Predicting sleep apnoea syndrome from heart period: a time-frequency wavelet analysis.

ABSTRACT

Heart rate fluctuations are a typical finding during obstructive sleep apnoea, characterised by bradycardia during the apnoeic phase and tachycardia at the restoration of ventilation. In this study, a time-frequency domain analysis of the nocturnal heart rate variability (HRV) was evaluated as the single diagnostic marker for obstructive sleep apnoea syndrome (OSAS). The predictive accuracy of time-frequency HRV variables (wavelet (Wv) decomposition parameters from level 2 (Wv2) to level 256 (Wv256)) obtained from nocturnal electrocardiogram Holter monitoring were analysed in 147 consecutive patients aged 53.8+/-11.2 yrs referred for possible OSAS. OSAS was diagnosed in 66 patients (44.9%) according to an apnoea/hypopnoea index > or = 10. Using receiver-operating characteristic curves analysis, the most powerful predictor variable was Wv32 (W 0.758, p<0.0001), followed by Wv16 (W 0.729, p<0.0001) and Wv64 (W 0.700, p<0.0001). Classification and Regression Trees methodology generated a decision tree for OSAS prediction including all levels of Wv coefficients, from Wv2 to Wv256 with a sensitivity reaching 92.4% and a specificity of 90.1% (percentage of agreement 91.2%) with this nonparametric analysis. Time-frequency parameters calculated using wavelet transform and extracted from the nocturnal heart period analysis appeared as powerful tools for obstructive sleep apnoea syndrome diagnosis.