Functional Near Infrared Spectroscopy in the Investigation of Hemodynamic Changes during General Anesthesia

Anesthesiologists, physiologists and medical device professionals have long been working to design advanced depth of anesthesia monitoring systems in addition to routine physiological measures such as blood pressure (BP), heart rate (HR), respiration rate (RR), peripheral oxygen saturation (SpO2) and end tidal CO₂ (EtCO₂). Various devices have been designed to monitor depth of anesthesia. However, none of the systems have gained widespread use mainly because the warning signal associated with the depth of anesthesia is often evidenced after the anesthesiologist has already noted the change in anesthetic state. Further, there still remains an unmet clinical need for a practical, inexpensive tool for the reliable and objective assessment of the effects of general anesthetics during surgery. Advanced functional imaging modalities such as functional magnetic resonance imaging (fMRI) and positron emission tomography (PET) scans deliver superior spatial information, which comes at a high equipment and maintenance cost. Therefore, they are not readily accessible for routine clinical use in general anesthesia. On the other hand, functional near infrared spectroscopy (fNIRS) is non-invasive, safe, portable, and affordable and has a better temporal resolution with a short set up time which makes it more suitable for operating room settings. In this thesis we explore the feasibility of using the fNIRS measures to detect the transition from maintenance to emergence during general anesthesia. The goals of this thesis are to i) examine the effect of volatile anesthetics, specifically sevoflurane, on hemodynamic parameters measured by the fNIRS, the heart rate, end-tidal carbon dioxide concentration and peripheral oxygen saturation, ii) investigate features extracted from these signals that contain information related to anesthetic states, and iii) propose a method for the automatic classification of maintenance and emergence during general anesthesia with sevoflurane using machine learning algorithms. The results reveal the ability of fNIRS biomarkers to automatically differentiate between maintenance and emergence states in real time during general anesthesia. The maintenance state was identified as a period of relative signal stability, while the emergence state was characterized by signal variability. This suggests that the hemodynamic changes observed during emergence reflect the competing effects of increased vasoconstriction and increased cerebral metabolic rate that occur during sevoflurane washout. We examined linear and non-linear classification methods and concluded that a non-linear method can increase the performance of an fNIRS based classifier. Furthermore, this investigation determined that fNIRS based classifiers are able to outperform the EEG based Bispectral Index (BIS) and minimum alveolar concentration (MAC) when examined individually. When combined, fNIRS and MAC demonstrated the better performance. In summary, this thesis provides evidence that biomarkers derived from the fNIRS measures can reveal differences between anesthetic depths and can be used for real time and automatic detection of maintenance and emergence states during the delivery of general anesthesia with sevoflurane.