Adaptive filtering method for the hemodynamic evaluation of the ventricular assist device

A least-mean-square (LMS) adaptive filtering method was applied to multichannel recordings (ECG, radial arterial pressure, pulmonary arterial pressure, right atrial pressure, respiratory signal, and pressure signals related to the function of artificial ventricles) in patients affected with refractory heart failure assisted by the Pierce Donachy biventricular assist device. The signals derived from an intensive care unit monitoring system were recorded on an FM magnetic tape recorder and later digitized. In this clinical model the morphology of the right atrial, pulmonary, and radial arterial pressure curves results from the interaction of independent pumps, namely the right and left ventricular assist device, by the respiratory pump, and the heart. A cascade of adaptive filters was used, each linear time-varying module trying to identify the contribution of different input signals. The analysis performed on signals recorded in two patients, both under mechanical and spontaneous respiration, makes it possible to extract each pump contribution from each pressure signal