Search Results (5)

This study investigates a noninvasive continuous glucose monitoring (NI-CGM) system optimized for earlobe application, leveraging the site’s anatomical advantages—absence of bone, muscle, and thick skin—for enhanced optical transmission. The system integrates multimodal sensing, combining near-infrared (NIR) diffuse transmission with temperature and pressure sensors. A novel Multi-Wavelength Slope Efficiency Near-Infrared Spectroscopy (MW-SE-NIRS) method is introduced, enhancing noise robustness through the slope efficiency-based parameterization of NIR signal dynamics. By employing three NIR wavelengths with distinct scattering and absorption properties, the method improves glucose detection reliability, addressing tissue heterogeneity and physiological noise in noninvasive monitoring. To validate the feasibility, a pilot clinical trial enrolled five participants with normal or pre-diabetic glucose profiles. Continuous glucose data capturing pre- and postprandial variations were analyzed using a 1D convolutional neural network (Conv1D). For three subjects under stable physiological conditions, the model achieved 97.0% Clarke error grid (CEG) A-Zone accuracy and a mean absolute relative difference (MARD) of 5.2%. Across all participants, results showed 90.9% CEG A-Zone accuracy and a MARD of 8.4%, with performance variations linked to individual factors such as earlobe thickness variability and physical activity. These outcomes demonstrate the potential of the MW-SE-NIRS system for noninvasive glucose monitoring and highlight the importance of future work on personalized modeling, sensor optimization, and larger-scale clinical validation. Full article

Diabetes requires effective monitoring of the blood glucose level (BGL), traditionally achieved through invasive methods. This study addresses the non-invasive estimation of BGL by utilizing heart rate variability (HRV) features extracted from photoplethysmography (PPG) signals. A systematic feature selection methodology was developed employing advanced metaheuristic algorithms, specifically the Improved Dragonfly Algorithm (IDA), Binary Grey Wolf Optimizer (bGWO), Binary Harris Hawks Optimizer (BHHO), and Genetic Algorithm (GA). These algorithms were integrated with machine learning (ML) models, including Random Forest (RF), Extra Trees Regressor (ETR), and Light Gradient Boosting Machine (LightGBM), to enhance predictive accuracy and optimize feature selection. The IDA-LightGBM combination exhibited superior performance, achieving a mean absolute error (MAE) of 13.17 mg/dL, a root mean square error (RMSE) of 15.36 mg/dL, and 94.74% of predictions falling within the clinically acceptable Clarke error grid (CEG) zone A, with none in dangerous zones. This research underscores the efficiency of utilizing HRV and PPG for non-invasive glucose monitoring, demonstrating the effectiveness of integrating metaheuristic and ML approaches for enhanced diabetes monitoring. Full article

To reduce the error induced by overfitting or underfitting in predicting non-invasive fasting blood glucose (NIBG) levels using photoplethysmography (PPG) data alone, we previously demonstrated that incorporating HbA1c led to a notable 10% improvement in NIBG prediction accuracy (the ratio in zone A of Clarke’s error grid). However, this enhancement came at the cost of requiring an additional HbA1c measurement, thus being unfriendly to users. In this study, the enhanced HbA1c NIBG deep learning model (blood glucose level predicted from PPG and HbA1c) was trained with 1494 measurements, and we replaced the HbA1c measurement (explicit HbA1c) with “implicit HbA1c” which is reversely derived from pretested PPG and finger-pricked blood glucose levels. The implicit HbA1c is then evaluated across intervals up to 90 days since the pretest, achieving an impressive 87% accuracy, while the remaining 13% falls near the CEG zone A boundary. The implicit HbA1c approach exhibits a remarkable 16% improvement over the explicit HbA1c method by covering personal correction items automatically. This improvement not only refines the accuracy of the model but also enhances the practicality of the previously proposed model that relied on an HbA1c input. The nonparametric Wilcoxon paired test conducted on the percentage error of implicit and explicit HbA1c prediction results reveals a substantial difference, with a p-value of 2.75 × 10^–7. Full article

Monitoring diabetes saves lives. To this end, we introduce a novel, unobtrusive, and readily deployable in-ear device for the continuous and non-invasive measurement of blood glucose levels (BGLs). The device is equipped with a low-cost commercially available pulse oximeter whose infrared wavelength (880 nm) is used for the acquisition of photoplethysmography (PPG). For rigor, we considered a full range of diabetic conditions (non-diabetic, pre-diabetic, type I diabetic, and type II diabetic). Recordings spanned nine different days, starting in the morning while fasting, up to a minimum of a two-hour period after eating a carbohydrate-rich breakfast. The BGLs from PPG were estimated using a suite of regression-based machine learning models, which were trained on characteristic features of PPG cycles pertaining to high and low BGLs. The analysis shows that, as desired, an average of 82% of the BGLs estimated from PPG lie in region A of the Clarke error grid (CEG) plot, with 100% of the estimated BGLs in the clinically acceptable CEG regions A and B. These results demonstrate the potential of the ear canal as a site for non-invasive blood glucose monitoring. Full article

To assess intermittently scanned continuous glucose monitoring (isCGM) performance for different rates of change in plasma glucose (RCPG) during glycemic challenges in type 1 diabetes (T1D). Nineteen people with T1D (7 females; age 35 ± 11 years; HbA_1c 7.3 ± 0.6% (56 ± 7 mmol/mol)) performing two glycemic challenges (OGTT) were included. During OGTTs, plasma glucose was compared against sensor glucose for timepoints 0 min (pre-OGTT), +15 min, +30 min, +60 min, +120 min, +180 min, and +240 min by means of median absolute (relative) difference (MARD and MAD) and Clarke Error Grid (CEG), then was stratified for RCPG and glycemic ranges. Overall, MARD was 8.3% (4.0–14.8) during hypoglycemia level 1 18.8% (15.8–22.0), euglycemia 9.5% (4.3–15.1), hyperglycemia level 1 9.4% (4.0–17.2), and hyperglycemia level 2 7.1% (3.3–11.9). The MARD was associated with the RCPG (p < 0.0001), detailing significant differences in comparison of low, moderate, high, and very high RCPG (p = 0.014). Overall, CEG resulted in 88% (212 values) of comparison points in zone A, 12% (29 values) in zone B, and 0.4% (1 value) in zone D. The isCGM system was accurate during OGTTs. Its performance was dependent on the RCPG and showed an overestimation of the actual reference glucose during hypoglycemia. Full article