Abstract We investigated whether including arterial pressure of nitrogen (PaN 2 ) in a deep‐learning analysis of single measurements of arterial blood gases, cardiac output, and indirect calorimetry enables individualized quantification of West's ventilation/perfusion (V/Q) lung model. West's key parameters are shunt (% cardiac output supplying lung units with V/Q = 0), logSD (log standard deviation of unit V/Q ratios), and meanV/Q (mean unit V/Q ratio). By processing randomized combinations of shunt, logSD, meanV/Q, indirect calorimetry, and cardiac output data in a Python computerization of West's model, 2,010,000 blood gases including PaN 2 combined with their input variables completed a simulated monitoring dataset covering broad ranges of oxygenation and acid–base equilibria. Deep‐learning applications trained on these data successfully predicted withheld values of shunt, logSD, and meanV/Q from a separate test dataset of 43,915 samples. Linear regression of predicted versus true values produced R 2 ≥ 0.99 with slopes 0.98–1.00. Kernel density estimates confirmed close agreement. Sensitivity analyses demonstrated high dependence upon PaN 2 . Deep‐learning analysis of single measurements of arterial blood gases, which include PaN 2 , when combined with cardiac output and indirect calorimetry data, can quantify individual lung function with high fidelity in terms of key parameters of West's V/Q model.
Scott et al. (Sun,) studied this question.