Enhancing Blood Glucose Prediction with Meal Absorption and Physical Exercise Information

TL;DR

The paper addresses the challenge of forecasting blood glucose in type 1 diabetes by integrating meal-absorption and physical-exercise information with CGM signals. It proposes a composite minimal model of glucose–insulin dynamics augmented by deconvolution-based state estimation, incorporating exogenous inputs for meals and activity, and validates the approach on in silico UVa-Padova data and real clinical data against the LVX method. Results show RMSE reductions from $26.68$ to $23.89$ mg/dL (in silico) and from $37.02$ to $35.96$ mg/dL (clinical), with statistically superior hypoglycemia and hyperglycemia prediction, demonstrating the practical potential for safer closed-loop control. The study highlights that meal-absorption information yields substantial gains, with greatest improvement when combined with exercise, and discusses implications for deployment in predictive systems such as PEPPER, along with future work on circadian insulin sensitivity and long-term exercise effects.

Abstract

Objective: Numerous glucose prediction algorithm have been proposed to empower type 1 diabetes (T1D) management. Most of these algorithms only account for input such as glucose, insulin and carbohydrate, which limits their performance. Here, we present a novel glucose prediction algorithm which, in addition to standard inputs, accounts for meal absorption and physical exercise information to enhance prediction accuracy. Methods: a compartmental model of glucose-insulin dynamics combined with a deconvolution technique for state estimation is employed for glucose prediction. In silico data corresponding from the 10 adult subjects of UVa-Padova simulator, and clinical data from 10 adults with T1D were used. Finally, a comparison against a validated glucose prediction algorithm based on a latent variable with exogenous input (LVX) model is provided. Results: For a prediction horizon of 60 minutes, accounting for meal absorption and physical exercise improved glucose forecasting accuracy. In particular, root mean square error (mg/dL) went from 26.68 to 23.89, p<0.001 (in silico data); and from 37.02 to 35.96, p<0.001 (clinical data - only meal information). Such improvement in accuracy was translated into significant improvements on hypoglycaemia and hyperglycaemia prediction. Finally, the performance of the proposed algorithm is statistically superior to that of the LVX algorithm (26.68 vs. 32.80, p<0.001 (in silico data); 37.02 vs. 49.17, p<0.01 (clinical data). Conclusion: Taking into account meal absorption and physical exercise information improves glucose prediction accuracy.

Figures (5)

• Figure 1: $R_a$ profiles corresponding to the fast, slow and medium meals of the employed UVa-Padova simulator for a 60 grams intake of carbohydrates.
• Figure 2: Mean $RMSE$, in $mg/dL$, for the four configurations and the LVX algorithm against to evaluated prediction horizons corresponding to the 10 virtual adults.
• Figure 3: Two-day period close-up of the prediction results for subject adult $1$. The simulated continuous glucose measurements are showed in solid blue line, the prediction results of the LVX method are showed in dotted green line, and the prediction results of the $Configuration~1$ are showed in dashed red line. Vertical pink bars indicate carbohydrate intakes (grams) and vertical light blue bars indicate insulin boluses (units).
• Figure 4: Mean $RMSE$, in $mg/dL$, corresponding to $Configuration ~3$, $Configuration ~1$ and $LVX$ algorithm for different prediction horizons and evaluated on 10 adults subjects.
• Figure 5: two-day period close-up of the prediction results for $Configuration~1$ and $LVX$ method. The simulated continuous glucose measurements are showed in solid blue line, the prediction results of the LVX method are showed in dotted green line, and the prediction results of the $Configuration~1$ are showed in dashed red line. Vertical pink bars indicate carbohydrate intakes (grams) and vertical light blue bars indicate insulin boluses (units).