Multiobjective Optimization of Interpretable Fuzzy Systems and Applicable Subjects for Fast Estimation of Obstructive Sleep Apnea-hypopnea Severity

This paper proposes an interpretable fuzzy estimation system (IFES) for fast estimation of severity of obstructive sleep apnea-hypopnea syndrome (OSAHS) using three easily available physiological variables: neck circumference, waist circumference, and average blood pressure after waking up. The IFES...

Full description

Saved in:
Bibliographic Details
Published inIEEE transactions on fuzzy systems Vol. 31; no. 7; pp. 1 - 13
Main Authors Juang, Chia-Feng, Pan, Guan-Ren, Huang, Wei-Chang, Wen, Chih-Yu, Wu, Ming-Feng
Format Journal Article
LanguageEnglish
Published New York IEEE 01.07.2023
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects
apnea-hypopnea index (AHI)
Estimation
evolutionary fuzzy systems
Fuzzy neural networks
Fuzzy systems
Genetic algorithms
Hypertension
Indexes
interpretable AI
Medical diagnostic imaging
multiobjective optimization
Multiple objective analysis
Obstructive sleep apnea
Optimization
Performance evaluation
Screening
Sleep apnea
Online AccessGet full text

Cover

More Information
Summary:This paper proposes an interpretable fuzzy estimation system (IFES) for fast estimation of severity of obstructive sleep apnea-hypopnea syndrome (OSAHS) using three easily available physiological variables: neck circumference, waist circumference, and average blood pressure after waking up. The IFES aims to decrease the waiting list of polysomnography (PSG) so that OSAHS treatment can be performed as soon as possible. The IFES is optimized to achieve three objectives: high estimation accuracy, the largest number of applicable subjects, and high model interpretability. The applicable subjects are determined by evaluating the influence of six screening factors, such as smoking, hypertension, and sleep efficiency, on the estimation performance. This paper finds solutions of this multiobjective optimization problem using a multiobjective genetic algorithm. A total of 1197 participants are enrolled with the five-fold cross-validation scheme employed to evaluate estimation performance. Experimental results show that the proposed method successfully finds the influence of different screening factors and the found IFESs outperform different estimation models used for comparison.
ISSN:1063-6706
1941-0034
DOI:10.1109/TFUZZ.2022.3222033