Model Predictive Valve Control for Lung Pressure Profile Tracking Assistance

Over 12,000 people in the UK are diagnosed with a brain tumour every year. Current diagnostic methods are expensive and typically invasive, meaning that early, widespread screening is not possible. This paper develops a control set-up that enables a new, low-cost, non invasive eardrum sensor to provide accurate intracranial pressure measurements, thereby facilitating more effective diagnosis. The controller works by assisting participants to accurately track airway pressure profiles via blowing into a tube. This has the effect of removing blood flow fluctuations that would otherwise corrupt the eardrum sensor readings used to compute intracranial pressure.A new experimental hardware set-up, identification process and model predictive control approach are developed to compute the optimal sequence of valve positions, and experimental results show a 21% improvement in pressure profile tracking compared with existing clinical methods.