RAPIDEST: A Framework for Obstructive Sleep Apnea Detection

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Sleep Medicine

Background:

• Traditional polysomnography (PSG) for Obstructive Sleep Apnea (OSA) diagnosis is complex, requiring multiple sensors and extensive data analysis.
• The Apnea-Hypopnea Index (AHI) calculation from PSG data is inconvenient for patients and clinicians.
• There is a need for simplified, accurate diagnostic methods for OSA.

Purpose of the Study:

• To propose an AI-based framework, RAPIDEST, for detecting OSA.
• To simplify OSA diagnosis by utilizing only electroencephalogram (EEG) signals.
• To establish a novel rarity score for assessing OSA severity based on sleep stage transitions.

Main Methods:

• Developed the RAre Pattern Identification and DEtection for Sleep-stage Transitions (RAPIDEST) framework.
• Utilized sequences of sleep stages derived from EEG signals.
• Defined a novel rarity score to quantify the unusualness of sleep stage sequences for OSA detection.

Main Results:

• Demonstrated that the rarity score effectively captures unusual sleep stage sequences indicative of OSA.
• Established a significant relationship between the proposed rarity score and the traditional Apnea-Hypopnea Index (AHI).
• Validated the effectiveness of the AI-based framework in detecting OSA using simplified data collection.

Conclusions:

• The RAPIDEST framework offers a significantly simplified approach to OSA diagnosis compared to traditional PSG.
• Using only EEG signals and a rarity score provides a convenient and potentially more accessible method for OSA evaluation.
• The proposed AI-based method shows promise for improving the efficiency and reducing the complexity of OSA diagnosis.