A Lightweight Obstructive Sleep Apnea Detection Method with Millimeter-Wave Radar and Oximeter

Long Pu, Xueqian Wang, Zetao Wang, Zhaoxi Chen, Jian Guan, Gang Li

Confirmation: I have read and agree with the IEEE BHI 2025 conference submission's policy on behalf of myself and my co-authors.

Keywords: obstructive sleep apnea, millimeter-wave radar, oxygen saturation, feature fusion

TL;DR: We propose a sleep monitoring method based on the fusion of millimeter-wave radar and oxygen saturation to accurately detect sleep apnea events, while ensuring the comfort of the monitoring process.

Abstract: Obstructive Sleep Apnea (OSA) is a prevalent disorder characterized by intermittent cessation of breathing during sleep. The established gold standard for OSA diagnosis, Polysomnography (PSG), is uncomfortable for patients. This paper proposes a user-friendly and fine-grained method for detecting OSA events with millimeter-wave radar and oximeter. To adequately fuse the two sensors, we introduce the multi-scale feature extraction strategy and neighboring short-term feature enhancement strategy (Ms&Ne). Key features are extracted at long, medium, and short scales, capturing both long-term characteristics and detailed variations of the signals, effectively addresses the signal misalignment issue due to oxygen desaturation delay. Short-scale features are further incorporated to enhance short-term variation detection. The eXtreme Gradient Boosting (XGBoost) is utilized for tree-based feature interactions. Clinical trials involving 121 patients at Shanghai Sixth People’s Hospital demonstrate that our method achieves an highest F1-score of 0.7713 for OSA detection at the second-by-second level.

Track: 1. Biomedical Sensor Informatics

Registration Id: 9BN5DHXZ69T

Submission Number: 123