Modelling of Microcirculatory Dynamics with Auto-regressive Models

Ageing adversely affects most physiological processes, including cardiovascular dynamics. Indeed, vascular ageing plays a major role in several cardiovascular diseases, currently deemed the main cause of death in the world. Vascular ageing produces measurable effects at cardiac level, on major vessels, but also at capillary level. While the effects on the first two components are well known, the dynamics of microcirculation still lacks a complete understanding. In this work, we analyzed the predictability of the pulse waveform at capillary level, using a linear model, aiming at relating the estimated prediction model with the subject age. The pulse waveform was described fitting each pulse with a sum of Gaussian curves, that showed good representation capabilities; each pulse was thus associated with a series of physiologically relevant parameters, derived from the Gaussian curves. For each subject, we estimated a set of auto-regressive models that represent the temporal sequence of each parameter. Then, a classifier was trained to discriminate auto-regressive models according to subject age. Results indicate that, in accordance with literature, the parameter that presents the higher discriminative power is the cardiac cycle duration, with an accuracy in assigning individuals to their age group of 90%, and an area under the ROC curve of 0.915. However, we observed also good performances by modelling the sequence of pulse amplitudes.