Robust Nonlinear State Space Model Identification for Hemorrhage Resuscitation

This paper presents a novel method for predicting hemodynamic responses in hemorrhage resuscitation. The proposed approach, namely, robust nonlinear state space modeling (RNSSM), aims to overcome challenges of identifying reliable models using limited and noisy critical care data by innovatively integrating autoencoder learning and variational Gaussian inference in a unified framework. Simulation results demonstrate the initial feasibility and performance evidence of the RNSSM approach as a digital twin of an animal study in hemorrhage resuscitation scenarios. Clinical Relevance- Enabling reliable, personalized hemodynamic models amenable to the closed-loop control design can potentially lead to development of efficient model-informed precision dosing strategies, promoting patient safety and outcomes in critical care.