No abstract

Extracorporeal membrane oxygenation (ECMO) is employed to treat critical patients for one to a few days of life support in intensive care units. Venovenous (VV) and venoarterial (VA) ECMO configurations are the most commonly used rescue strategies for temporary cardiac and respiratory function support. However, both ECMO modes sometimes cannot meet a patient’s demands because of (a) less oxygenated blood in either the upper body or lower body, or (b) a deterioration in the patient’s hemodynamic status. Veno-Venoarterial (VVA) ECMO is an upgraded system that provides sufficiently oxygenated blood to the systemic and pulmonary circulation systems. Drainage cannulas and gas flow exchanges are determined to provide the maximum drainage blood flow required by the patient through a servo-regulator that adjusts the motor speed. A generalized regression neural network (GRNN) based estimator is created to automatically estimate the desired pump speed and then provide sufficient drainage flow for temporary life support. To achieve stability flow in an ECMO circuit, a bisection approach algorithm (BAA) is employed to improve the performance of transient responses in step controls and steady state controls. Experimental studies are used to validate the proposed model and it is compared with conventional controllers to indicate good performance in clinical VVA ECMO applications.