AI-Assisted Awake Endoscopic Video Analysis for Obstructive Sleep Apnea Detection

Area of Science:

• Artificial Intelligence in Medicine
• Medical Imaging Analysis
• Sleep Medicine Diagnostics

Background:

• Obstructive sleep apnea (OSA) is a prevalent sleep disorder linked to severe health issues.
• Current diagnostic methods like polysomnography and DISE are often invasive, costly, or time-consuming.
• There is a need for rapid, non-invasive, and automated diagnostic tools for OSA.

Purpose of the Study:

• To develop a fully automated, AI-assisted platform for detecting obstructive sleep apnea (OSA).
• To utilize nasopharyngoscopic videos acquired during wakefulness for OSA detection.
• To establish diagnostic criteria comparable to the apnea-hypopnea index (AHI).

Main Methods:

• An AI system analyzed flexible nasopharyngoscopic videos of awake, supine patients.
• Image classifiers identified anatomical regions, and segmentation models extracted airway features.
• Support vector regression predicted OSA using variables like airway cross-sectional area and wall/tongue base ratios.

Main Results:

• The AI system achieved high classification accuracy for various anatomical regions (up to 98.5%).
• The prediction model demonstrated 97.14% accuracy on the test set.
• Diagnostic thresholds for OSA were identified, comparable to AHI, with the entire workflow averaging 85 seconds.

Conclusions:

• This study presents the first fully automated AI-based dynamic endoscopic video analysis for OSA detection in awake patients.
• The system non-invasively predicts OSA and identifies potential obstruction sites in real-time.
• This offers a practical outpatient screening tool to identify candidates for further evaluation.