Mamba model improves ECG signal cleaning for longer recordings
DR-net-Mamba: Selective State-Space Modeling for Long-Range ECG Time-Series Denoising
Machine LearningArtificial Intelligence
Summary
Long ECG recordings can have a lot of noise that makes it hard to understand heart signals correctly. The authors created a new method called DR-net-Mamba that cleans these signals better by combining two ways of looking at the data: one focusing on small details and the other on bigger time patterns, without slowing down too much. Their tests showed it works better than previous methods, especially with very long or noisy recordings, and it helped computers identify heart problems more accurately. The biggest improvements happened with certain heart signal changes that need broad context to detect well.
What this means in practice
- •For ambulatory care teams: Improve quality of long-term ECG monitoring by better cleaning of noisy signals to support more reliable heart diagnosis.
- •For medical device developers: Integrate efficient denoising methods that handle long ECG recordings without high computational costs, enabling advances in wearable heart monitors.$Commercial implications: Enables building commercial wearable ECG devices with improved signal quality and longer recording capabilities due to efficient long-range noise removal.
Authors
Basile Morel, Samuel Ruiperez-Campillo, Andreas P. Streich, Julia E. Vogt, Thomas Hofmann
Abstract
Electrocardiogram (ECG) recordings are corrupted by non-stationary noise sources that degrade diagnostic reliability, particularly in ambulatory and long-duration recordings. Deep learning denoisers exist, but convolutional architectures are limited by their receptive field, transformer-based models scale quadratically with sequence length, and diffusion-based approaches incur prohibitive inference cost. We propose a Mamba-augmented model that inserts selective state-space blocks at the convolutional bottleneck, combining local feature extraction with long-range temporal modeling at linear complexity. We comprehensively evaluate the proposed model with respect to reconstruction fidelity, noise robustness, recording-length scaling, and downstream diagnostic classification across over 40 pathology classes. On synthetic and real datasets, our model achieves the highest SNR and lowest RMSE, with the Mamba advantage increasing with sequence length and in low-SNR regimes. On classification with two independent classifiers, the proposed Mamba-based models achieve the best macro AUROC among all denoisers and improve over their convolutional base models. Calibration is more nuanced and classifier-dependent: denoising improves Binary Cross-Entropy and Brier score on Inception1D but often fails to beat the noisy input on ResNet1D-Wang, and the lead-specific Mamba variant is the only denoiser to improve both calibration metrics over the noisy baseline on both classifiers. Per-class analysis reveals a morphology-dependent benefit: Mamba substantially improves ST/T-change diagnoses, which depend on broad, context-sensitive waveforms.