Advancing wakefulness screening of sleep apnea disorder through breathing sound analysis and hardware optimization

Marian University Inc. received a $130,600 Project Grant awarded August 4, 2025, by the National Institute of Biomedical Imaging and Bioengineering (NIBIB) under the Discovery and Applied Research for Technological Innovations to Improve Human Health program (CFDA 93.286). The award supports research and development of an enhanced wakefulness-based sleep apnea screening tool utilizing breathing sound analysis technology. The project addresses the significant public health challenge of obstructive sleep apnea (OSA), which affects over one billion people worldwide but remains substantially underdiagnosed, particularly in underserved populations lacking access to gold-standard polysomnography diagnostics.

The funded initiative will deliver technological improvements to advance the screening tool's clinical utility through three primary objectives: refining the algorithm to reduce anthropometric factor influence and incorporate intra-subject variance using deterministic and non-deterministic machine learning approaches; identifying acoustic features that correlate with and predict the apnea-hypopnea index using advanced signal processing and statistical analysis; and optimizing hardware design and affordability for large-scale implementation. This non-invasive, objective, and accessible screening methodology is designed to improve diagnostic accuracy compared to existing questionnaire-based approaches while expanding accessibility for populations with limited access to traditional sleep diagnostic services. The project performance period extends through July 31, 2027, with work conducted in Indiana.

Abstract

Sleep apnea affects over one billion people worldwide, with obstructive sleep apnea (OSA) accounting for more than 85% of cases. OSA is characterized by repeated episodes of airway obstruction during sleep, leading to fragmented sleep and reduced oxygen levels. OSA is linked to several health complications, including cardiovascular disease, diabetes, and cognitive impairment. Despite its prevalence, OSA remains significantly underdiagnosed, especially in underserved populations, due to limited access to gold-standard diagnostics like overnight polysomnography (PSG). Early diagnosis is essential to prevent long-term health issues and alleviate the associated healthcare burden. Additionally, a quick and objective OSA screening could prevent unnecessary surgical precautions due to the low specificity (~40%) of widely used questionnaires, such as stop-bang. This project focuses on enhancing a wakefulness-based sleep apnea screening tool that utilizes breathing sound analysis to assess sleep apnea risk. The tool offers a non-invasive, objective, and accessible screening method, especially for populations with limited access to traditional sleep studies. However, the tool's screening accuracy, affordability, and usability must be improved for large-scale implementation. We propose three specific aims to address these challenges. Aim 1 seeks to refine the tool's algorithm by reducing the influence of anthropometric factors on sound feature classification and incorporating intra-subject variance to enhance reliability. Deterministic and non-deterministic machine learning algorithms will be integrated to improve classification outcomes. Aim 2 will apply machine learning, advanced signal processing, and statistical analysis techniques to identify acoustic features that correlate with and predict the apnea-hypopnea index (AHI) and other sleep metrics, such as oxygen desaturation index and apnea index. Aim 3 focuses on the development of custom hardware and a microphone chamber for improved breathing sound recording and processing, aiming to create a more portable, cost-effective, and user-friendly device for broad adoption. Our long-term goal is to provide an accessible, wakefulness-based screening tool for early detection of sleep apnea and assessment of sleep quality metrics, particularly in underserved populations and individuals undergoing surgeries requiring full anesthesia. This tool has the potential to reduce diagnostic time and costs, improve early intervention rates, and decrease the risk of untreated sleep apnea leading to severe health complications. The proposed R03 project will generate key insights and models for future clinical trials and broader implementation. By improving the accuracy and usability of this screening tool, the research aims to address critical public health needs, especially in rural or low-resource settings where access to conventional sleep diagnostics is limited. If successful, the tool could significantly impact sleep apnea diagnosis and treatment, leading to better patient outcomes and a reduction in healthcare costs associated with untreated sleep apnea.