Monitoring hemodynamic parameters during anesthesia is essential for maintaining patient stability. In cases of intraoperative hypotension (IOH), an anesthesiologist can respond by increasing systemic vascular resistance, enhancing cardiac contractility, or adding blood volume to restore arterial blood pressure (ABP) and cardiac output. The choice depends on the estimated state of the cardiovascular system (CVS), which is often based on experience and can be non-specific. The goal of this thesis is to explore whether a computational CVS model can support clinical decision-making.
In this work, I analyzed measured ABP waveforms from anesthetized patients undergoing abdominal surgery. I used a signal processing algorithm based on the slope sum function (SSF) to segment individual pulses, applied temporal normalization, and calculated 13 features for each wave. To estimate feature uncertainty a metric I developed, I considered the effect of the transfer between central and brachial ABP as well as signal distortions (simulated Gaussian noise, spikes, baseline shifts, and motion artifacts) and measuring error.
Using the CircAdapt model, I obtained simulated ABP waves for various CVS conditions, defined by primary parameters (vascular resistance, contractility and complience of veins as proxy for blood volume) and secondary parameters (arterial elasticity, heart rate,...). By comparing measured and simulated waves, I computed the likelihood of each real waveform matching a specific point in the parametric space, accounting for feature uncertainty.
Finally, I analyzed how feature uncertainty affects the size of the probable volume in parametric space and how the center of volume shifts over time in relation to intraoperative interventions. This study demonstrates how the integration of simulation, signal analysis, and clinical data can improve the interpretation of ABP waveforms and aid in assessing the physiological state of the CVS.
|
