Pulmonary gas exchange evaluated by machine learning: a computer simulation

Abstract

Using computer simulation we investigated whether machine learning (ML) analysis of selected ICU monitoring data can quantify pulmonary gas exchange in multi-compartment format. A 21 compartment ventilation/perfusion (V/Q) model of pulmonary blood flow processed 34,551 combinations of cardiac output, hemoglobin concentration, standard P50, base excess, VO₂ and VCO₂ plus three model-defining parameters: shunt, log SD and mean V/Q. From these inputs the model produced paired arterial blood gases, first with the inspired O₂ fraction (FiO₂) adjusted to arterial saturation (SaO₂) = 0.90, and second with FiO₂ increased by 0.1. 'Stacked regressor' ML ensembles were trained/validated on 90% of this dataset. The remainder with shunt, log SD, and mean 'held back' formed the test-set. 'Two-Point' ML estimates of shunt, log SD and mean utilized data from both FiO₂ settings. 'Single-Point' estimates used only data from SaO₂ = 0.90. From 3454 test gas exchange scenarios, two-point shunt, log SD and mean estimates produced linear regression models versus true values with slopes ~ 1.00, intercepts ~ 0.00 and R² ~ 1.00. Kernel density and Bland-Altman plots confirmed close agreement. Single-point estimates were less accurate: R² = 0.77-0.89, slope = 0.991-0.993, intercept = 0.009-0.334. ML applications using blood gas, indirect calorimetry, and cardiac output data can quantify pulmonary gas exchange in terms describing a 20 compartment V/Q model of pulmonary blood flow. High fidelity reports require data from two FiO₂ settings.

Keywords: Computer simulation; Gas exchange; Lung model; MIGET format; Machine learning.

Fig. 1

Graphical illustration of modelled blood flow through 20 gas exchanging compartments plus a single shunt compartment (V/Q = 0). Shunt is set at 10% of total pulmonary blood flow. Note the log normal distribution of the non-shunt pulmonary blood flow according to compartment V/Q ratios. In this example log SD = 2.0 and flow distributional V/Q mean = 0.35

Fig. 2

Shunt (single-point). Two subplots are illustrated. The Bland–Altman (BA) plot illustrates the 3454 points. For clarity, each point is horizontally jittered by ± 1% of the value of the independent variable. Horizontal plot lines indicate the median and 95% confidence interval for the difference (enumerated in Table 5). The kernel density estimate (KDE) plot illustrates the distribution of observations for the independent variable. The solid line is the true value of the variable with the dashed line indicating the modeled variable. Each subplot shares the same X-axis scale. Both X-axis units and the Y-axis units in the BA plot are defined by the independent variable. The Y-axis in the KDE plot is dimensionless

Fig. 3

Shunt (two-point). Description as for Fig. 2

Fig. 4

Mean (single-point). Description as for Fig. 2

Fig. 5

Mean (two-point). Description as for Fig. 2

Fig. 6

Log SD (single-point). Description as for Fig. 2

Fig. 7

Log SD (two-point). Description as for Fig. 2