“人工胰脏”中的血糖监测与预测研究

Diabetes has become the global severe issue which is dangerous to people's health and development of society and economy. The research of artificial pancreas (AP) has changed the traditional treatment way of diabetes mellitus and brought hope to the diabetes patients. Although China has the greatest number of diabetes patients in the world, however, the research on AP is still in the preliminary stage with relative weak basis. Glucose monitoring and prediction are the basis and key in AP. However, most of the existing monitoring and prediction methods are limited to the further improvement of previous methods and incomplete discussions, which are insufficient to well solve the practical problems. Up to now, there is no systematic and feasible solution yet. This project will further analyze the underlying glucose-insulin dynamics. Based on multivariate statistical analysis techniques, it will conduct strategic research on the glucose monitoring and prediction problems in AP and develop the feasible and systematic solution. The major research work include: analyze the multimode characteristics of glucose-insulin dynamics and develop the multimode partion method; develop the statistical models of glcuose monitoring and prediction as well as the monitoring charts and confidence interval for glucose prediction; develop the online implementation strategy, solve the problems of new mode judging and model updating; develop the strategy of model transfer for glucose prediction. The project will promote the invomation and progress of study of AP, showing strategic significance.

糖尿病已经成为危害人类健康、社会和经济发展的全球性突出问题。人工胰脏的研究革新了传统的糖尿病治疗模式，为广大糖尿病患者带来了希望。尽管中国拥有世界第一的糖尿病人口，但对人工胰脏的研究尚处于起步阶段，基础相对薄弱。人工胰脏的基础和关键是血糖监测与预测。但是，目前的血糖监测与预测方法大多是对传统方法的改进或是局部的讨论，对问题的解决和分析难免显现出局限性，并没有形成一个系统性的解决方案。本项目将深入分析血糖-胰岛素的潜在特性，基于多元统计分析技术，面向人工胰脏中的血糖监测与预测问题展开有针对性的重点研究，建立切实可行的系统性解决方案。主要研究内容包括：揭示血糖-胰岛素多模态特性，设计模态划分方法；研究血糖监测、预测的建模方法，设计血糖监测图与预测置信区间；建立在线实施策略，解决新模态判断与模型更新问题；建立血糖预测的"模型迁移"策略。本项目的开展将会推动人工胰脏研究的创新与进步，具有战略意义。

糖尿病已经成为危害人类健康、社会和经济发展的全球性突出问题。针对人工胰脏的研究具有重要的科学研究意义和应用价值。人工胰脏的基础和关键是血糖监测与预测。本项目从深入分析血糖-胰岛素的潜在特性出发，基于多元统计分析技术，面向人工胰脏中的血糖监测与预测问题展开了深入的研究，建立了切实可行的血糖预测与监测系统性解决方案。揭示了血糖变化多模态特性，设计了多模态自动识别与划分方法；针对血糖预测的研究，分析了血糖自相关性以及血糖-外源输入动态性，建立了血糖预测的通用模型策略和“模型迁移”策略，从不同出发点解决了快速经济建模问题；针对血糖监测的可靠性进行了分析，首次将多元统计分析方法用于该领域，提出了血糖仪故障检测方法以及信号噪声滤波处理策略，提高了异常血糖报警的可靠性。通过对本项目的研究，我们在国际杂志及重要国际国内会议上发表论文27篇，包括SCI收录20篇，EI收录7篇（与SCI不重复统计）；已申请5项发明专利，其中授权3项；出版专著1部。在该项目基础上，我们获得了1项专利转化和2项国际合作，将相关理论成果在不同领域进行了成功推广应用。本项目的开展推动了人工胰脏研究的创新与进步，具有重要的实用价值。