Abstract
In this PhD research project, the frequency spectral analysis of non-invasively measured photoplethysmography (PPG) variability and heart rate variability (HRV) signals was applied to monitor the pathophysiologic response of cardiovascular system control in sepsis, and also to derive potential clinically useful markers for early sepsis diagnosis. From the study of animal sepsis model, all sympathetic related spectral powers in toe- and ear-PPG variabilities, and HRV were significantly altered (p < 0.05), after the onset of endotoxin-induced hypotension. Toe-PPG variability, in particular, displayed a substantial but transient rise in sympathetic-related spectral power at the onset of hypotension, which might be related to the activation and subsequent withdrawal of some non-baroreflex sympathetic drive. Significantly reduced coherence between HRV and systolic BP variability (p < 0.05), on the other hand, might be regarded as the evidence of severe impairment in baroreflex heart rate control. The potential of these non-invasively measured PPG variability and HRV indices was further accredited in a clinical study of sepsis patients, and the outcomes of the study were in good agreement with the animal sepsis model, showing that the normalized MF power of ear-PPG variability was significantly reduced in severe sepsis patients (p < 0.05). These cardiovascular spectral indices were eventually used in a nonlinear support vector machine (SVM) classifier to discriminate sepsis patients into two distinct pathological groups (i.e., systemic inflammatory response syndrome and severe sepsis), showing good classification performance. The use of combined frequency spectral analysis technique and SVM in the identification of sepsis continuums has produced significant outcomes, and in future, more efforts should be devoted into this kind of research work to facilitate early diagnosis of sepsis progression.