Possible predictors of apneic-events during sleep: advanced snoring sound analysis

Snoring is the most common symptom in patients experiencing obstructive sleep apnea (OSA) and is reported by up to 94% of those affected. It has already been extensively proven that snoring of patients with OSA acoustically differs significantly from habitual snoring. For this study, snoring in patients with OSA is analyzed by its different components: pitch, formants (F1-F3), formant-ratio F2/F1, harmonic-noise-ratio (HNR), spectral energy ratio (SER). The objective of this study is to determine whether one or more acoustic parameters exhibit significant alterations or distinctive features immediately preceding an apneic episode, as compared to the patterns observed during simple snoring in individuals with obstructive sleep apnea (OSA). Identifying such preapneic indicators could facilitate the prediction of subsequent apnea events. With further technological advances and greater standardization, acoustical analysis could play a central role in the diagnosis and treatment of snoring and sleep apnea.

Based on the "Munich-Passau Snore Sound Corpus" (MPSSC) as one of the world's largest data sets for snoring sounds, retrospective data, consisting of 219 patients with OSA, were obtained using drug-induced sleep endoscopy (DISE). The mentioned components were retrieved and analyzed with the open source software "Praat". Statistical testing was done with t-tests (Student and Welch) and Mann-Whitney-U-test.

For four parameters (F1, F2/F1, HNR, logSER) there are statistically significant differences between continuous/simple and preapneic snoring sounds.

Preapneic snoring sounds seem to have different characteristics than continuous/simple snoring sounds. With further research, a simple acoustic differentiation for various diagnostic purposes could become possible.