Session S83.6

Detection Of Obstructive Sleep Apnea From The ECG

Z. Shinar. A. Baharav, S. Akselrod

Tel-Aviv University
Ramat-Aviv, Israel

Recently, it was estimated that almost 1 in 10 men in mid-life is affected by sleep apnea syndrome. Since the traditional gold standard sleep study polysomnography is both cumbersome and expensive, there is an obvious need for a simple and cost effective test to screen large populations for suspected sleep apnea syndrome. Furthermore, it is known that obstructive sleep apnea is often related to the sleeping position, and in some cases completely position dependent (usually supine). Many studies have shown different statistical behavior of the heart rate or its derivatives in normal versus disordered patients, yet the main problem is to define the time segments of HR within which the calculations are performed. In a previous study, R wave duration changes were found to be a good indicator of body position changes [1]. We used this parameter to divide the data into subsequent subsets containing a single body position. Data sets for the study were downloaded from the PhysioNet database, as part of CinC challenge 2000. The data consist of 70 ECG records, divided into a learning set of 35 records, which includes apnea annotations, and a test set of 35 records. The ECG is digitized at 100Hz with 12-bit resolution. Data were analyzed using a combined approach: 1) R peaks detection using a computer program based on second derivative computation, followed by a correction procedure of the miss-detected peaks. 2) R wave duration calculation based on inflection points calculations [1]. 3) Segmentation of each record according to changes in position. 4) Time dependent spectral analysis of each segment. 5) Classification of each record as type A or B or C (defined in http://www.physionet.org/cinc-challenge-2000.shtml) according to the total length of the segments that have high power content in the frequency band of 0.01-0.05Hz (equivalent to periods of 28-100sec). 6) Examination of the RR interval of each suspected segment to identify the pattern of acceleration and deceleration typical to apnea. The results of the classification were 93% correct (entry 20000429.201539, entrant 6), indicating that the combined approach (1-5) constitute a good screening procedure for sleep apnea patients.
Reference: 1. Z. Shinar, A. Baharav, and S. Akselrod, "R Wave duration as a measure of body position changes during sleep". In: Computers in Cardiology. IEEE Comp. Soc., 1999, pp. 49-52.