Enhancing Obstructive Apnea Disease Detection Using Dual-Tree Complex Wavelet Transform-Based Features and the Hybrid
Javad Ostadieh1, Mehdi Chehel Amirani1, Morteza Valizadeh2
1Department of Electrical Engineering and, Urmia University, Urmia, Iran.
Journal of Medical Signals and Sensors
|February 12, 2021
Summary
This study presents a novel, low-complexity method for obstructive sleep apnea (OSA) detection, achieving high accuracy. The new technique significantly reduces computational load compared to existing Support Vector Machine (SVM) methods.
Area of Science:
- Biomedical Signal Processing
- Machine Learning for Healthcare
Background:
- Obstructive sleep apnea (OSA) is a high-risk disease, making its detection a critical research area.
- Current detection methods often involve high computational complexity.
Purpose of the Study:
- To evaluate powerful, low-computational signal processing techniques for OSA detection.
- To compare the performance of these techniques against established methods.
Main Methods:
- Utilized Dual-tree complex wavelet transform (DT-CWT) for feature coefficient extraction.
- Extracted eight non-linear features, reduced using the Multi-cluster feature selection (MCFS) algorithm.
- Employed a hybrid K-means, RLS Radial Basis Function (RBF) network for classification.
Main Results:
- Achieved a high OSA detection accuracy of approximately 96%.
- Demonstrated a significant reduction in computational complexity, nearly one-third of SVM-based methods.
Conclusions:
- The proposed hybrid RBF network offers an effective and computationally efficient solution for OSA detection.
- This method presents a viable alternative to complex SVM-based approaches for OSA diagnosis.
