The ECMO(Extracorporeal Membrane Oxygenator) is a procedure used to maintain oxygen supply and facilitate smooth blood circulation in patients with severely impaired cardiopulmonary function. There are two methods: VV(Veno-Venous) mode for pulmonary a...
The ECMO(Extracorporeal Membrane Oxygenator) is a procedure used to maintain oxygen supply and facilitate smooth blood circulation in patients with severely impaired cardiopulmonary function. There are two methods: VV(Veno-Venous) mode for pulmonary assistance and VA(Veno-Arterial) mode for cardiac support. In VA mode, ECMO directly supplies blood flow to the aorta, leading to high pressure in the aorta and causing issues with the heart's blood ejection. This problem can be mitigated by supplying blood in a counter-pulsation manner, avoiding synchronization with the heartbeats via pulsatile ECMO, thereby alleviating the adverse effects of the VA mode. Additionally, by utilizing IBP(Invasive Blood Pressure) instead of ECG, it becomes possible to control counter-pulsation even in situations where ECG is unavailable. Due to the need to differentiate between the pulses of the heart and the pulsatile ECMO from complex IBP data, this study aims to utilize DNN(Deep Neural Networks) to distinguish between two distinct pulses and subsequently create and evaluate a counter-pulsation control system for pulsatile ECMO. In this study, the DNN for blood pressure waveform analysis was trained using IBP data obtained by connecting the pulsatile VAD(Ventricular Assist Device), which acted as the heart model, and the pulsatile ECMO to a mock-circulation system simulating human blood pressure, along with the activation signal from the pulsatile ECMO. The trained DNN, acting as a classifier, was evaluated for its accuracy by using data not involved in the training process to generate ROC curves, assessing its classification capability. The trained DNN and counter-pulsation control algorithm, configured similarly to the DNN training, were utilized in an in-vitro setting to measure the occurrence rate of counter-pulsation during counter-pulsation control and, for comparison, measured the occurrence rate of counter-pulsation during synchronous pulsation. Additionally, to assess the capability of counter-pulsation control, the duration from simultaneous pulsation to the achievement of counter-pulsation was measured, and to evaluate the effect of counter-pulsation, the pressure and flow rate of the heart model were measured in situations where counter-pulsation occurred compared to scenarios where co-pulsation occurred. The accuracy in distinguishing between the pulse of heart model and the pulsatile ECMO using the blood pressure waveform analysis DNN was measured at 87.55% and 88.76%, respectively. The rate of counter-pulsation during asynchronous pulsation was measured at 25.75%. However, when utilizing the blood pressure waveform analysis DNN for counter-pulsation control, the counter-pulsation rate was measured at 78.62%. Additionally, the duration from co-pulsation until the achievement of counter-pulsation, excluding specific situations at high bpm, was measured within an average of 20 seconds. It was observed that during counter-pulsation, the heart model experienced reduced load compared to when co-pulsation occurred, and there was an increase in the heart model's output flow rate during counter-pulsation. In this study, a PLL(Phase-Locked Loop) method was utilized, which, although having longer transition periods compared to the beat-to-beat method, allows stable execution of counter-pulsation by accumulating previous bpm. Through the blood pressure waveform analysis DNN and the counter-pulsation control algorithm, it was confirmed that the occurrence of counter-pulsation in counter-pulsation control of pulsatile ECMO increased by more than 50% compared to asynchronous pulsation. Furthermore, based on measurements of changes in heart model’s load and output flow rate during both counter-pulsation and co-pulsation, it is possible to anticipate that counter-pulsation in pulsatile ECMO may reduce cardiac load and increase cardiac output.