A new paper on arXiv describes a deep-learning-based Signal Quality Assessment (SQA) model for ambulatory electrocardiograms (ECG). The model is designed to distinguish between clean and noisy signals collected in real-world, ambulatory settings rather than in controlled clinical environments.
The authors report that the model was trained on data from the Copenhagen Center for population studies, though the abstract does not detail the full dataset composition. The approach combines deep learning with ambulatory context-awareness, suggesting that the model accounts for the conditions under which wearable ECG recordings are typically made.
The paper evaluates the model's ability to separate clean from noisy ECG traces, which is a key step for improving downstream analysis of wearable heart data. No comparison with other SQA methods is described in the abstract, so the reported performance should be read as the authors' initial assessment.