COMPUTERIZED DETECTION OF RESPIRATORY EVENTS DURING SLEEP FROM RAPID INCREASES IN OXYHEMOGLOBIN SATURATION

A computerized search for rapid resaturation (RES)-defined as increases in oxyhemoglobin saturation (SaO2) of 3% or more within 10 s-was used to detect apneas and hypopneas during sleep by the episodes of compensatory hyperventilation following them. Results were compared to those from computerized search for desaturations (DESAT)-defined as decreases in SaO2 of 4% or more within 40 s-and to simultaneous polysomnography. We studied 30 patients with obstructive sleep apnea (OSA) with an apnea plus hypopnea index (AHI) of 30.8 +/- 6.9 (median +/- SEM) and 23 habitual snorers (HSN) with an AHI of 7 +/- 1.5. Manual scoring of polysomnography revealed 7965 respiratory events (6192 apneas, 1773 hypopneas) in OSA patients and 940 events (411 apneas, 529 hypopneas) in the HSN group. In OSA patients, the computer found 96% of events by searching for RES and 87% by searching for DESAT. The percentage of computer-found events in OSA classified as true positive was 91% for RES and 97% for DESAT. In the HSN group, 83% of polysomnographically scored events were found by RES and 55% by DESAT, with 72% of RES and 84% of DESAT being true positive. The correlation of the number of computer-found RES with the number of events from polysomnography was better in OSA (r = 0.862, p < 0.0001) than in HSN (r = 0.722, p < 0.001). The same was true for DESAT (OSA: r = 0.896, p < 0.0001; HSN: r = 0.637, p < 0.01). In conclusion, computer-found rapid resaturations are more sensitive than desaturations for the detection of respiratory events during sleep from oximetry. Increased sensitivity of RES is accompanied by relatively well-preserved specificity.