Figure 2
Examples of learned non-linear pairwise relationships between non-key predictors and the key predictor. The key predictor on the x-axes, previous pH (pH[t-1]), is shown with non-key predictors etCO2 and ΔSpO2 (left Y-axes). The predicted pH (pH[t]) is the sum of contribution from all predictors. Contribution of each non-key and key predictor pair to the total estimated pH is color-coded, with white indicating higher contributions and black indicating lower contributions (right Y-axes). A prediction example is shown for a hypothetical patient with previously measured pH of 7.25, current etCO2 of 20,and ΔSpO2 of 10 (denoted by the crossing points of the horizontal and vertical yellow bars). The predicted contribution for pH is read from the colormap, denoted by the yellow tick mark. The etCO2, ΔSpO2, and previous pH contributions are 0.91 (12% of predicted pH) and 1.02 (13% of predicted pH) from the learned relationships, respectively, and the total predicted pH is the sum of all contributions. The symbol ‘…’ denotes other predictor contributions not shown.
