基于粒子群和人工神经网络的近红外光谱血糖建模方法研究
发布时间：2018-10-09 13:58
【摘要】：现有的近红外光谱无创血糖建模方法大多是基于多波长近红外光谱信号,不利于无创血糖仪在家庭中普及,并且这些建模方法没有考虑单个个体每天血糖变化规律的差异性。针对这些问题,本文以血糖吸收最强的1 550 nm近红外光吸光度为自变量、血糖浓度为因变量,结合粒子群(PSO)算法和人工神经网络(ANN)建立了一种无创血糖检测模型——PSO-2ANN模型。该模型以两个结构和参数确定的人工神经网络为基本的子模块,通过粒子群算法优化两个子模块的权重系数得到最终的模型。使用PSO-2ANN模型对10名志愿者的实验数据进行预测。结果表明,其中9名志愿者的预测相对误差率均小于20%;通过PSO-2ANN模型得到的血糖浓度预测值分布在克拉克误差网格A、B区域的比重为98.28%,证实了PSO-2ANN模型具有比传统人工神经网络模型更为理想的预测精度和稳健性。另外,单个个体由于外界环境、心情、精神状态等因素的影响,每天血糖的变化规律可能会出现一定程度的差异性,PSO-2ANN模型只需要调节一个参数便能修正这种差异性。本文提出的PSO-2ANN模型为克服血糖浓度预测的个体差异性提供了新的思路。
[Abstract]:Most of the existing NIR modeling methods are based on multi-wavelength NIR signals, which is not conducive to the popularity of non-invasive blood glucose meters in the family, and these modeling methods do not take into account the differences of individual daily blood glucose changes. Aiming at these problems, a noninvasive blood glucose detection model, PSO-2ANN model, is established by combining particle swarm (PSO) algorithm and artificial neural network (ANN). The model is based on two artificial neural networks whose structure and parameters are determined, and the final model is obtained by optimizing the weight coefficients of the two sub-modules by particle swarm optimization (PSO). PSO-2ANN model was used to predict the experimental data of 10 volunteers. The results show that The predicted relative error rate of 9 volunteers was less than 20, and the predicted blood glucose concentration by PSO-2ANN model was 98.28% in the Clark error grid, which proved that the PSO-2ANN model had a better performance than the traditional artificial neural network model. A more ideal prediction accuracy and robustness. In addition, due to the influence of external environment, mood, mental state and other factors, the variation of daily blood glucose may be different to a certain extent. PSO-2ANN model only need to adjust one parameter to correct this difference. The PSO-2ANN model proposed in this paper provides a new idea for overcoming the individual differences in blood glucose concentration prediction.
【作者单位】： 重庆大学生物工程学院;重庆市医疗电子工程技术中心;
【基金】：国家自然科学基金项目(81371713) 中央高校基本科研业务费专项(106112015CDJZR235522)
【分类号】：R587.1;TN219
本文编号：2259595
[Abstract]:Most of the existing NIR modeling methods are based on multi-wavelength NIR signals, which is not conducive to the popularity of non-invasive blood glucose meters in the family, and these modeling methods do not take into account the differences of individual daily blood glucose changes. Aiming at these problems, a noninvasive blood glucose detection model, PSO-2ANN model, is established by combining particle swarm (PSO) algorithm and artificial neural network (ANN). The model is based on two artificial neural networks whose structure and parameters are determined, and the final model is obtained by optimizing the weight coefficients of the two sub-modules by particle swarm optimization (PSO). PSO-2ANN model was used to predict the experimental data of 10 volunteers. The results show that The predicted relative error rate of 9 volunteers was less than 20, and the predicted blood glucose concentration by PSO-2ANN model was 98.28% in the Clark error grid, which proved that the PSO-2ANN model had a better performance than the traditional artificial neural network model. A more ideal prediction accuracy and robustness. In addition, due to the influence of external environment, mood, mental state and other factors, the variation of daily blood glucose may be different to a certain extent. PSO-2ANN model only need to adjust one parameter to correct this difference. The PSO-2ANN model proposed in this paper provides a new idea for overcoming the individual differences in blood glucose concentration prediction.
【作者单位】： 重庆大学生物工程学院;重庆市医疗电子工程技术中心;
【基金】：国家自然科学基金项目(81371713) 中央高校基本科研业务费专项(106112015CDJZR235522)
【分类号】：R587.1;TN219
【相似文献】
相关期刊论文 前1条
1 郭奕瑞;李玉倩;王高帅;刘晓田;张路宁;张红艳;王炳源;王重建;;人工神经网络模型在2型糖尿病患病风险预测中的应用[J];郑州大学学报(医学版);2014年02期
相关重要报纸文章 前1条
1 记者 周前进;人工神经网络可筛查糖尿病[N];健康报;2000年
相关硕士学位论文 前1条
1 洪烨;基于机器学习算法的糖尿病预测模型研究[D];哈尔滨工业大学;2016年
，本文编号：2259595
本文链接：https://www.wllwen.com/kejilunwen/dianzigongchenglunwen/2259595.html
