Intelligent closed-loop insulin delivery systems for ICU patients

Area of Science:

• Biomedical Engineering
• Intensive Care Medicine
• Control Systems

Background:

• Effective glycemic control in intensive care units (ICUs) is crucial for reducing patient mortality.
• Individualized insulin dosing models for ICU patients are scarce, posing a significant challenge.
• Existing protocols may not offer optimal glucose management for diverse patient populations.

Purpose of the Study:

• To develop and evaluate a novel, individualized insulin delivery strategy for ICU patients.
• To combine particle swarm optimization (PSO) and model predictive control (MPC) for automated insulin rate calculation.
• To demonstrate the superiority of the proposed method over existing protocols like the Yale protocol.

Main Methods:

• A hybrid PSO-MPC approach was employed for online model identification and insulin dose optimization.
• Ten typical linear dynamic models were used to approximate individual patient models.
• PSO updated model weights, while MPC determined the insulin delivery rate.
• The strategy was validated against the Yale protocol using 30 virtual subjects.

Main Results:

• The proposed PSO-MPC strategy achieved 100% performance in the A+B zone of the control-variability grid.
• The Yale protocol achieved only 51% performance in the A+B zone.
• This indicates significantly improved glycemic control and reduced variability with the new strategy.

Conclusions:

• The combined PSO-MPC strategy represents a promising advancement in closed-loop insulin delivery for ICU patients.
• This method effectively and promptly controls glucose levels, bringing them within a safe range.
• The enhanced glycemic control demonstrated by this strategy has the potential to significantly reduce mortality in ICUs.