Key result
Machine learning approaches for wearable ECG lack a single superior architecture, driven primarily by dataset characteristics.
Why the study?
Wearable and mobile ECG single- or few-lead recordings and susceptibility to motion artefacts complicate clinical-grade interpretation despite expanded access to rhythm monitoring.
Future progress in wearable ECG machine learning depends on evaluation rigor, representative datasets, and reproducible preprocessing rather than architectural novelty.
Answers are grounded in this paper and cite their sources.
Dataset quality may outweigh architecture in wearable ECG ML; leaves open standardized benchmarks before clinical adoption.
Wearable and mobile electrocardiography (ECG) has rapidly expanded access to rhythm monitoring outside of clinical settings, but the single-or few-lead nature of these recordings and their susceptibility to motion artefacts complicate clinical-grade interpretation. This review synthesises recent machine learning (ML) approaches for wearable ECG across three problem domains -signal-quality assessment (SQA), single-lead interpretation, and multi-lead interpretation -covering preprocessing pipelines, quality handling strategies, and 18 benchmark datasets. Across studies, no single architectural paradigm consistently outperforms others when evaluated fairly: classical feature-based methods remain competitive for rhythm tasks under resource constraints, while deep discriminative models show stronger cross-database generalisation when training data are sufficient. The most consistent predictor of reported performance is dataset characteristics -particularly scale, demographic composition, and whether signals were acquired from wearable devices in free-living conditions -rather than architectural choice. Three recurring evaluation deficiencies are identified: subject-level data leakage inflating classification metrics, SQA performance reported without acceptance rates, and reconstructed ECG leads validated by waveform similarity alone without downstream diagnostic concordance. A structured comparative summary of all reviewed methods is provided. The analysis clarifies that future progress depends primarily on evaluation rigour, representative datasets, and reproducible preprocessing rather than on architectural novelty alone.
No takes yet. Share an insight, caveat, or question.
Bartosz Puszkarski (2026) conducted a review in Wearable and mobile electrocardiography (ECG). Machine learning approaches was evaluated. Machine learning approaches for wearable ECG showed no single architectural paradigm consistently outperforms others, with performance primarily driven by dataset characteristics.
Synapse has enriched 10 related papers. Consider them for comparative context: