RESUMO
Cerebral blood flow (CBF) reflects the rate of delivery of arterial blood to the brain. Since no nutrients, oxygen or water can be stored in the cranial cavity due to space and pressure restrictions, a continuous perfusion of the brain is critical for survival. Anesthetic procedures are known to affect cerebral hemodynamics, but CBF is only monitored in critical patients due, among others, to the lack of a continuous and affordable bedside monitor for this purpose. A potential solution through bioelectrical impedance technology, also known as rheoencephalography (REG), is proposed, that could fill the existing gap for a low-cost and effective CBF monitoring tool. The underlying hypothesis is that REG signals carry information on CBF that might be recovered by means of the application of advanced signal processing techniques, allowing to track CBF alterations during anesthetic procedures. The analysis of REG signals was based on geometric features extracted from the time domain in the first place, since this is the standard processing strategy for this type of physiological data. Geometric features were tested to distinguish between different anesthetic depths, and they proved to be capable of tracking cerebral hemodynamic changes during anesthesia. Furthermore, an approach based on Poincaré plot features was proposed, where the reconstructed attractors form REG signals showed significant differences between different anesthetic states. This was a key finding, providing an alternative to standard processing of REG signals and supporting the hypothesis that REG signals do carry CBF information. Furthermore, the analysis of cerebral hemodynamics during anesthetic procedures was performed by means of studying causal relationships between global hemodynamics, cerebral hemodynamics and electroencephalogram (EEG) based-parameters. Interactions were detected during anesthetic drug infusion and patient positioning (Trendelenburg positioning and passive leg raise), providing evidence of the causal coupling between hemodynamics and brain activity. The provided alternative of REG signal processing confirmed the hypothesis that REG signals carry information on CBF. The simplicity of the technology, together with its low cost and easily interpretable outcomes, should provide a new opportunity for REG to reach standard clinical practice. Moreover, causal relationships among the hemodynamic physiological signals and brain activity were assessed, suggesting that the inclusion of REG information in depth of anesthesia monitors could be of valuable use to prevent unwanted CBF alterations during anesthetic procedures.
RESUMO
An effective data representation methodology on high-dimension feature spaces is presented, which allows a better interpretation of subjacent physiological phenomena (namely, cardiac behavior related to cardiovascular diseases), and is based on search criteria over a feature set resulting in an increase in the detection capability of ischemic pathologies, but also connecting these features with the physiologic representation of the ECG. The proposed dimension reduction scheme consists of three levels: projection, interpretation, and visualization. First, a hybrid algorithm is described that projects the multidimensional data to a lower dimension space, gathering the features that contribute similarly in the meaning of the covariance reconstruction in order to find information of clinical relevance over the initial training space. Next, an algorithm of variable selection is provided that further reduces the dimension, taking into account only the variables that offer greater class separability, and finally, the selected feature set is projected to a 2-D space in order to verify the performance of the suggested dimension reduction algorithm in terms of the discrimination capability for ischemia detection. The ECG recordings used in this study are from the European ST-T database and from the Universidad Nacional de Colombia database. In both cases, over 99% feature reduction was obtained, and classification precision was over 99% using a five-nearest-neighbor classifier (5-NN).
