Fig 1.
Cardiovascular circuit structure for the BioGears circulatory system.
Respiratory, cerebral, tissue, and renal circuits are omitted for brevity. Left and right ventricle drive the system through elastance changes. Pulmonary arteries oversee oxygen diffusion between the respiratory (gas) circuit into the cardiovascular (liquid) circuit.
Fig 2.
Overview of the implemented model of physiology.
A time step in the model proceeds down in the diagram from preprocess to postprocess. Within preprocess there is an ordering of system execution that proceeds left to right. The heart driver updates elastance as a function of the heart cycle, then the oxygen binding is computed and associated pH calculations. Here the hypovolemia model implements lactate generation. Next the nervous system responds to hypovolemia by adjusting the baroreceptor response. Next the circuit is solved for pressures and flow rates and finally postprocess updates the circuit data values based upon the matrix inversion.
Table 1.
Parameter values for the nervous system models as they respond to pressure stimulus. Signals are sent to update cardiovascular system. Parameter values were taken from prominent prior results with minimal qualitative adjustments.
Fig 3.
Overview of the C++ structure of the BioGears physiology engine.
Common data model handles data input and output as well as standardization of inputs such as patient definition. The synthetic environment wraps the models to generate a complete application programming interface. The engine implements a model of the physiology for a given patient.
Fig 4.
Cardiovascular metrics reported for varying levels of initial hemorrhage.
Patient time series data for the nervous system response and lactate molarity due to hemorrhage for a single patient over multiple initial bleed rates.
Fig 5.
Patient trajectories over multiple hemorrhage events.
Figures are shown after smoothing. Dots denote the computed max curvature for the given trajectory. Blood volume was chosen as a parameter for each physiology value to compare results. A non-linear trend to each set of maximum curvature points is denoted amongst the plots.
Table 2.
Descriptive statistics for the simulated patient population. The targeted cohort was constructed to be a healthy middle age population with no comorbidities. Tight variance was arbitrarily chosen for this investigation but cohort development, to understand model sensitivity, is something that is supported by the BioGears engine.
Fig 6.
Two sets of trajectories with the minimum convex hull computed with vertices at each maximum curvature point.
The top set are simulations initialized with 100 ml/min bleed rate and the bottom are patients initialized with a 200 ml/min bleed rate. Each trajectory corresponds to a distinct patient.
Fig 7.
Convex hulls computed for 5 patients with different bleed rates.
The left most hull corresponds to 200 mL/min, the middle to 150 mL/min, and the left most hull to 100 mL/min. Specific blood volume amounts were not reported in the experimental data, only “severity” of injury. We approximate severity with total blood volume for class 3, 2, and 1 hemorrhage.
Fig 8.
Nonlinear logistic function fit to a collection of maximum curvature points over the entire simulated dataset.
We note a good fit up to the presented confidence interval.