Comparison of entropy measures for estimating severity of obstructive sleep apnea from overnight pulse oximetry data

Regularity of overnight pulse oximetry data, which can be quantified using entropy, may be used to characterize severity of obstructive sleep apnea syndrome (OSAS). In this paper three entropy measures, namely Shannon-, approximate-and sample entropy, were compared in terms of their ability to discriminate between mild and severe OSAS. In a group of 5 mild OSAS subjects, with apnea hypopnea index (AHI) <;10, and 20 severe OSAS subjects, with AHI>;10, approximate and sample entropy, but not Shannon entropy, produced significantly higher values in the severe OSAS group. More importantly, it was shown that parameters, including the data sampling frequency, the window length m and the tolerance r, may considerably affect the estimated entropy values. It is concluded that approximate as well as sample entropy can both be efficiently used to estimate severity of OSAS from pulse oximetry, provided the above mentioned parameters are carefully defined.