The sound of snoring: AI-based acoustic phenotyping of upper airway obstruction in Obstructive Sleep Apnea
- 1Dept. of Otorhinolaryngology, Sleep Surgery Center, Sleep Disorders Center, Istituto Auxologico Italiano IRCCS, Milan, Italy, Dept. of Clinical Sciences and Co
Abstract
BackgroundDrug-Induced Sleep Endoscopy (DISE) is the current reference standard for identifying the anatomical site of upper airway obstruction in obstructive sleep apnea (OSA), but its invasiveness and limited availability restrict its routine use. Because snoring is generated by vibration of the obstructing upper-airway structures, its acoustic characteristics may provide a non-invasive biomarker of the anatomical site of obstruction. This study investigated whether machine learning could reliably distinguish palatal from epiglottic snoring using acoustic information alone.
MethodsA retrospective analysis was performed on 159 DISE recordings obtained from adult patients with moderate, non-positional OSA. Snoring events were independently identified and anatomically classified by two blinded expert examiners. Only events with complete inter-observer agreement were included. The corresponding audio segments were extracted and characterized using spectral, cepstral, temporal and harmonic acoustic features. Feature selection was performed using recursive feature elimination. Support Vector Machine (SVM) and Multilayer Perceptron (MLP) classifiers were developed using a patient-independent nested cross-validation framework.
ResultsThe final dataset comprised 1,759 snoring events, including 1,054 palatal and 705 epiglottic recordings. Both classifiers demonstrated robust discrimination between the two anatomical classes under patient-independent validation, achieving ROC AUC values of 0.90{+/-} 0.03 (SVM) and 0.90{+/-} 0.03 (MLP), with balanced accuracies of 0.83 and 0.81, respectively. The most informative predictors were spectral features, particularly spectral energy distribution, spectral flux, Mel-Frequency Cepstral Coefficients (MFCCs), and spectral flatness, whereas fundamental frequency contributed minimally to classification.
ConclusionsSnoring contains reproducible acoustic information reflecting the anatomical origin of upper-airway obstruction. By combining high-confidence DISE-derived anatomical labels with a rigorous machine learning framework, this study demonstrates the feasibility of non-invasive acoustic phenotyping of clinically relevant obstruction sites. Rather than proposing a novel artificial intelligence algorithm, our work establishes a clinically oriented framework that may support future decision-support tools for patient selection, treatment planning, and multicentre development of comprehensive acoustic phenotyping models.
Related
- Hypertrophic obstructive cardiomyopathy complicated by severe midventricular obstruction treated with percutaneous transluminal septal myocardial ablation using simultaneous contrast imaging with left ventriculography and coronary angiography: a case report
- Airway Obstruction With Severe Hypoxemia Caused by Malformation of an Endotracheal Tube Cuff: A Case Report
- Beyond the Apnea-Hypopnea Index: Cardiovascular and Metabolic Phenotyping of Hypertensive Patients with Obstructive Sleep Apnea—A Narrative Review
- The use of auto-titrating positive airway pressure in children with obstructive sleep apnea: a scoping review
- Nasal and Nasopharyngeal Airway Obstruction in Neonates: A Pictorial Essay.
BibTeX
@misc{leone2026sound,
title = {{The sound of snoring: AI-based acoustic phenotyping of upper airway obstruction in Obstructive Sleep Apnea}},
author = {F. Leone and F. Pietrogiacomi and L. Fiorini and E. Agrimi and F. Vultaggio and A. Bianchi and V. Cesarini and G. Costantini and G. Gnecco and F. Salamanca and F. Mozzanica},
howpublished = {medRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.14.26362626},
url = {https://www.medrxiv.org/content/10.64898/2026.09.14.26362626v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - The sound of snoring: AI-based acoustic phenotyping of upper airway obstruction in Obstructive Sleep Apnea
AU - F. Leone
AU - F. Pietrogiacomi
AU - L. Fiorini
AU - E. Agrimi
AU - F. Vultaggio
AU - A. Bianchi
AU - V. Cesarini
AU - G. Costantini
AU - G. Gnecco
AU - F. Salamanca
AU - F. Mozzanica
PY - 2026
DA - 2026/09/14
PB - medRxiv
DO - 10.64898/2026.09.14.26362626
UR - https://www.medrxiv.org/content/10.64898/2026.09.14.26362626v1
AB - BackgroundDrug-Induced Sleep Endoscopy (DISE) is the current reference standard for identifying the anatomical site of upper airway obstruction in obstructive sleep apnea (OSA), but its invasiveness and limited availability restrict its routine use. Because snoring is generated by vibration of the obstructing upper-airway structures, its acoustic characteristics may provide a non-invasive biomarker of the anatomical site of obstruction. This study investigated whether machine learning could reliably distinguish palatal from epiglottic snoring using acoustic information alone. MethodsA retrospective analysis was performed on 159 DISE recordings obtained from adult patients with moderate, non-positional OSA. Snoring events were independently identified and anatomically classified by two blinded expert examiners. Only events with complete inter-observer agreement were included. The corresponding audio segments were extracted and characterized using spectral, cepstral, temporal and harmonic acoustic features. Feature selection was performed using recursive feature elimination. Support Vector Machine (SVM) and Multilayer Perceptron (MLP) classifiers were developed using a patient-independent nested cross-validation framework. ResultsThe final dataset comprised 1,759 snoring events, including 1,054 palatal and 705 epiglottic recordings. Both classifiers demonstrated robust discrimination between the two anatomical classes under patient-independent validation, achieving ROC AUC values of 0.90{+/-} 0.03 (SVM) and 0.90{+/-} 0.03 (MLP), with balanced accuracies of 0.83 and 0.81, respectively. The most informative predictors were spectral features, particularly spectral energy distribution, spectral flux, Mel-Frequency Cepstral Coefficients (MFCCs), and spectral flatness, whereas fundamental frequency contributed minimally to classification. ConclusionsSnoring contains reproducible acoustic information reflecting the anatomical origin of upper-airway obstruction. By combining high-confidence DISE-derived anatomical labels with a rigorous machine learning framework, this study demonstrates the feasibility of non-invasive acoustic phenotyping of clinically relevant obstruction sites. Rather than proposing a novel artificial intelligence algorithm, our work establishes a clinically oriented framework that may support future decision-support tools for patient selection, treatment planning, and multicentre development of comprehensive acoustic phenotyping models.
ER -