Testing the Short-term Blood Glucose Prediction Algorithm Using DirecNet Clinical Database

Abstract

The goal of this research was testing the short-term blood glucose (BG) prediction algorithm for its applicability in automated insulin-therapy device. The testing was performed using DirecNet database. The patient data was distorted for estimating algorithm stability by adding noise signal. Noise level was set to 10%, 15%, 20% and 25%. The scientific novelty of this research was that the prediction algorithm testing was performed in several aspects: prediction results sensitivity to BG measure error value, BG value prediction quality in case of different patients physiological parameters and algorithm prediction results reproducibility. The obtained results showed the average prediction error detected at levels 2.0%, 3.0%, 6.6%, 7.4% and 13.7% respectively.