WO2022237162A1 - 一种血糖预测方法、系统及其应用 - Google Patents

一种血糖预测方法、系统及其应用 Download PDF

Info

Publication number
WO2022237162A1
WO2022237162A1 PCT/CN2021/138105 CN2021138105W WO2022237162A1 WO 2022237162 A1 WO2022237162 A1 WO 2022237162A1 CN 2021138105 W CN2021138105 W CN 2021138105W WO 2022237162 A1 WO2022237162 A1 WO 2022237162A1
Authority
WO
WIPO (PCT)
Prior art keywords
prediction
blood sugar
blood glucose
candidate
value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2021/138105
Other languages
English (en)
French (fr)
Inventor
聂泽东
李景振
刘宇航
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Institute of Advanced Technology of CAS
Original Assignee
Shenzhen Institute of Advanced Technology of CAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Institute of Advanced Technology of CAS filed Critical Shenzhen Institute of Advanced Technology of CAS
Publication of WO2022237162A1 publication Critical patent/WO2022237162A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/145—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue
    • A61B5/14532—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue for measuring glucose, e.g. by tissue impedance measurement
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
    • G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records

Definitions

  • the present application belongs to the field of medical and health technologies, and in particular relates to a blood sugar prediction method, system and application thereof.
  • Blood glucose prediction refers to estimating changes in blood glucose in the future based on current and historical blood glucose values, and is an integral part of diabetes management. By predicting blood sugar and taking certain measures according to the prediction results, diabetic patients can prevent the occurrence of adverse blood sugar events such as hypoglycemia and hyperglycemia in daily life, and improve the quality of life.
  • Blood sugar prediction refers to estimating changes in blood sugar in the future based on current and historical blood sugar values, and is an important part of the diabetes management system. Accurate blood sugar prediction helps diabetic patients to take appropriate measures in advance to prevent the occurrence of adverse blood sugar events such as hypoglycemia and hyperglycemia in daily life.
  • blood glucose prediction is an indispensable part of the closed-loop artificial pancreas system, and the predicted value of blood glucose determines the dose of insulin infusion, etc.
  • Current blood glucose prediction methods are mainly based on algorithms such as autoregressive models, jumping neural networks, recurrent neural networks, and long-term and short-term memory networks.
  • the application Based on the fact that for the same person, the current blood sugar prediction algorithm has relatively high accuracy in the stage of relatively stable blood sugar changes, but its blood sugar prediction accuracy is significantly reduced in the stage of large blood sugar fluctuations (such as after meals), the application provides a blood sugar Forecasting methods, systems and their applications.
  • the present application provides a blood sugar prediction method, the method includes acquiring blood sugar data, performing a preliminary prediction on the blood sugar value, and displaying the prediction result after fusing the preliminary prediction results.
  • the blood glucose data includes current blood glucose value and historical blood glucose value
  • the preliminary prediction of blood glucose value adopts a candidate blood glucose prediction algorithm
  • the candidate blood glucose prediction algorithm is based on the current blood glucose value
  • Preliminary blood glucose prediction is performed based on the historical blood glucose value.
  • the candidate blood sugar prediction algorithm includes support vector machine, partial least square method, ridge regression, kernel ridge regression, long short-term memory network, autoregressive, random forest, decision tree and various machines learning and deep learning algorithms.
  • the fusing of the preliminary prediction results includes using multiple indicators to evaluate the prediction performance of different candidate blood sugar prediction algorithms to obtain evaluation results, and according to the evaluation results, different The candidate blood sugar prediction algorithms are comprehensively scored, and the two candidate blood sugar prediction algorithms with the highest comprehensive scores are selected, and the two candidate blood sugar prediction algorithms are fused.
  • Another embodiment provided by the present application is: the fusion of the two candidate blood sugar prediction algorithms adopts a cluster intelligent optimization algorithm.
  • the present application also provides a blood sugar prediction system, which includes sequentially connected blood sugar data acquisition modules, several candidate blood sugar prediction modules, fusion modules, and output modules; the blood sugar data acquisition module is used to obtain blood sugar values, and transmitted to the candidate blood sugar prediction module; the candidate blood sugar prediction module is used to predict blood sugar; the fusion module is used to fuse different blood sugar prediction results; the output module is used to display blood sugar prediction results .
  • the fusion module includes a multi-indicator evaluation sub-module, a comprehensive scoring sub-module and a prediction result fusion sub-module connected in sequence;
  • the multi-indicator evaluation sub-module is used to adopt multiple different Evaluation of the prediction performance of different candidate blood sugar prediction modules according to the index;
  • the comprehensive scoring sub-module is used to comprehensively score different candidate blood sugar prediction modules according to the evaluation results;
  • the prediction result fusion sub-module is used for comprehensive scoring results,
  • the two candidate blood glucose prediction modules with the highest comprehensive scores are selected, and according to the blood glucose prediction values of the two candidate blood glucose prediction modules, the cluster intelligent optimization algorithm is used to realize the fusion of the candidate blood glucose prediction algorithms.
  • the fitness function of the prediction result fusion sub-module is calculated as follows:
  • the output module includes an early warning sub-module, and the early warning sub-module is used to issue an early warning when the predicted blood sugar value is low blood sugar or high blood sugar value.
  • the present application also provides an application of a blood sugar prediction method, which is applied to the blood sugar prediction of different groups of people.
  • the blood sugar prediction method provided by this application is a multi-mode fusion method based on dynamic multi-index evaluation and cluster intelligent optimization, which is used to realize blood sugar prediction, and provides a real-time, high-accuracy blood sugar prediction solution for diabetic patients.
  • the blood sugar prediction method provided in the present application can realize high-accuracy blood sugar prediction for different groups of people and at different times, and has good generalization and versatility.
  • the blood sugar prediction method provided in this application can be better applied to the blood sugar prediction of different groups of people, and has higher accuracy.
  • the blood sugar prediction method provided in this application is a performance evaluation method for candidate blood sugar prediction algorithms based on multi-index evaluation and comprehensive scoring.
  • the blood sugar prediction method provided by this application is based on the fusion method of swarm intelligence optimization and the design of fitness function.
  • FIG. 1 is a schematic diagram of the blood sugar prediction system of the present application
  • Fig. 2 is a schematic diagram of the blood glucose prediction fusion module of the present application.
  • the present application provides a method for predicting blood sugar.
  • the method includes acquiring blood sugar data, performing a preliminary prediction on the blood sugar level, and displaying the prediction result after fusing the preliminary prediction results.
  • the blood glucose data includes the current blood glucose value and the historical blood glucose value
  • the preliminary prediction of the blood glucose value adopts a candidate blood glucose prediction algorithm
  • the candidate blood glucose prediction algorithm is initially performed according to the current blood glucose value and the historical blood glucose value. Blood sugar prediction.
  • the candidate blood sugar prediction algorithm includes support vector machine, partial least square method, ridge regression, kernel ridge regression, long short-term memory network, autoregressive, random forest, decision tree and various machine learning and deep learning algorithms.
  • the fusing of the preliminary prediction results includes evaluating the prediction performance of different candidate blood sugar prediction algorithms by using multiple indicators to obtain evaluation results, and synthesizing the different candidate blood sugar prediction algorithms according to the evaluation results Scoring, screening out the two candidate blood sugar prediction algorithms with the highest comprehensive score, and fusing the two candidate blood sugar prediction algorithms.
  • the fusion of the two candidate blood sugar prediction algorithms adopts a cluster intelligent optimization algorithm.
  • the present application also provides a blood sugar prediction system, which includes sequentially connected blood sugar data acquisition modules, several candidate blood sugar prediction modules, fusion modules, and output modules; the blood sugar data acquisition module is used to obtain blood sugar values, and transmitted to the candidate blood sugar prediction module; the candidate blood sugar prediction module is used to predict blood sugar; the fusion module is used to fuse different blood sugar prediction results; the output module is used to display blood sugar prediction results .
  • the main function of the blood glucose data acquisition module is to obtain the blood glucose value of diabetic patients in real time by using an implanted minimally invasive blood glucose monitoring sensor or a wearable non-invasive blood glucose monitoring sensor, and transmit the obtained blood glucose value to the candidate blood glucose prediction through wired or wireless means. module.
  • Preliminary predictions of blood glucose values are respectively realized through the candidate blood glucose prediction modules, and the prediction results are transmitted to the fusion module.
  • the fusion module includes a multi-indicator evaluation sub-module, a comprehensive scoring sub-module and a prediction result fusion sub-module connected in sequence; the multi-indicator evaluation sub-module is used to use a plurality of different indicators to evaluate different candidate blood sugar prediction modules The prediction performance is evaluated; the comprehensive scoring submodule is used to comprehensively score different candidate blood glucose prediction modules according to the evaluation results; the prediction result fusion submodule is used to comprehensively score the results and select the two with the highest comprehensive scores
  • the candidate blood sugar prediction module according to the blood sugar prediction values of the two candidate blood sugar prediction modules, adopts a cluster intelligent optimization algorithm to realize the fusion of candidate blood sugar prediction algorithms.
  • r 1 represents the error between the predicted value of blood sugar and the real value at the previous N-5 moments
  • r 3 represents the error between the predicted value and the real value of blood glucose at the previous N-3 moments
  • r 5 represents The error between the predicted blood glucose value and the real value at the previous N-1 moments.
  • serial number direction trend Analyzing conditions 1 keeps falling r 1 ⁇ r 3 ⁇ r 5 2 Rising first and then falling, generally falling r 1 ⁇ r 3 , r 3 ⁇ r 5 , r 1 ⁇ r 5 3 Decline first and then rise, overall decline r 1 ⁇ r 3 , r 3 ⁇ r 5 , r 1 ⁇ r 5 4 Rising first and then falling, generally rising r 1 ⁇ r 3 , r 3 ⁇ r 5 , r 1 ⁇ r 5 5 Decline first and then rise, overall rise r 1 ⁇ r 3 , r 3 ⁇ r 5 , r 1 ⁇ r 5 6 keep rising r 1 ⁇ r 3 ⁇ r 5
  • the main function of this sub-module is to comprehensively score different candidate blood sugar prediction algorithms according to the results of the multi-index evaluation sub-module.
  • it is illustrated by comprehensively scoring four candidate blood sugar prediction algorithms.
  • the four candidate blood glucose prediction algorithms are assigned scores of 3, 2, 1, and 0 points according to the average error value in ascending order.
  • the four candidate blood sugar prediction algorithms are assigned scores of 3, 2, 1, and 0 points according to the error standard deviation from small to large.
  • the score assigned by the algorithm is 3 points; if the direction trend of the candidate blood sugar prediction algorithm is judged to be rising first and then falling ( overall decline) or decline first and then rise (general decline), the score assigned by the algorithm is 2 points; overall rising), the algorithm assigns a score of 1; if the direction trend of a candidate blood glucose prediction algorithm is judged to be rising all the time, the algorithm assigns a score of 0.
  • the comprehensive score is used as the performance evaluation standard of the candidate blood glucose prediction algorithm.
  • prediction result fusion sub-module fitness function is calculated as follows:
  • the output module includes an early warning submodule, which is used to issue an early warning when the predicted blood sugar value is low blood sugar or high blood sugar.
  • Blood sugar prediction refers to estimating changes in blood sugar in the future based on current and historical blood sugar values, which is of great significance for diabetic patients to control adverse blood sugar events and reduce the occurrence of diabetic complications. For example, when it is predicted that hypoglycemia may occur in the next 30 minutes, diabetic patients can take measures in advance, such as suspending insulin infusion, food intake, etc., to prevent the occurrence of hypoglycemia. When it is predicted that hyperglycemia will occur in the future, the infusion of insulin can be increased in advance to avoid the occurrence of diabetic complications such as ketoacidosis.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Physics & Mathematics (AREA)
  • Public Health (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Veterinary Medicine (AREA)
  • Animal Behavior & Ethology (AREA)
  • Epidemiology (AREA)
  • General Physics & Mathematics (AREA)
  • Primary Health Care (AREA)
  • Surgery (AREA)
  • Molecular Biology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computing Systems (AREA)
  • Psychiatry (AREA)
  • Physiology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Databases & Information Systems (AREA)
  • Signal Processing (AREA)
  • Mathematical Physics (AREA)
  • Emergency Medicine (AREA)
  • Optics & Photonics (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)
  • Investigating Or Analysing Biological Materials (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

一种血糖预测方法、系统及其应用，属于医疗健康技术领域。目前的血糖预测算法在血糖变化比较平稳阶段具有较高准确度，但在血糖波动较大的阶段(如餐后)其血糖预测准确度明显降低。该血糖预测方法包括：获取血糖数据，对血糖值进行初步预测，对所述初步预测结果进行融合后显示预测结果。能更好的适用于不同人群的血糖预测，且准确度更高。

Description

一种血糖预测方法、系统及其应用 技术领域
本申请属于医疗健康技术领域，特别是涉及一种血糖预测方法、系统及其应用。
背景技术
糖尿病是一种严重危害人类健康的代谢性疾病。血糖预测是指根据当前和历史血糖值估计未来一段时间内血糖的变化，是糖尿病管理中不可或缺的一部分。糖尿病患者通过预测血糖，根据预测结果采取一定的措施，可预防日常生活中低血糖、高血糖等不良血糖事件的发生，提高生活质量。
血糖预测是指根据当前和历史血糖值估计未来一段时间内血糖的变化，是糖尿病管理系统的重要组成部分。准确的血糖预测有助于糖尿病患者提前采取适当措施，预防日常生活中低血糖、高血糖等不良血糖事件的发生。此外，血糖预测是闭环人工胰腺系统中不可缺少的部分，血糖的预测值决定了胰岛素输注的剂量等。目前血糖预测方法主要是基于自回归模型、跳跃神经网络、递归神经网络、长短期记忆网络等算法。然而，由于人体血糖变化受饮食、情绪、运动等多重因素影响，血糖变化具有时变性、非线性和个性化，导致目前的血糖预测算法在一部分人群中具有较高的准确度，但在另一部分人群中则准确度较低。此外，对于同一个人，目前的血糖预测算法在血糖变化比较平稳阶段具有较高准确度，但在血糖波动较大的阶段(如餐后)其血糖预测准确度明显降低。
发明内容
1.要解决的技术问题
基于对于同一个人，目前的血糖预测算法在血糖变化比较平稳阶段具有较高准确度，但在血糖波动较大的阶段(如餐后)其血糖预测准确度明显降低，本申请提供了一种血糖预测方法、系统及其应用。
2.技术方案
为了达到上述的目的，本申请提供了一种血糖预测方法，所述方法包括获取血糖数据，对血糖值进行初步预测，对所述初步预测结果进行融合后显示预测结果。
本申请提供的另一种实施方式为：所述血糖数据包括当前血糖值和历史血糖值，所述对血糖值进行初步预测采用候选血糖预测算法，所述候选血糖预测算法根据所述当前血糖值和所述历史血糖值初步进行血糖预测。
本申请提供的另一种实施方式为：所述候选血糖预测算法包括支持向量机、偏最小二乘 法、岭回归、核岭回归、长短期记忆网络、自回归、随机森林、决策树多种机器学习及深度学习算法。
本申请提供的另一种实施方式为：所述对所述初步预测结果进行融合包括采用多个指标对不同所述候选血糖预测算法的预测性能进行评估得到评估结果，根据所述评估结果对不同所述候选血糖预测算法进行综合评分，筛选出综合得分最高的2种所述候选血糖预测算法，对2种所述候选血糖预测算法进行融合。
本申请提供的另一种实施方式为：所述对2种所述候选血糖预测算法进行融合采用群集智能优化算法。
本申请还提供一种血糖预测系统，包括依次连接的血糖数据获取模块、若干候选血糖预测模块、融合模块和输出模块；所述血糖数据获取模块，用于获取血糖值，并将所述血糖值传输至所述候选血糖预测模块；所述候选血糖预测模块，用于对血糖进行预测；所述融合模块，用于将不同的血糖预测结果进行融合；所述输出模块，用于显示血糖预测结果。
本申请提供的另一种实施方式为：所述融合模块包括依次连接的多指标评估子模块、综合评分子模块和预测结果融合子模块；所述多指标评估子模块，用于采用多个不同的指标对不同候选血糖预测模块的预测性能进行评估；所述综合评分子模块，用于根据评估结果对不同候选血糖预测模块进行综合评分；所述预测结果融合子模块，用于综合评分结果，选择出综合得分最高的2种候选血糖预测模块，根据所述2种候选血糖预测模块的血糖预测值，采用群集智能优化算法实现候选血糖预测算法的融合。
本申请提供的另一种实施方式为：所述预测结果融合子模块适应度函数计算如下：
Figure PCTCN2021138105-appb-000001
其中p i表示第i个历史时刻的预测血糖值的遗忘因子；p表示当前时刻预测血糖值的遗忘因子；pre 1his_i表示候选血糖预测算法1在第i个历史时刻的血糖预测值，pre 2his_i表示候选血糖预测算法2在第i个历史时刻的血糖预测值；true his_i表示在第i个历史时刻的血糖真实值；pre 1cur表示表示候选血糖预测算法1在当前时刻的血糖预测值；pre 2cur表示表示候选血糖预测算法2在当前时刻的血糖预测值；pre pso表示将两种候选血糖预测算法融合后的当前时刻血糖预测值；k 1为候选血糖预测算法1的权重值，k 2为候选血糖预测算法2的权重值；y为适应度函数的值；通过利用蚂蚁算法或者粒子群优化等群集智能优化算法，求解k 1，k 2，pre pso，从而得出当前时刻的血糖预测结果。
本申请提供的另一种实施方式为：所述输出模块包括预警子模块，所述预警子模块用于当预测的血糖值为低血糖或者高血糖值，发出预警。
本申请还提供一种血糖预测方法的应用，将所述的血糖预测方法应用于不同人群的血糖预测。
3.有益效果
与现有技术相比，本申请提供的一种血糖预测方法、系统及其应用的有益效果在于：
本申请提供的血糖预测方法，一种基于动态多指标评估及群集智能优化的多模融合方法用于实现血糖预测，为糖尿病患者提供一种实时、高准确度的血糖预测方案。
本申请提供的血糖预测方法，对不同的人群、不同的时间均可以实现高准确度的血糖预测，具有良好的泛化性和通用性。
本申请提供的血糖预测方法，能更好的适用于不同人群的血糖预测，且准确度更高。
本申请提供的血糖预测方法，基于多指标评估和综合评分的候选血糖预测算法性能评估方法。
本申请提供的血糖预测方法，基于群集智能优化的融合方法及适应度函数的设计。
附图说明
图1是本申请的血糖预测系统示意图；
图2是本申请的血糖预测融合模块示意图。
具体实施方式
在下文中，将参考附图对本申请的具体实施例进行详细地描述，依照这些详细的描述，所属领域技术人员能够清楚地理解本申请，并能够实施本申请。在不违背本申请原理的情况下，各个不同的实施例中的特征可以进行组合以获得新的实施方式，或者替代某些实施例中的某些特征，获得其它优选的实施方式。
参见图1～2，本申请提供一种血糖预测方法，所述方法包括获取血糖数据，对血糖值进行初步预测，对所述初步预测结果进行融合后显示预测结果。
进一步地，所述血糖数据包括当前血糖值和历史血糖值，所述对血糖值进行初步预测采用候选血糖预测算法，所述候选血糖预测算法根据所述当前血糖值和所述历史血糖值初步进行血糖预测。
进一步地，所述候选血糖预测算法包括支持向量机、偏最小二乘法、岭回归、核岭回归、长短期记忆网络、自回归、随机森林、决策树多种机器学习及深度学习算法。
进一步地，所述对所述初步预测结果进行融合包括采用多个指标对不同所述候选血糖预测算法的预测性能进行评估得到评估结果，根据所述评估结果对不同所述候选血糖预测算法 进行综合评分，筛选出综合得分最高的2种所述候选血糖预测算法，对2种所述候选血糖预测算法进行融合。
进一步地，所述对2种所述候选血糖预测算法进行融合采用群集智能优化算法。
本申请还提供一种血糖预测系统，包括依次连接的血糖数据获取模块、若干候选血糖预测模块、融合模块和输出模块；所述血糖数据获取模块，用于获取血糖值，并将所述血糖值传输至所述候选血糖预测模块；所述候选血糖预测模块，用于对血糖进行预测；所述融合模块，用于将不同的血糖预测结果进行融合；所述输出模块，用于显示血糖预测结果。
血糖数据获取模块主要功能是通过利用植入式微创血糖监测传感器或者可穿戴无创血糖监测传感器实时获取糖尿病患者的血糖值，并将所获取的血糖值通过有线或者无线的方式传输至候选血糖预测模块。
通过所述候选血糖预测模块，分别实现血糖值的初步预测，并将预测结果传输至融合模块。
进一步地，所述融合模块包括依次连接的多指标评估子模块、综合评分子模块和预测结果融合子模块；所述多指标评估子模块，用于采用多个不同的指标对不同候选血糖预测模块的预测性能进行评估；所述综合评分子模块，用于根据评估结果对不同候选血糖预测模块进行综合评分；所述预测结果融合子模块，用于综合评分结果，选择出综合得分最高的2种候选血糖预测模块，根据所述2种候选血糖预测模块的血糖预测值，采用群集智能优化算法实现候选血糖预测算法的融合。
1)多指标评估子模块：
该子模块主要是采用多个不同的指标对不同候选血糖预测算法的预测性能进行评估。具体来说，分别计算每一种候选血糖预测算法的预测值与真实值的误差，并从三个指标评估每种候选血糖预测误差：指标1，计算每一种候选血糖预测算法的前N(N可以取不同的整数，以N＝5为例)个时刻的历史的预测值及历史真实值的误差的平均值；指标2，计算每一种候选血糖预测算法的前N个历史的预测值及历史真实值的误差的标准差；指标3：计算每一种候选血糖预测算法的前N个历史的预测值及历史真实值的误差的方向趋势，方向趋势计算如表1所示，共有6种变化趋势，包括一直下降、先上升后下降(总体为下降)、先下降后上升(总体为下降)、先上升后下降(总体为上升)、先下降后上升(总体为上升)、一直上升。各种趋势的判定方法如表1，其中r 1表示前N-5时刻的血糖预测值与真实值的误差，r 3表示前N-3时刻的血糖预测值与真实值的误差，r 5表示前N-1时刻的血糖预测值与真实值的误差。
表1方向趋势判断表
序号 方向趋势 判断条件
1 一直下降 r 1≥r 3≥r 5
2 先上升后下降，总体为下降 r 1≤r 3,r 3≥r 5,r 1≥r 5
3 先下降后上升，总体为下降 r 1≥r 3,r 3≤r 5,r 1≥r 5
4 先上升后下降，总体为上升 r 1≤r 3,r 3≥r 5,r 1≤r 5
5 先下降后上升，总体为上升 r 1≥r 3,r 3≤r 5,r 1≤r 5
6 一直上升 r 1≤r 3≤r 5
2)综合评分子模块：
该子模块主要实现的功能是根据多指标评估子模块的结果对不同候选血糖预测算法进行综合评分。在本发明中以对四种候选血糖预测算法进行综合评分来说明。首先，根据多指标评估子模块的指标1结果，按误差平均值按从小到大，分别给四种候选血糖预测算法分配得分为3，2，1，和0分。相似的，根据多指标评估子模块的指标2结果，按误差标准差按从小到大，分别给四种候选血糖预测算法分配得分为3，2，1，和0分。根据多指标评估子模块的指标3结果，如果某候选血糖预测算法的方向趋势判定为一直下降，则该算法分配的得分为3分；如果候选血糖预测算法的方向趋势判定为先上升后下降(总体为下降)或者先下降后上升(总体为下降)，则该算法分配的得分为2分；如果候选血糖预测算法的方向趋势判定为先上升后下降(总体为上升)或者先下降后上升(总体为上升)，则该算法分配的得分为1分；如果某候选血糖预测算法的方向趋势判定为一直上升，则该算法分配的得分为0分。通过统计各个候选血糖预测算法在指标1、指标2及指标3中的综合得分，将其综合得分作为该候选血糖预测算法的性能评估标准。
进一步地，所述预测结果融合子模块适应度函数计算如下：
Figure PCTCN2021138105-appb-000002
其中p i表示第i个历史时刻的预测血糖值的遗忘因子；p表示当前时刻预测血糖值的遗忘因子；pre 1his_i表示候选血糖预测算法1在第i个历史时刻的血糖预测值，pre 2his_i表示候选血糖预测算法2在第i个历史时刻的血糖预测值；true his_i表示在第i个历史时刻的血糖真实值；pre 1cur表示表示候选血糖预测算法1在当前时刻的血糖预测值；pre 2cur表示表示候选血糖预测算法2在当前时刻的血糖预测值；pre pso表示将两种候选血糖预测算法融合后的当前时刻 血糖预测值；k 1为候选血糖预测算法1的权重值，k 2为候选血糖预测算法2的权重值；y为适应度函数的值；通过利用蚂蚁算法或者粒子群优化等群集智能优化算法，求解k 1，k 2，pre pso，从而得出当前时刻的血糖预测结果。
进一步地，所述输出模块包括预警子模块，所述预警子模块用于当预测的血糖值为低血糖或者高血糖值，发出预警。
本申请还提供一种血糖预测方法的应用，将所述的血糖预测方法应用于不同人群的血糖预测。血糖预测是指根据当前和历史血糖值估计未来一段时间内血糖的变化，其对糖尿病病人控制不良血糖事件发生以及减少糖尿病并发症发生具有重要的意义。例如，当预测在未来30分钟内可能发生低血糖时，糖尿病患者可以提前采取措施，如暂停胰岛素的输注、摄入食物等，以阻止低血糖症的发生。当预测在未来将会发生高血糖时，则可以提前加大胰岛素的输注，以避免酮症酸中毒等糖尿病并发症的发生。
尽管在上文中参考特定的实施例对本申请进行了描述，但是所属领域技术人员应当理解，在本申请公开的原理和范围内，可以针对本申请公开的配置和细节做出许多修改。本申请的保护范围由所附的权利要求来确定，并且权利要求意在涵盖权利要求中技术特征的等同物文字意义或范围所包含的全部修改。

Claims (10)

  1. 一种血糖预测方法，其特征在于：所述方法包括获取血糖数据，对血糖值进行初步预测，对所述初步预测结果进行融合后显示预测结果。
  2. 如权利要求1所述的血糖预测方法，其特征在于：所述血糖数据包括当前血糖值和历史血糖值，所述对血糖值进行初步预测采用候选血糖预测算法，所述候选血糖预测算法根据所述当前血糖值和所述历史血糖值初步进行血糖预测。
  3. 如权利要求2所述的血糖预测方法，其特征在于：所述候选血糖预测算法包括支持向量机、偏最小二乘法、岭回归、核岭回归、长短期记忆网络、自回归、随机森林、决策树多种机器学习及深度学习算法。
  4. 如权利要求2所述的血糖预测方法，其特征在于：所述对所述初步预测结果进行融合包括采用多个指标对不同所述候选血糖预测算法的预测性能进行评估得到评估结果，根据所述评估结果对不同所述候选血糖预测算法进行综合评分，筛选出综合得分最高的2种所述候选血糖预测算法，对2种所述候选血糖预测算法进行融合。
  5. 如权利要求4所述的血糖预测方法，其特征在于：所述对2种所述候选血糖预测算法进行融合采用群集智能优化算法。
  6. 一种血糖预测系统，其特征在于：包括依次连接的血糖数据获取模块、若干候选血糖预测模块、融合模块和输出模块；
    所述血糖数据获取模块，用于获取血糖值，并将所述血糖值传输至所述候选血糖预测模块；
    所述候选血糖预测模块，用于对血糖进行预测；
    所述融合模块，用于将不同的血糖预测结果进行融合；
    所述输出模块，用于显示血糖预测结果。
  7. 如权利要求6所述的血糖预测系统，其特征在于：所述融合模块包括依次连接的多指标评估子模块、综合评分子模块和预测结果融合子模块；
    所述多指标评估子模块，用于采用多个不同的指标对不同候选血糖预测模块的预测性能进行评估；
    所述综合评分子模块，用于根据评估结果对不同候选血糖预测模块进行综合评分；
    所述预测结果融合子模块，用于综合评分结果，选择出综合得分最高的2种候选血糖预测模块，根据所述2种候选血糖预测模块的血糖预测值，采用群集智能优化算法实现候选血糖预测算法的融合。
  8. 如权利要求7所述的血糖预测系统，其特征在于：所述预测结果融合子模块适应度函 数计算如下：
    Figure PCTCN2021138105-appb-100001
    其中p i表示第i个历史时刻的预测血糖值的遗忘因子；p表示当前时刻预测血糖值的遗忘因子；pre 1his_i表示候选血糖预测算法1在第i个历史时刻的血糖预测值，pre 2his_i表示候选血糖预测算法2在第i个历史时刻的血糖预测值；true his_i表示在第i个历史时刻的血糖真实值；pre 1cur表示表示候选血糖预测算法1在当前时刻的血糖预测值；pre 2cur表示表示候选血糖预测算法2在当前时刻的血糖预测值；pre pso表示将两种候选血糖预测算法融合后的当前时刻血糖预测值；k 1为候选血糖预测算法1的权重值，k 2为候选血糖预测算法2的权重值；y为适应度函数的值；通过利用蚂蚁算法或者粒子群优化等群集智能优化算法，求解k 1，k 2，pre pso，从而得出当前时刻的血糖预测结果。
  9. 如权利要求6所述的血糖预测系统，其特征在于：所述输出模块包括预警子模块，所述预警子模块用于当预测的血糖值为低血糖或者高血糖值，发出预警。
  10. 一种血糖预测方法的应用，其特征在于：将权利要求1～5中任一项所述的血糖预测方法应用于不同人群的血糖预测。
PCT/CN2021/138105 2021-05-12 2021-12-14 一种血糖预测方法、系统及其应用 Ceased WO2022237162A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110518780.9A CN113205884B (zh) 2021-05-12 2021-05-12 一种血糖预测方法、系统及其应用
CN202110518780.9 2021-05-12

Publications (1)

Publication Number Publication Date
WO2022237162A1 true WO2022237162A1 (zh) 2022-11-17

Family

ID=77030993

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/138105 Ceased WO2022237162A1 (zh) 2021-05-12 2021-12-14 一种血糖预测方法、系统及其应用

Country Status (2)

Country Link
CN (1) CN113205884B (zh)
WO (1) WO2022237162A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113205884B (zh) * 2021-05-12 2023-02-24 中国科学院深圳先进技术研究院 一种血糖预测方法、系统及其应用
CN114550940A (zh) * 2022-04-26 2022-05-27 广东工业大学 基于Stacking模型融合的无创血糖估计方法、系统及存储介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180150609A1 (en) * 2016-11-29 2018-05-31 Electronics And Telecommunications Research Institute Server and method for predicting future health trends through similar case cluster based prediction models
CN108766578A (zh) * 2018-05-16 2018-11-06 清华大学深圳研究生院 一种血糖预测方法及装置
CN109935331A (zh) * 2019-04-02 2019-06-25 北京航空航天大学 一种基于多模型动态综合的血糖预测方法及系统
CN111631704A (zh) * 2020-05-29 2020-09-08 中国科学院深圳先进技术研究院 基于心电与脑电信息结合的糖尿病前期检测系统和方法
CN112102953A (zh) * 2020-10-22 2020-12-18 平安科技（深圳）有限公司 个性化糖尿病健康管理系统、设备及存储介质
CN113205884A (zh) * 2021-05-12 2021-08-03 中国科学院深圳先进技术研究院 一种血糖预测方法、系统及其应用

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110443314A (zh) * 2019-08-08 2019-11-12 中国工商银行股份有限公司 基于机器学习的景区客流量预测方法及装置
CN110517748A (zh) * 2019-08-29 2019-11-29 燕山大学 一种基于三支决策的双激素人工胰脏模型预测控制算法
CN111128327B (zh) * 2019-12-30 2023-05-26 中国科学院深圳先进技术研究院 一种低血糖预警方法和装置
CN111144950B (zh) * 2019-12-30 2023-06-30 北京顺丰同城科技有限公司 模型筛选方法、装置、电子设备及存储介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180150609A1 (en) * 2016-11-29 2018-05-31 Electronics And Telecommunications Research Institute Server and method for predicting future health trends through similar case cluster based prediction models
CN108766578A (zh) * 2018-05-16 2018-11-06 清华大学深圳研究生院 一种血糖预测方法及装置
CN109935331A (zh) * 2019-04-02 2019-06-25 北京航空航天大学 一种基于多模型动态综合的血糖预测方法及系统
CN111631704A (zh) * 2020-05-29 2020-09-08 中国科学院深圳先进技术研究院 基于心电与脑电信息结合的糖尿病前期检测系统和方法
CN112102953A (zh) * 2020-10-22 2020-12-18 平安科技（深圳）有限公司 个性化糖尿病健康管理系统、设备及存储介质
CN113205884A (zh) * 2021-05-12 2021-08-03 中国科学院深圳先进技术研究院 一种血糖预测方法、系统及其应用

Also Published As

Publication number Publication date
CN113205884A (zh) 2021-08-03
CN113205884B (zh) 2023-02-24

Similar Documents

Publication Publication Date Title
CN111105860B (zh) 面向慢性病康复的精准运动大数据智能预测、分析及优化系统
CN118711822B (zh) 心脑血管疾病风险预测方法及系统
CN108648827A (zh) 心脑血管疾病风险预测方法及装置
EP3759722A1 (en) Method, computing device and wearable device for sleep stage detection
Utomo et al. Automatic sleep stage classification using weighted ELM and PSO on imbalanced data from single lead ECG
CN113205884B (zh) 一种血糖预测方法、系统及其应用
CN119791621A (zh) 一种多功能生命体征监测方法及系统
CN120340847A (zh) 一种基于体检报告的癌症预警管理方法及系统
CN118737359A (zh) 一种心脏外科术后护理监控方法及系统
CN111916179A (zh) 一种基于人工智能自适应个体体征进行“量身定制”饮食滋补模型方法
Arifuzzaman et al. Heart disease prediction through enhanced machine learning and diverse feature selection approaches
Kranzinger et al. Generalisability of sleep stage classification based on interbeat intervals: validating three machine learning approaches on self-recorded test data
CN120340817B (zh) 基于改进群体算法的老年肿瘤患者衰弱前期风险预测系统
CN119541872B (zh) 一种基于智能监测的体脂率bmi与占胃胶囊效果评估管理方法
CN120183731A (zh) 基于因果机器联合模型的衰老时钟构建方法
CN120413011A (zh) 一种基于人工智能的慢病风险预警系统及方法
Saravanan et al. Reliable diabetes mellitus forecasting using artificial neural network multilayer perceptron
CN118824542A (zh) 基于切换神经过程和胰岛素灵敏度估计的血糖预测系统
Haritha et al. Cuckoo Search-Optimized Residual Networks for Identification of Diabetic Foot Ulcers
Ponugoti et al. MNAFH-Net: Maxout neuron attention forward harmonic network for human stress level detection
CN117854709B (zh) 一种基于深度对比聚类的糖尿病六分型方法和系统
CN120977588B (zh) 基于患者促炎膳食行为分析的用户画像生成系统
Zhang et al. Automatic grading assessments of wearable ECG critical value via deep adaptive-asymmetric PRank algorithm
CN118866365B (zh) 一种基于人工智能判断的远程糖尿病监测管理方法及系统
Tašić et al. Classification of personalized daily data from continuous glucose monitoring

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21941724

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21941724

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 21941724

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 27/05/2024)

122 Ep: pct application non-entry in european phase

Ref document number: 21941724

Country of ref document: EP

Kind code of ref document: A1