Physiological Validation of Current Machine Learning Models for Hemodynamic Instability in Humans

Death from exsanguination continues to be a major problem in combat casualty care and the care of civilians subjected to trauma. The ability to detect significant blood loss using traditional vital signs (heart rate, blood pressure) is marginal due to a variety of compensatory mechanisms that maintain blood pressure in the face of marked reductions in circulating blood volume. Along these lines, it is critical to develop monitoring devices and algorithms to non-invasively assess central blood volume in humans for the purposes of facilitating more timely interventions. The standard way to simulate hemorrhage in humans is to use Lower Body Negative Pressure (LBNP). In this technique, the lower body of a supine volunteer is enclosed in an airtight container while suction is applied to cause venous pooling in the legs and reduce central blood volume. This technique generates many physiological adaptations that are similar to that observed during blood loss studies in animals and human volunteers. Thus, the purpose of this study is to obtain data for use in developing monitoring technology to detect hemorrhage and other surrogate markers of central blood volume in humans.