Biomedical applications

Overview

Biomedical devices (and biomedical applications in general) are among the most successful (and most noble) applications of control theory. By successfully modeling a natural phenomena in the human body, e.g., blood sugar control loops, we can use the powerful tools of control theory to safely design devices that monitor, sense and take actions that regulate a potentially unstable system due to illness, e.g., diabetes.

One particular application that I investigated during my master thesis is closed-loop fluid resuscitation. This topic concerns the case when patients experience sudden hemorrhage or blood/fluid loss and are in need of rapid fluid resuscitation. In a typical critical care situation, this would be done manually by the medical staff by infusing boluses and carefully monitoring the response. Clearly, such a method is prone to errors and automating it will bring great advantage. Closed-loop fluid resuscitation is a term which describes a control system, consisting of actuators, sensors and regulators, that collectively ensure accurate intravenous fluid infusion based on a certain hemodynamic measurement.

In current clinical practice, blood pressure is measured to a very high accuracy. Therefore, it represents a reliable hemodynamic measurement which can be used to design, e.g., a feedback controller. But first, one must study the effects of hemorrhage/fluid loss on blood pressure perturbations. To that end, we worked on modeling the input-output relationship between fluid infusion/hemorrhage (inputs) and blood pressure perturbations (outputs). We then verified the efficacy of the proposed model using experimental data. It is worth noting that due to the close similarity (on an abstract level) of the physiological systems of humans and sheep, the experimental data that was used was that of sheep. Then, we used the proposed model to design a model-based model-reference adaptive control scheme using blood pressure feedback. We rigorously tested the proposed controller in-silico on a well known and highly nonlinear physiological model of the human cardiovascular system. The results showed that such automated systems outperform manual bolus administration as well as standard PID controllers.

Selected publications