for different population data.