研究目的
To develop a wearable, cost-efficient, and non-invasive continuous blood glucose monitoring system using combined visible-near infrared spectroscopy to address the limitations of existing methods such as invasiveness, high cost, and long settling times.
研究成果
The proposed wearable Vis-NIR biosensor achieves non-invasive continuous blood glucose monitoring with high accuracy (SEP < 6.16 mg/dl, ???? > 0.86) using a 10-second data window. It successfully estimates full-day glucose changes, demonstrating potential for integration into smartwatches and clinical use, though further validation in dynamic environments and with larger cohorts is recommended.
研究不足
The sensor placement on the wrist may suffer from low signal-to-noise ratio and baseline wander during movement; experiments were conducted in controlled conditions (25°C, 60% humidity) with subjects at rest, limiting applicability to active scenarios. The subject group was homogeneous and small; longer-term validation on a larger, diverse population is needed. Calibration relies on PLS, and other methods like AI/ML could be explored for improvement.
1：Experimental Design and Method Selection:
The study uses combined visible-near infrared (Vis-NIR) spectroscopy to measure blood glucose non-invasively by analyzing photoplethysmography (PPG) signals from the wrist tissue. A multivariate analysis with partial least squares (PLS) is employed for calibration.
2：Sample Selection and Data Sources:
12 healthy volunteers (age 22-30, 2 females, 4 males) with no hypertension history participated. Blood glucose range was 70-152 mg/dl. Reference glucose measurements were taken using a finger-prick device (HealthProTM).
3：List of Experimental Equipment and Materials:
LEDs at wavelengths 530 nm, 660 nm, 850 nm, 950 nm; photodiode (SFH7060 package); IR LED VSMY2853G; trans-impedance amplifier; filters (LPF, HPF, BPF); Arduino Due microcontroller; MATLAB R2017b for data analysis; commercial devices for validation (ECG AD8232, PPG RP520).
4：0).
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: Subjects sat comfortably; baseline glucose measured every 10 minutes for 20 minutes, then after carbohydrate-rich meals, data collected every 20 minutes for 120 minutes. PPG signals were acquired from the wrist, pre-processed using digital wavelet transform and moving average, segmented into 10-second windows, features extracted, and PLS model built and validated with 10-fold cross-validation.
5：Data Analysis Methods:
Features extracted include amplitude of AC and DC components, optical density difference, and Teager-Kaiser energy operator metrics. PLS regression used for modeling, with performance evaluated by correlation coefficient (????) and standard error of prediction (SEP).
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
