Professor Marzia Cescon at ECE Spring 2025 Colloquium

Learning-enabled control methods for personalized health: the case of type 1 diabetes

The future of healthcare delivery will hinge on the personalization of therapies, moving progressively away from the one-size-fits-all approach to treatments tailored to the individual needs, medical history and genetic profile. When the therapeutic regime involves administering a drug, such personalization can be obtained with the use of data-enabled feedback control technology. One prime candidate for the effective development of personalized and adaptive treatment strategies by means of controls is Type 1 Diabetes (T1D). In this chronic disease, the auto-immune destruction of the beta-cells in the pancreas prevents the body from producing insulin, a hormone that is required for the glucose homeostasis feedback loop. As a consequence, patients diagnosed with T1D must rely on exogenous insulin for survival.

In our research group, we aim at artificially recreating the individual-specific glucose feedback loop using a combination of medical devices and learning-enabled control algorithms, to realize a fully automated insulin delivery system tailored to the patient. In this presentation, I will outline the challenges inherent to controlling physiological variables and will describe several learning-enabled control engineering algorithms that we are developing in-silico in our lab for closed-loop glucose control, to improve patients outcomes and quality of life.