Unit of Medical Technology and Intelligent Information Systems, Department of Materials Science and Engineering, University of Ioannina, Ioannina, Greece.
Computational NeuroEngineering Laboratory, University of Florida, Gainesville, FL, USA.
Med Biol Eng Comput. 2019 Jan;57(1):27-46. doi: 10.1007/s11517-018-1859-3. Epub 2018 Jul 2.
This study aims at presenting a nonlinear, recursive, multivariate prediction model of the subcutaneous glucose concentration in type 1 diabetes. Nonlinear regression is performed in a reproducing kernel Hilbert space, by either the fixed budget quantized kernel least mean square (QKLMS-FB) or the approximate linear dependency kernel recursive least-squares (KRLS-ALD) algorithm, such that a sparse model structure is accomplished. A multivariate feature set (i.e., subcutaneous glucose, food carbohydrates, insulin regime and physical activity) is used and its influence on short-term glucose prediction is investigated. The method is evaluated using data from 15 patients with type 1 diabetes in free-living conditions. In the case when all the input variables are considered: (i) the average root mean squared error (RMSE) of QKLMS-FB increases from 13.1 mg dL (mean absolute percentage error (MAPE) 6.6%) for a 15-min prediction horizon (PH) to 37.7 mg dL (MAPE 20.8%) for a 60-min PH and (ii) the RMSE of KRLS-ALD, being predictably lower, increases from 10.5 mg dL (MAPE 5.2%) for a 15-min PH to 31.8 mg dL (MAPE 18.0%) for a 60-min PH. Multivariate data improve systematically both the regularity and the time lag of the predictions, reducing the errors in critical glucose value regions for a PH ≥ 30 min. Graphical abstract ᅟ.
本研究旨在提出一种用于 1 型糖尿病患者皮下血糖浓度的非线性、递归、多变量预测模型。通过固定预算量化核最小均方（QKLMS-FB）或近似线性相关核递归最小二乘（KRLS-ALD）算法，在再生核希尔伯特空间中进行非线性回归，从而实现稀疏模型结构。使用了一个多变量特征集（即皮下血糖、食物碳水化合物、胰岛素方案和身体活动），并研究了其对短期血糖预测的影响。该方法使用 15 名 1 型糖尿病患者在自由生活条件下的数据进行评估。在考虑所有输入变量的情况下：（i）QKLMS-FB 的平均均方根误差（RMSE）从 15 分钟预测期（PH）的 13.1mg/dL（平均绝对百分比误差（MAPE）为 6.6%）增加到 60 分钟 PH 的 37.7mg/dL（MAPE 为 20.8%）；（ii）KRLS-ALD 的 RMSE 可预测地增加，从 15 分钟 PH 的 10.5mg/dL（MAPE 为 5.2%）增加到 60 分钟 PH 的 31.8mg/dL（MAPE 为 18.0%）。多变量数据系统地改善了预测的规律性和时滞性，减少了 PH≥30 分钟时关键血糖值区域的误差。
