基于群智能的胰岛素泵疗法优化策略研究

发布时间：2018-03-07 08:56

本文选题：糖尿病　切入点：粒子群优化算法　出处：《北京化工大学》2015年硕士论文　论文类型：学位论文

【摘要】：糖尿病作为一种常见的内分泌疾病,严重困扰了患者的日常生活并危害了患者的健康,长期高血糖会引发一系列的并发症。目前临床上常用的方法是胰岛素强化治疗,可以分为每日多次皮下胰岛素注射和胰岛素持续皮下注射(胰岛素泵)。胰岛素持续皮下注射更接近于人体的生理模式。胰岛素泵可以通过调整胰岛素基础率使血糖保持稳定,通过输入大剂量有效控制餐后引起的高血糖。尽管胰岛素泵的应用日益广泛,但关于胰岛素泵基础量和大剂量输注模式研究却很少。临床中常根据医师的经验和胰岛素泵使用手册的一些公式估算,然而糖尿病人差异很大,不同病人之间的体重差异很大,可以从50千克到100千克不等,不同病人之间对于胰岛素的敏感性也不一样,这些影响因素差异很大,不能为每个病人设计一个合理的胰岛素基础输注率和大剂量,容易引起高血糖或低血糖的产生,危害糖尿病患者的生命。因此对基础量和大剂量的优化极为重要。本文提出了基于改进的粒子群算法的优化方法,首先提出了一种基于智能权重机制的粒子群算法,主要是提出了一个框架,是由任意搜索方法组合而成的,由一个时变的权重将其结合在一起,进一步提高优化的结果。该方法的核心是如何选取权重,本文权重的选取是与优化效果成正比的。本文采用非均匀变异算子、微分变异算子和随机局部搜索算法三种搜索算法,利用了非均匀变异的局部搜索的优点、微分变异算子使粒子保持多样性的优点以及随机局部搜索算法平衡局部搜索和全局搜索的能力的优点。因此改进粒子群算法有着很好的局部搜索和全局搜索的能力。本文对提出的新的粒子群算法进行了性能测试,主要在15个基准测试函数上进行了详细的测试,并与其它四个优化算法的测试效果进行了比较,仿真结果表明,相对于其它四个优化算法,改进的粒子群优化算法具有很好的收敛速度和搜索精度。其次本文基于改进粒子群算法提出了一种自动调节胰岛素泵基础输注率和大剂量的优化方法,该方法能够根据病人的血糖数据计算并自动调节病人的基础输注率和大剂量,不需要人为的干预,并在10个虚拟病人身上进行了仿真测试,并且与其它四个优化算法进行比较,结果表明该方法能够很好地控制病人的血糖,将病人血糖很快的控制在安全范围内。
[Abstract]:As a common endocrine disease, diabetes seriously disturbs patients' daily life and endangers their health. Long-term hyperglycemia can lead to a series of complications. At present, intensive insulin therapy is commonly used in clinical practice. It can be divided into multiple daily subcutaneous insulin injections and continuous insulin subcutaneous injections. Insulin continuous subcutaneous injection is closer to the physiological model of the human body. Insulin pumps can stabilize blood glucose by adjusting the insulin base rate. Effective control of postprandial hyperglycemia by infusion of large doses. Despite the increasing use of insulin pumps, However, little research has been done on the basic quantity of insulin pump and the model of high dose infusion. In clinical practice, we often estimate the basic amount of insulin pump and the formula of insulin pump use manual. However, there is a great difference in diabetes mellitus patients, and there is a great difference in body weight among different patients. It can range from 50 kg to 100 kg, and the sensitivity to insulin varies from patient to patient, and these factors vary greatly, and it's not possible to design a reasonable basal insulin infusion rate and a large dose for each patient. It is easy to cause hyperglycemia or hypoglycemia and endanger the lives of diabetic patients. Therefore, it is very important to optimize the basic quantity and large dose. In this paper, an optimization method based on improved particle swarm optimization (PSO) is proposed. Firstly, a particle swarm optimization algorithm based on intelligent weight mechanism is proposed. A framework is proposed, which is composed of arbitrary search methods and is combined by a time-varying weight. The core of this method is how to select the weight, and the selection of the weight is proportional to the optimization effect. In this paper, three search algorithms, namely, non-uniform mutation operator, differential mutation operator and random local search algorithm, are used. Take advantage of the advantage of local search of non-uniform variation, Differential mutation operator has the advantages of maintaining diversity of particles and the ability of stochastic local search algorithm to balance local search and global search. Therefore, the improved particle swarm optimization algorithm has good ability of local and global search. In this paper, the performance of the new particle swarm optimization algorithm is tested. The test results of 15 benchmark functions are compared with those of the other four optimization algorithms. The simulation results show that, compared with the other four optimization algorithms, The improved particle swarm optimization algorithm has good convergence speed and searching accuracy. Secondly, based on the improved particle swarm optimization algorithm, an optimization method is proposed to automatically adjust the basic infusion rate and large dose of insulin pump. This method can calculate and automatically adjust the basic infusion rate and large dose of patients according to the blood sugar data of patients, without the need of human intervention, and carry out simulation tests on 10 virtual patients, and compare it with the other four optimization algorithms. The results show that the method can control the blood sugar of the patients very well, and the blood sugar of the patients can be controlled quickly within the safe range.
【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R587.1;TP18

【参考文献】