Impacts of respiratory fluctuations on cerebral circulation: a machine-learning-integrated 0–1D multiscale hemodynamic model

Abstract

Abstract Objective. This study aims to accurately identify the effects of respiration on the hemodynamics of the human cardiovascular system, especially the cerebral circulation. Approach: we have developed a machine learning (ML)-integrated zero–one-dimensional (0–1D) multiscale hemodynamic model combining a lumped-parameter 0D model for the peripheral vascular bed and a one-dimensional (1D) hemodynamic model for the vascular network. In vivo measurement data of 21 patients were retrieved and partitioned into 8000 data samples in which respiratory fluctuation (RF) of intrathoracic pressure (ITP) was fitted by the Fourier series. ML-based classification and regression algorithms were used to examine the influencing factors and variation trends of the key parameters in the ITP equations and the mean arterial pressure. These parameters were employed as the initial conditions of the 0–1D model to calculate the radial artery blood pressure and the vertebral artery blood flow volume (VAFV). Main results: during stable spontaneous respiration, the VAFV can be augmented at the inhalation endpoints by approximately 0.1 ml s−1 for infants and 0.5 ml s−1 for adolescents or adults, compared to those without RF effects. It is verified that deep respiration can further increase the ranges up to 0.25 ml s−1 and 1 ml s−1, respectively. Significance. This study reveals that reasonable adjustment of respiratory patterns, i.e. in deep breathing, enhances the VAFV and promotes cerebral circulation.