S Reisch1, J Daniuk, H Steltner, K H Rühle, J Timmer, J Guttmann. 1. Section of Experimental Anesthesiology, University Hospital Freiburg, Center of Data Analysis and Model Building, University of Freiburg, and Hospital Ambrock, Hagen, Germany.
Abstract
BACKGROUND: The forced oscillation technique (FOT) allows analysis of the upper airway impedance and, hence, detection of obstructive sleep apnea. OBJECTIVE: To evaluate FOT with respect to sensitivity and to specificity in online detection of sleep-disordered breathing patterns and to compare algorithmic onset detection time with manual onset time markers of staff physicians. METHODS: We compared the absolute value mid R:Zmid R: of the impedance with three routinely obtained polysomnographic signals - nasal airflow V(nasal), thoracic excursion Thox and esophageal pressure P(es) - by retrospective analysis of the diagnostic polysomnograms of 51 patients. For each signal we evaluated algorithms for online detection of respiratory events. For each out of five apnea classes, 50 respiratory events marked by staff physicians were drawn randomly from the 51 polysomnograms to optimize the online detection algorithms (learning set). The algorithm analyzes relative changes of signal baseline and amplitude. Again 50 respiratory events were drawn randomly for each apnea class to examine to what extent it is possible to detect event onsets with the algorithms (test set). RESULTS: The sensitivity of the signals varied between 56 and 94% and was on average 74%. The specificity was 96 +/- 1.5% on average. The onset was detected 4-6 s after the initially evaluated onset of the staff physicians. CONCLUSION: We conclude that nasal airflow and FOT are equivalent sensitive measurands for detection of respiratory events. Copyright 2000 S. Karger AG, Basel
BACKGROUND: The forced oscillation technique (FOT) allows analysis of the upper airway impedance and, hence, detection of obstructive sleep apnea. OBJECTIVE: To evaluate FOT with respect to sensitivity and to specificity in online detection of sleep-disordered breathing patterns and to compare algorithmic onset detection time with manual onset time markers of staff physicians. METHODS: We compared the absolute value mid R:Zmid R: of the impedance with three routinely obtained polysomnographic signals - nasal airflow V(nasal), thoracic excursion Thox and esophageal pressure P(es) - by retrospective analysis of the diagnostic polysomnograms of 51 patients. For each signal we evaluated algorithms for online detection of respiratory events. For each out of five apnea classes, 50 respiratory events marked by staff physicians were drawn randomly from the 51 polysomnograms to optimize the online detection algorithms (learning set). The algorithm analyzes relative changes of signal baseline and amplitude. Again 50 respiratory events were drawn randomly for each apnea class to examine to what extent it is possible to detect event onsets with the algorithms (test set). RESULTS: The sensitivity of the signals varied between 56 and 94% and was on average 74%. The specificity was 96 +/- 1.5% on average. The onset was detected 4-6 s after the initially evaluated onset of the staff physicians. CONCLUSION: We conclude that nasal airflow and FOT are equivalent sensitive measurands for detection of respiratory events. Copyright 2000 S. Karger AG, Basel