Validation of Short-Term Blood Glucose Prediction Algorithms
Evgeniia L. Litinskaia1, Pavel A. Rudenko1, Kirill V. Pozhar1, and
Nikolai A. Bazaev2
1.Institute of Biomedical Engineering, National Research University of Electronic Technology, Zelenograd, Moscow, Russian Federation
2.Sechenov University, Moscow, Russian Federation
2.Sechenov University, Moscow, Russian Federation
Abstract—Algorithms for model predictive control as well as mathematical models themselves need effectiveness evaluation. In the work are considered physiological, neural network based and empirical models, their special aspects and methods of approbation. DirecNet open-access database clinical protocols were processed and used for empirical sigma-model based algorithm tests. The general concept of developed short-term prediction algorithm based sigma-model is to compare the measured and the modeled BG. Processing this data the algorithm generates its outputs and performs further BG prediction. The DirecNet data allows providing effective prediction algorithm and empirical mathematical model evaluation. Primary tests show that sigma-model based algorithm is unsusceptible to patient physiological quasi-constant parameters variability and is susceptible to noise level. Relative deviation of prognosis with added 25% normal noise is less than 20%.
Index Terms—diabetes mellitus, closed-loop system, blood glucose prediction, mathematical model, approbation
Index Terms—diabetes mellitus, closed-loop system, blood glucose prediction, mathematical model, approbation
Cite:Evgeniia L. Litinskaia, Pavel A. Rudenko, Kirill V. Pozhar, and Nikolai A. Bazaev, "Validation of Short-Term Blood Glucose Prediction Algorithms," International Journal of Pharma Medicine and Biological Sciences, Vol. 8, No. 2, pp. 34-39, April 2019. doi: 10.18178/ijpmbs.8.2.34-39