Key result
A deep learning-based noise detect and drop-based filter significantly enhanced ECG classification performance, outperforming noisy data by >20% and blind filtering by >40% with a low latency of 0.5 ms.
Why the study?
ECG signals from wearable sensors are often contaminated by noise that impairs automated analysis, while traditional blind filtering may be suboptimal.
A deep learning-based ECG noise detector improves automated ECG classification performance and reduces latency for wearable sensors, enhancing reliability for real-time IoMT applications.
Answers are grounded in this paper and cite their sources.
Blind filtering may degrade automatic ECG analysis accuracy in wearables; leaves open optimal noise-aware filtering strategies.
The Electrocardiogram (ECG) provides essential information for analysing heart activity. However, the ECG signal is often contaminated with different types of noise, such as baseline wander, power line interference, and muscle artefacts, which severely affect the performance of automatic ECG analysis. Automated analysis is crucial as it allows for rapid, consistent, and accurate interpretation of ECG data, which is essential for timely diagnosis and treatment of cardiovascular conditions. This paper introduces a deep learning (DL) based ECG noise detector in contrast to traditional blind filtering and assesses the impact on the performance of DL-based automated ECG analysis on wearable ECG sensors (e.g. Polar H10). Experimental results demonstrate that our noise detect and drop-based filter before automatic ECG analysis significantly enhances ECG classification performance and outperforms compared to noisy (more than 20%) and blind filtering (more than 40%). Additionally, the proposed noise detection and drop-based filtering exhibits a low latency of 0 . 5 ms , demonstrating its suitability for real-time applications and potential to reduce latency within the Internet of Medical Things (IoMT) settings. Further, we have integrated our noise detect and drop-based filter into a smartphone using a wearable ECG sensor via Bluetooth Low Energy (BLE). This real-world test validates its suitability for practical IoMT uses. The experimental results indicate that the automated ECG analysis based on DL can be more reliable and robust in the presence of noise, bringing the technology closer to standard clinical practice.
No takes yet. Share an insight, caveat, or question.
Rahman et al. (2025) studied ECG analysis for cardiovascular conditions. Deep learning-based ECG noise detector (noise detect and drop-based filter) vs. Noisy data and traditional blind filtering was evaluated on ECG classification performance. A deep learning-based noise detect and drop-based filter significantly enhanced ECG classification performance, outperforming noisy data by >20% and blind filtering by >40% with a low latency of 0.5 ms.
Synapse has enriched 10 related papers. Consider them for comparative context: