Multichannel Analysis of an Unobtrusive Sensor for Sleep Apnea-Hypopnea Detection

Abstract

This manuscript presents an unobtrusive method for sleep apnea-hypopnea syndrome (SAHS) detection. The airflow is indirectly measured through a sensitive mattress (Pressure Bed sensor, PBS) that incorporates multiple pressure sensors into a bed mattress. The instantaneous amplitude of each sensor signal is calculated through Hilbert transform, and then, the information is reduced via principal component analysis. The respiratory events (ERs -apneas/hypopneas) are detected as a reduction in the resulting instantaneous amplitude and accounted in the respiratory event index (IER), which is a severity indicator similar to the official apnea-hypopnea index (AHI). The respiratory signals extracted from PBS are analyzed first by clustering the information coming from channel pairs and then using the eight channels. The IER performance is compared with the AHI for different severity categories. For the diagnosis of healthy and pathological patients, we obtain a sensitivity, specificity, and accuracy of 92%, 100% and 96%, respectively using two or eight PBS channels. These results suggest the possibility to propose PBS as an alternative tool for SAHS diagnosis in a home environment.