Hylke Salverda

30 Chapter 2 Algorithm design Different approaches have been used for designing AOC algorithms. Generally, a combination of three methods is used: rule-based, proportional-integral-derivative (PID) and adaptive. Central to each of these approaches is the derivation of an SpO2 error, the deviation from the desired value, i.e. the positive or negative difference between the current measured SpO2 and the middle or upper/lower limit of the target range (TR). Rule-based A rule-based algorithm uses a set of rules to decide on an FiO2 adjustment much like a decision tree. The rules are often derived from expert knowledge (for example: “in case of mild hyperoxia, lower FiO2 by 0.02”; Figure 1). Incorporating clinical experience makes this type of algorithm quick to develop and intuitive for clinicians. By combining a large set of rules an attempt is made to cover all possible scenarios. However, there is great heterogeneity in response to a change in FiO2 22, 23, making it virtually impossible to have an exhaustive set of rules for each circumstance. Oxygen saturation trend 100 90 80 70 Time SpO No change FiO + 0.02 FiO – 0.02 No change Start Enter SpO value Hypoxia FiO + 0.02 FiO – 0.02 Hyperoxia Normoxia Oxygenation status FiO alteration Figure 1: A simple rule-based algorithm and its possible effect on the oxygen saturation trend.