Machine learning for cardio- and cerebrovascular monitoring

Adequate tissue perfusion depends on the tightly regulated interplay between cardiovascular and respiratory organ systems. When cardiac preload — the blood immediately available to the left heart — declines, the perfusion of tissue relies on the activation of cardiovascular autonomic reflexes to maintain blood pressure: heart rate and cardiac inotropy increase alongside enhanced vasoconstriction. The large variability in these responses between humans makes it challenging to define the current volume status in an individual. Similarly, the perfusion of the brain depends on multiple factors, obfuscating its direct approximation from systemic blood pressure. Within this dissertation we set out to quantify these hemodynamic parameters by the design of experiments that specifically expose healthy subjects and patients to hemodynamic challenges. Then we created machine learning models that learned from the gathered data and extracted the most sensitive information to detect hypovolemia and track brain blood flow during anesthesia.