Rationale: The main polysomnography (PSG) indices during obstructive sleep apnea syndrome (OSAS), including obstructive respiratory effort (RE) and arousals are closely reflected by specific patterns of vertical mandibular movement (MM) signal. However the potential of MM signal has never been fully explored. Objective and method: We used a new MM sensor coupled with a machine learning based technology (Sunrise, Namur, Belgium) to generate 3 indices from the MM signal recorded in 400 patients with OSAS: total sleep duration in RE (MM-REdt), 2) averaged hourly indices of arousals (MM-Arl) and 3) obstructive respiratory disturbances index (MM-ORDI); then performed a principal component analysis (PCA) to determine how much important these MM derived indices could contribute to the diagnosis and severity stratification compared with the PSG indices. The PCA allows to reduce the dimensionality (400×17) of original data space to only 2 components. Results: As showed in the figure, both MM-ORDI and MM-Arl are well correlated with the key PSG indices and contribute as important as AHI, RDI and ODI to the diagnosis and severity stratification of OSAS. We also found that MM-REdt is an original index which contributes independently and even more importantly than PSG to the diagnosis of OSAS.

2020