A model to simulate human cardio-respiratory responses to different fluid resuscitation treatments after hemorrhagic injury.

Decision-support systems based on artificial intelligence and machine learning algorithms can enhance the capability and capacity of medics to provide care for combat casualties during large-scale combat operations. The training and validation of such algorithms require large amounts of vital-sign data, which can be generated using computational models with the appropriate fidelity. Previously, we developed and validated a human cardio-respiratory (CR) model that captures the essential features of the cardiovascular and respiratory responses to hemorrhage and fluid resuscitation. Here, we extended the CR model by adding oxygen transport and fluid exchange between the capillaries and the interstitial space, which allowed us to represent the effect of different resuscitation fluid types, including saline, blood, and blood products, on vital signs and blood variables. We calibrated and validated the model using hemorrhagic-injury and resuscitation data from four experimental swine studies, involving six different types of resuscitation fluids. We captured the general trend of the experimental vital signs and blood variables with average root mean square errors of 6.91 mmHg for mean arterial pressure, 0.49 L/min for cardiac output, 0.72 g/dL for hemoglobin, and 0.70 mL/(kg·min) for delivered oxygen. In addition, model simulations showed that oxygen delivery increased during fluid resuscitation, regardless of the resuscitation fluid type. The extended CR model, with its ability to account for responses to the most widely used resuscitation fluids, will allow us to generate more realistic synthetic data of trauma casualties.