Signaalverwerking voor het opvolgen van cerebrale hemodynamische bij pasgeborenen

Disturbances in cerebral hemodynamics are one of the principal causes of
cerebral damage in premature infants. Specifically, changes in cerebral blood flow
might cause ischemia or hemorrhage that can lead to motor and developmental
disabilities. Under normal circumstances, there are several mechanisms that act
jointly to preserve cerebral hemodynamics homeostasis. However, in case that
one of these mechanisms is disrupted the brain is exposed to damage. Premature
infants are susceptible to variations in cerebral circulation due to their fragility.
Therefore, monitoring cerebral hemodynamics is of vital importance in order to
prevent brain damage in this population and avoid subsequent sequelae. This
thesis is oriented to the development of signal processing techniques that can
be of help in monitoring cerebral hemodynamics in neonates.
There are several problems that hinder the use in clinical practice of monitoring
cerebral hemodynamics. On one hand, continuous measurements of cerebral
blood flow, or hemodynamical variables, are difficult to obtain in premature
infants. In this context, Near Infrared Spectroscopy (NIRS) is one of the few
technologies that is available for the measurement of hemodynamical variables in
this population. NIRS is a noninvasive and safe technology that is based on light
radiation. NIRS allows the continuous measurement of cerebral oxygenation
that under certain considerations reflects changes in cerebral blood flow. On the
other hand, cerebral hemodynamics assessment is performed by evaluating the
strength of the relationship between some systemic variables, e.g. mean arterial
blood pressure and concentration of CO2, and the cerebral hemodynamics
variables. Under normal conditions cerebral hemodynamics variables should be
independent of systemic variations. Coupled dynamic between systemic and
cerebral hemodynamics variables represents a high risk situation for the patient.
Among the techniques available for the monitoring of cerebral hemodynamics,
most of them assume that the mechanisms responsible for its control are
linear and univariate. In reality, these mechanisms are nonlinear, multivariate,
nonstationary and highly coupled.
This thesis, on one hand, introduces the use of more sophisticated signal
processing techniques for monitoring cerebral hemodynamics, which can address
the multivariate and/or the nonlinear nature of the mechanisms involved in
its control. Linear techniques such as canonical correlation analysis, subspace
projections and wavelet based transfer function; and nonlinear techniques
such as least squares support vector machines and kernel principal component
regression, have been introduced for the NIRS-based monitoring of cerebral
hemodynamics. On the other hand, kernel principal component regression is a
nonlinear methodology that produces as result a black box model, which lacks
clinical interpretability. Therefore, in this thesis attention has been given to
the development of methodologies that allow to interpret the results produced
by this nonlinear model in a clinical framework. For this purpose a method
based on subspace projections is proposed. In addition, in this thesis, results
from several clinical studies related to monitoring cerebral hemodynamics are
presented.