Intensive care patients with acute respiratory failure usually need support of lung function, which is accomplished with mechanical ventilation and, in most severe cases, extracorporeal gas-exchange. Although mechanical ventilation is a life-saving therapy, it has the potential to worsen lung injury and impair the hemodynamics. Currently, there are different strategies to protect the lungs from injury by the ventilator. Yet, settings may differ substantially regarding to the mechanical energy transferred to lungs, and its distribution across the parenchyma. Variables at the ventilator and extracorporeal lung support device can be set automatically using optimization functions and clinical recommendations, but the handling of experts may still deviate from those settings depending upon clinical characteristics of individual patients. Artificial intelligence can be used to learn from those deviations as well as the patient’s condition in an attempt to improve the combination of settings and accomplish lung support with reduced risk of damage. The project proposed herein aims at developing a hybrid mechanical ventilator/extracorporeal lung support device, where elements communicate wireless, using artificial intelligence-based algorithms to improve the care of invasively mechanically ventilated patients with acute respiratory failure.
Here at NIMI, we are working of the core AI-based system using reinforcement learning in order to provide support and recommendation for experts in finding best settings for the mechanical ventilator system.