Model Predictive Valve Control of Lung Pressure Profile Tracking

Abstract

Measuring changes in intracranial pressure (ICP) is critical for diagnosing many cerebral pathologies. However noninvasive methods require airway pressure to be precisely controlled. In clinical practice, this is currently performed by the subject breathing into a tube, attempting to follow a target pressure profile. They are assisted by an operator manually releasing airway pressure via a cap, however tracking is poor. This paper develops the first automatic solution, taking the form of model predictive control (MPC) of a variable release valve to assist the subject in tracking the target trajectory. This differs from conventional MPC since the controlled variable is a system parameter rather than an input signal. A novel identification approach for the combined lung model, muscle dynamics and voluntary respiration time-varying system is also proposed. Numerical results validate the approach and show a 44% reduction in tracking error compared with manual assistance.