Development of Adaptive and Bioinspired Systems for Glycemic Control using Insulin Pumps and Continuous Glucose Monitors in Patients with Diabetes Mellitus
Publications of TIN2014-54806-R
Summary of TIN2014-54806-R
Type 1 diabetes mellitus is a chronic disease characterized by the increase of the amount of glucose, due to a lack of insulin, and that only in Spain affects more than 600.000 people. Patients with this disease need, for life, both to measure their glucose levels and to inject subcutaneous insulin. In clinical practice, blood glucose can be measured by continuous glucose monitoring systems, and insulin is injected either manually or by a continuous subcutaneous insulin pump or insulin pump. On the other hand, for a fully autonomous glycemic control, a predictive model to estimate future values of blood glucose would be necessary. With this information, a control algorithm could determine the dose of insulin to be delivered through an insulin pump.
Continuous monitoring systems have problems related to the degeneration of their sensors, the need (in most of them) for frequent calibrations and a lack of adaptation to the particular characteristics of each patient. Furthermore predictive models only consider measurements under controlled conditions of the patients, which in most cases do not reflect neither the real world conditions nor the characteristics of the patients, and use a restricted number of input parameters.
In this project we propose the use of adaptive and bioinspired systems to improve the glycemic control of patients with diabetes mellitus type 1 using insulin pumps and continuous glucose monitors. In the case of continuous glucose monitors, we will address the problems of the signal generated by the sensor, without studying, modifying or changing the type of sensor. In this regard, we will design adaptive digital filters using evolvable hardware that will process the data generated by the sensor in real time, resulting in more reliable and accurate measurements. Furthermore we will use Grammatical Evolution to get more reliable and individualized predictive models of the gluco-regulatory system, eliminating restrictions, such as linearity or the limitation on the input parameters.
At the conclusion of the project we expect to have solved these two problems. On the one hand, we will have a family of evolutionary digital filters adaptable in real time to the sensor and patient characteristics. These filters will be integrated together with the sensor within the glucose monitoring system and in addition to giving precise and reliable measurements of blood glucose, they must be portable and with a set of physical properties (reduced size, performance and power consumption) to allow uninterrupted and unattended operation throughout all the operational life of the sensor. A further advantage will be that those filters can be useful even for glucose monitoring systems using different technologies. On the other hand, we will have implemented a set of predictive software tools based on Grammatical Evolution. These tools will use glycemic values, food intakes, levels of fatigue, stress, etc., to generate predictive models of blood glucose. Biometric sensors will provide the input values that in real time will be sent to the model and control algorithm to generate the action of the insulin pump.