Sleep Apnea Events Recognition Based on Polysomnographic Recordings: A Large-Scale Multi-Channel Machine Learning...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Sleep Medicine

Background:

• Manual polysomnography interpretation for apneic events is time-consuming, costly, and prone to errors.
• Current diagnostic protocols for sleep apnea-hypopnea syndrome require specialized facilities, leading to long waiting times.
• Artificial intelligence (AI) offers a promising solution to enhance the accuracy and efficiency of sleep apnea diagnosis.

Purpose of the Study:

• To develop and validate a machine learning-based approach for the automatic detection of apneic events.
• To improve the diagnostic process for sleep apnea-hypopnea syndrome using AI.
• To establish a more efficient and accurate method for identifying sleep-related breathing disorders.

Main Methods:

• Utilized the Wisconsin Sleep Cohort (WSC) database, a large and diverse dataset of subjects.
• Developed a machine learning model for the automated recognition of apneic events.
• Evaluated the model's performance on event detection and classification of different apnea types.

Main Results:

• Achieved an overall accuracy of 87.2 ± 1.8% for the automatic detection of apneic events.
• Demonstrated significantly higher accuracy compared to existing methods on the same dataset.
• Obtained an overall accuracy of 62.9 ± 4.1% for distinguishing between different types of apnea.

Conclusions:

• The proposed AI approach enhances sleep apnea event recognition, offering improved performance over the state-of-the-art.
• The validated method provides a simple and interpretable way to identify sleep apnea events using a subset of signals.
• This AI-driven solution expands possibilities for sleep apnea diagnosis, potentially increasing healthcare quality and accessibility.