Classification techniques on computerized systems to predict and/or to detect Apnea: A systematic review.

Area of Science:

• Computational intelligence
• Medical informatics
• Biomedical engineering

Background:

• Sleep apnea syndrome (SAS) is linked to severe health issues like cardiovascular disease, hypertension, and depression.
• Effective detection and management of SAS are crucial for improving patient quality of life.
• Computational intelligence offers promising approaches for SAS analysis and decision-making.

Purpose of the Study:

• To systematically review computational intelligence-based systems for sleep apnea syndrome detection and/or prediction.
• To evaluate the performance, benefits, and challenges of these systems.
• To analyze modeling approaches for decision-making in various SAS scenarios.

Main Methods:

• Systematic literature review focusing on classification models for apnea event detection/prediction.
• Analysis of 45 studies employing diverse classification techniques.
• Clustering of machine learning models using a mind map.

Main Results:

• Machine learning (ML) models were utilized in 85.25% of studies, significantly more than threshold-based methods (14.75%).
• Key ML models included neural networks (44.26%), instance-based learning (11.47%), and dimensionality reduction (8.19%).
• Model accuracy is strongly correlated with effective feature selection.

Conclusions:

• Classification models for SAS should be auto-adaptive and independent of external human input.
• Further high-quality research, including randomized controlled trials and validation on large datasets, is recommended.
• Feature selection is critical for enhancing the accuracy of SAS classification models.