Figure 4 | Scientific Reports

Figure 4

From: Continuous noninvasive blood gas estimation in critically ill pediatric patients with respiratory failure

Figure 4

(a) Demonstrates that using prediction uncertainty to abstain on high-uncertainty samples improves estimation accuracy, while randomly abstaining the same percentage of samples provides no accuracy improvement. (b) Compares the estimation performance between samples with different time lags, defined as the time passed since the last BG. (c) The percentage of estimations that fall in the correct range of pH after abstention.

Back to article page