Estimating apnea-hypopnea index only from an accelerometer signal measured on the chest.

Abstract

Sleep apnea syndrome (SAS) is becoming a public concern in the field of preventive medicine because it causes more deleterious diseases. However, the majority of patients are not diagnosed and an easier-to-use, lower-cost device for screening of SAS is needed compared to the existing multimodal devices. In this study, we developed and tested a method to estimate the apnea-hypopnea index (AHI), which is the measure of severity of SAS, only from an accelerometer signal measured on the chest. Simultaneous recordings of overnight sleep polysomnography and accelerometer signal were conducted for 50 participants. Three prediction variables were extracted from the accelerometer signal: (i) power of vibration caused by snoring, (ii) variability of frequency of vibration caused by snoring, and (iii) the number of times of rapid changing in the amplitudes of the chest movement accompanied by breathing efforts. Multiple regression analysis was applied to estimate AHI, and the agreement between estimated AHI and true AHI was evaluated by a leave-one-out strategy. The correlation coefficient between the estimated AHI and the true AHI was 0.759, and the sensitivity and the specificity with the threshold of AHI=15 were 83.3% and 88.5%, respectively. Considering the fact that our method requires only an accelerometer to estimate AHI, it has the potential to be a cost-effective way of screening SAS patients.