Neuro-fuzzy control for artificial pancreas: in silico development and validation

• Published in: Revista iberoamericana de automática e informática industrial Vol. 17; no. 4; pp. 390 - 400
• Main Authors: Y. Rios, J. García-Rodríguez, E. Sánchez, A. Alanis, E. Ruiz-Velázquez, A. Pardo
• Format: Journal Article
• Language: Spanish
• Published: Universitat Politecnica de Valencia 01.09.2020
• Subjects: controlador embebido; diabetes mellitus tipo 1; hardware en el lazo; páncreas artificial
• ISSN: 1697-7912; 1697-7920
• DOI: 10.4995/riai.2020.13035

Type 1 Diabetes Mellitus (DMT1) is currently one of the most harmful diseases that aect people of any age, including children from birth. Exogenous insulin injections remain the most common treatment for these patients, however, it is not the optimal one. The scientific community has endeavored to optimize insulin administration using electronic devices and thus improve the diabetics life expectancy. There are numerous limitations for this biomedical evolution to become a reality such as the control algorithms validation, experimentation with electronic devices, and applicability in patients age transcendence, among others. This work presents the prototyping of a neuro-fuzzy intelligent controller on the Texas Instruments LAUNCHXL-F28069M development board to form a hardware in the loop (HIL) scheme. That is, the embedded controller sends the insulin delivery rate data to the computer where it is captured by the Uva/Padova software and integrated into the metabolic simulation of virtual diabetic patients treated with an insulin pump. The main task of the embedded intelligent algorithm is to determine the optimal insulin infusion rate for each of the 30 virtual patients who follow a meal protocol. The novelty of this work focuses on overcoming current limitations through a first intelligent control algorithm approach applicable to artificial pancreas (AP) and analyzing the feasibility of this proposal in age transcendence since the results correspond to in-silico tests in populations of 10 adults, 10 adolescents and 10 children.