Electrocardiogram-Derived Respiration: A Systematic Review of Biomedical Signal Processing Approaches

Electrocardiogram-derived respiration (EDR) allows estimation of respiratory activity from electrocardiogram (ECG), providing a solution for continuous cardiorespiratory monitoring using wearable devices. This systematic review focuses on signal processing methods that derive EDR from single-lead ECG recordings. A literature search was conducted on Scopus and PubMed following PRISMA guidelines, targeting open-access studies published between 2015 and 2025. From an initial set of 226 articles, 10 studies met the inclusion criteria. The selected studies were analyzed in terms of dataset characteristics, EDR extraction techniques, signal processing complexity, extracted respiratory features, and validation metrics. Data sources included 13 databases, 9 of which are publicly available, with heterogeneous subject populations. EDR extraction methods were categorized into amplitude modulation, frequency modulation, statistical, slope-based, and area-based approaches. Breath rate was the dominant extracted feature, followed by waveform fidelity and other parameters such as cardiorespiratory interaction. Validation strategies were highly heterogeneous; only partial convergence on mean absolute error for BR. Overall, the review reveals high methodological heterogeneity, limited standardization of validation protocols, and scarce real-world wearable testing. These findings highlight the need for robust, standardized evaluation frameworks and the development of methods optimized for single-lead wearable ECG to enable reliable, practical cardiorespiratory monitoring.