UC Irvine

This proposal encompasses a comprehensive investigation into non-invasive cardiovascular monitoring techniques, which involved developing and integrating novel signal processing algorithms and machine learning (ML) architectures. We introduced a unique algorithm employing a novel intra-beat biomarker, designed for correcting raw BP data from various sensor sources. Our diastolic transit time (DTT)-based approach successfully eliminated stochastic baseline wander, enabling accurate, single-sensor beat-to-beat blood pressure (BP) measurements. This method not only maintained the spatiotemporal integrity of signals but also mitigated motion artifacts, paving the way for its application in continuous non-invasive BP technologies for acute, outpatient, and home settings. Our investigations into beat-to-beat heart rate and BP variability revealed significant physiological insights, particularly in stroke and sickle cell disease (SCD) patients. We observed increased systolic BP variability in stroke patients compared to healthy controls, a finding that underscores the potential of BP variability as a marker for long-term monitoring and management. In SCD patients, our study highlighted the impact of mental and psychological stressors on BP variability and vascular tone, revealing a complex interplay between autonomic nervous system dysregulation, endothelial dysfunction, and vascular tone alterations. These insights suggest the need for holistic management strategies that address both physiological and psychosocial aspects of disease. The application of ML in predictive health analytics marks a significant advancement in non-invasive monitoring of cardiac output (CO) and intracranial pressure (ICP). We developed a model for real-time, accurate CO monitoring using arterial line BP data, demonstrating the potential of ML to revolutionize acute care patient management. Additionally, our work in developing a non-invasive sensing system for continuous ICP measurement through beat-to-beat BP correlation offers a promising avenue for enhancing neurologic patient care, potentially reducing patient morbidity, mortality, and cost burden. Collectively, this research not only advances our understanding of cardiovascular physiology and its implications for patient care but also showcases the potential of integrating wearable technology and predictive analytics into clinical practice. Thus, the findings and methodologies presented in this dissertation offer significant implications for the development of non-invasive monitoring technologies, with the potential to improve patient outcomes across a broad spectrum of clinical settings.