The advantages of automatic control of the fraction of inspired oxygen in neonates have been documented in recently published clinical trials. Many control algorithms are available, but their comparison is missing in the literature. A mathematical model of neonatal oxygen transport could be a useful tool to compare and enhance both automatic control algorithms and manual control of fraction of inspired oxygen. Besides other components, the model of neonatal oxygen transport must include a module linking arterial (SaO2) and peripheral (SpO2) oxygen saturation. The pulse oximeter module must reflect issues of SpO2 measurement typical for clinical practice, such as overestimation of SpO2 over SaO2 documented by several studies, or inaccurate pulse oximeter readings due to high noise. The aim of this study was to describe both the bias between SaO2 and SpO2 and the noise, characteristic for continuous SpO2 recording, for a computer model of oxygenation of a premature infant. The SpO2-SaO2 bias, derived from available clinical data, describes a typical deviation of the SpO2 measurement as a function of the true SaO2 value in three different SaO2 intervals. The SpO2 measurement noise was considered as a random process that affects biased SpO2values at each time point with statistical properties estimated from SpO2 continuous recordings of 5 stable newborns. The results of the study will help to adjust a computer model of neonatal oxygenation to the real situations observed in the clinical practice.