The Respiratory Fluctuation Index: A global metric of nasal airflow or thoracoabdominal wall movement time series to diagnose obstructive sleep apnea. (March 2019)

• Record Type: Journal Article
• Authors: Wang, Fu-Tai; Hsu, Ming-Hung; Fang, Shih-Chin; Chuang, Li-Ling; Chan, Hsiao-Lung
• Is Part Of: Biomedical signal processing and control. Volume 49(2019)
• Page Start: 250
• Page End: 262
• Publication Date: 2019-03
• Subjects: Polysomnography -- Obstructive sleep apnea -- Apnea hypopnea index -- Respiratory fluctuation; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2018.12.015
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A new global measure, the respiratory fluctuation index, to assess obstructive sleep apnea. This parameter considers both degree and frequency of respiratory drops without the need for hand-counting episodes. Generalized quantification of thoracic or abdominal respiratory effort as well as nasal airflow. Respiratory fluctuation index from nasal airflow has a good efficacy to diagnose obstructive sleep apnea. Abstract: Polysomnography (PSG) recordings provide comprehensive physiological data to diagnose obstructive sleep apnea (OSA), a common breathing disorder. OSA severity is usually diagnosed using the apnea hypopnea index (AHI), the frequency of episodes of breathing cessation or reduction. This study proposes a new global measure, the Respiratory Fluctuation Index (RFI), which accurately characterizes the distribution of respiratory drops in a PSG time series without the need for hand-counting episodes. We test two approaches: linear regression models to estimate the hand-scored AHI from one or more RFIs, and threshold detection models to diagnose OSA severity based on a single RFI. Based on PSG data recorded from 60 adults, we find very good agreement between both types of models and the ground truth. Regression models based on the RFI derived from nasal airflow had the best agreement with manually scored AHIs, visualized with Bland-Altman plots. In addition, threshold detection models based on the RFI of nasal airflow achieved the highest sensitivities (78.97%, 87.50% and 92.31%) and specificities (78.26%, 86.46% and 91.82%) in detecting OSA with AHI ≥ 5, AHI ≥ 15 and AHI ≥ 30, respectively. The strong performance of the models demonstrates the efficacy of the proposed RFI metric for OSA assessment.