Quantifying Loop Gain using Dynamical Modelling of Ventilatory Control - Data and Code

Abstract

This repository provides data and code to accompany the manuscript, "Unravelling Sleep Apnea Dynamics: Quantifying Loop Gain using Dynamical Modeling of Ventilatory Control". The manuscript describes an an automated method for quantifying Loop Gain (LG) from respiratory inductance plethysmography signals to enhance precision management of sleep apnea. We analyzed data from 465 patients, including 400 from Massachusetts General Hospital and 65 heart failure patients. Our method accurately estimated LG across diverse apnea phenotypes. Patients with higher central apnea index, high self-similarity, or heart failure exhibited significantly higher median LG values (0.19, 0.27, and 0.41 respectively) compared to those with obstructive apnea (median LG = 0.11-0.14; p < 0.001). Additionally, LG was significantly elevated during non-rapid eye movement sleep and at higher altitudes. This automated LG estimation method provides a scalable, non-invasive tool for endotyping in sleep apnea to support personalized management strategies.

Background

Sleep apnea affects millions of people globally, with significant implications for cardiovascular health, cognitive function, and quality of life. Accurate identification of apnea phenotypes, such as obstructive sleep apnea (OSA) and central sleep apnea (CSA), is critical for effective treatment.

Ventilatory control, influenced by factors like chemoreflexes, lung mechanics, and arousal thresholds, plays a crucial role in apnea severity. Loop gain (LG) is a key measure of this control, indicating the respiratory system's sensitivity to changes in oxygen and carbon dioxide levels. High loop gain (HLG) signifies a more reactive system, potentially leading to breathing instability.

This study presents a model-based framework for automated LG assessment using respiratory inductance plethysmography signals.

Methods

Retrospective analysis of polysomnography data was approved by the MGH (protocol #2013P001024) and BIDMC Institutional Review Boards (#2016P000058) with waiver of informed consent. Data was collected from Massachusetts General Hospital sleep laboratory between 2008 and 2022, and from heart failure patients and high-altitude studies at Beth Israel Deaconess Medical Centre.

Cases included those with an apnea-hypopnea index >15 and ≥4 hours of recorded sleep. We defined four ventilatory control subgroups: REM OSA, NREM OSA, High CAI, and SS OSA. Additionally, a "SS range" group was selected to encompass the full spectrum of self-similarity levels.

Abdominal RIP signals were filtered, normalized, and segmented into 8-minute windows. LG estimation employed an augmented Mackey-Glass equation and an expectation-maximization algorithm. Simulation experiments on synthetic breathing data with known parameter values were conducted to quantify the accuracy of our parameter estimates.

Data Description

The dataset includes 465 patients (400 from Massachusetts General Hospital and 65 heart failure patients), along with altitude data from 8 patients with recordings at sea level, 5,000, 8,000, and 13,000 feet. From the MGH recordings with an AHI >15 and ≥4 hours of sleep, we sampled 100 patients from each of our defined ventilatory control subgroups. The heart failure group included all patients with AHI >10 and left ventricular ejection fraction <50%.

Usage Notes

Data and code to generate all results and figures from the publication are provided here, where you can find the code to reproduce all figures and tables for the paper.

Ethics

This study of human subjects was approved by the MGH (protocol #2013P001024) and BIDMC Institutional Review Boards (#2016P000058) with waiver of informed consent for the retrospective analysis of clinically acquired PSG data. All data is deidentified.

Acknowledgements

The authors would like to acknowledge contributions of Eline Oppersma, Dirk W Donker, and Nancy S Redeker.

Conflicts of Interest

Dr. Westover is a co-founder of and holds equity in Beacon Biosignals. Beacon Biosignals did not contribute funding nor played any role in the study. Dr. Thomas holds a licensed patent to MyCardio, LLC, for the use of ECG/PPG spectrogram technology to assess sleep quality, sleep apnea, and sleep apnea endotypes; has submitted a patent for Enhanced Expiratory Rebreathing Space (EERS), a CO₂-regulating treatment for central sleep apnea; and serves as a consultant for Guidepoint and GLG Councils. Drs. Thomas, Oppersma, and Westover, along with Thijs Nassi, have jointly submitted a patent for the detection and quantification of respiratory self-similarity.