[1]王美林,翁鑫凯,凌永权.基于PPG信号的双阶段机器学习无创血糖预测方法[J].中国医学物理学杂志,2026,43(4):489-496.[doi:DOI:10.3969/j.issn.1005-202X.2026.04.011]
　WANG Meilin,WENG Xinkai,LING Yongquan.Dual-stage machine learning framework for non-invasive blood glucose prediction based on photoplethysmography signals[J].Chinese Journal of Medical Physics,2026,43(4):489-496.[doi:DOI:10.3969/j.issn.1005-202X.2026.04.011]
点击复制

基于PPG信号的双阶段机器学习无创血糖预测方法()

《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]

卷:
43卷
期数:
2026年第4期
页码:
489-496
栏目:
医学信号处理与医学仪器
出版日期:
2026-04-28

文章信息/Info

Title:
Dual-stage machine learning framework for non-invasive blood glucose prediction based on photoplethysmography signals
文章编号:
1005-202X（2026）04-0489-08
作者:
王美林; 翁鑫凯; 凌永权
广东工业大学信息工程学院， 广东 广州 510006
Author(s):
WANG Meilin;  WENG Xinkai;  LING Yongquan
School of Information Engineering, Guangdong University of Technology, Guangzhou 510006, China
关键词:
血糖; PPG信号; Savitzky-Golay滤波; 特征重要性排序; 双阶段预测模型; 随机森林; XGBoost 模型
Keywords:
blood glucose photoplethysmography signals Savitzky-Golay filter feature importance ranking dual-stage prediction model random forest extreme gradient boosting model
分类号:
R318；TP391
DOI:
DOI:10.3969/j.issn.1005-202X.2026.04.011
文献标志码:
A
摘要:
提出一种基于光电容积脉搏波（PPG）信号的双阶段非侵入式血糖预测方法。共采集257例受试者的PPG信号，并采用Savitzky-Golay（SG）滤波进行信号去噪，结合稀疏加权非对称最小二乘法（SWALS）进行基线漂移的校正。随后提取涵盖时域、频域和形态学的多维特征，并使用极限梯度提升（XGBoost）与Pearson相关系数（R）递增阈值策略进行特征选择。预测模型方面，采用双阶段建模方案：第一阶段利用随机森林实现初步预测，第二阶段利用XGBoost修正预测残差。经过40例独立受试者数据验证，得到的预测相关系数（R）达到0.844 3，Clarke误差网格（CEG）A+B区覆盖率达到92.5%。结果表明所构建的方法在预测精度和CEG分布表现上优于现有同类无创血糖预测方法。
Abstract:
A dual-stage non-invasive blood glucose prediction method based on photoplethysmography signals (PPG) is proposed in this study. PPG signals collected from 257 subjects are denoised using the Savitzky-Golay filter and corrected for baseline drift using the sparse weighted asymmetric least squares method. Subsequently, multidimensional features covering time domain, frequency domain and morphology are extracted, and feature selection is performed using extreme gradient boosting (XGBoost) and Pearson correlation coefficient (R) incremental threshold strategy. A dual-stage modeling scheme is adopted for the prediction model, with a random forest in the first stage to achieve preliminary prediction and XGBoost in the second stage to correct the prediction residuals. After verification in 40 independent subjects, the prediction Pearson correlation coefficient (R) reaches 0.844 3, and the coverage rate of Clarke error grid (CEG) A+B region is 92.5%. The results demonstrate that the constructed method outperforms existing similar non-invasive blood glucose prediction methods in prediction accuracy and CEG distribution performance.

相似文献/References:

[1]李煜瑶,李静,焦青,等. 2型糖尿病患者功能磁共振成像的局部脑区活动时空四维一致性[J].中国医学物理学杂志,2019,36(10):1162.[doi:DOI:10.3969/j.issn.1005-202X.2019.10.009]
　LI Yuyao,LI Jing,JIAO Qing,et al. Four-dimensional (spatio-temporal) consistency of local neural activity in type 2 diabetes patient on fMRI[J].Chinese Journal of Medical Physics,2019,36(4):1162.[doi:DOI:10.3969/j.issn.1005-202X.2019.10.009]

备注/Memo

备注/Memo:
【收稿日期】2025-11-24 【基金项目】国家自然科学基金（61901124）；广东省自然科学基金（2021A1515012305）；广东省研究生创新教育创新计划（2023JGXM_048） 【作者简介】王美林，博士，副教授，主要研究方向：物联网信号处理、人体信号处理及应用等，E-mail: wml@gdut.edu.cn
更新日期/Last Update: 2026-04-29