DR-net-Mamba: Selective State-Space Modeling for Long-Range ECG Time-Series Denoising

Basile Morel; Samuel Ruiperez-Campillo; Andreas P. Streich; Julia E. Vogt; Thomas Hofmann

cs.LGSep 28, 2026

DR-net-Mamba: Selective State-Space Modeling for Long-Range ECG Time-Series Denoising

Authors: Basile Morel, Samuel Ruiperez-Campillo, Andreas P. Streich, Julia E. Vogt, Thomas Hofmann

Organizations: Department of Computer Science at ETH Zurich, Universitätstrasse 6, 8092 Zürich, Switzerland

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.

Figures & tables

Fig. 1: Overall Pipeline: clean ECG records (a) are corrupted with noise (b), denoised (c), and classified (d) to assess whether reconstruction gains translate to diagnostic accuracy (e).

Fig. 2: Architecture of UNet-Mamba1-3B. Three Mamba blocks, each preceded by layer normalization, go into the convolutional bottleneck to capture temporal dependencies beyond the ${\sim}1.5$ s receptive field of the encoder. The Mamba layer follows [ 25 ] .

Fig. 3: Reconstruction SNR (dB) on synthetic (a), PTB-XL (b), and European ST-T (c) datasets. Empirical noisy SNR = 0.96 dB (medium noise setting, see Supplementary Material B-C ). Analogous RMSE results are reported in Figure S8 in Supplementary Material C-D .

Fig. 4: Reconstruction SNR (dB) as a function of input SNR for each noise type (EM, BW, MA, AWGN) and the combined condition on European ST-T. Grey-shaded areas denote input SNR values outside clinically plausible ranges [ 58 ] . Shaded bands indicate 95% bootstrap CIs.

Fig. 5: Reconstruction SNR (dB) as a function of recording length for UNet and UNet-Mamba (without signal compression), on synthetic (a) and European ST-T (b). Shaded bands indicate 95% bootstrap CIs.

Fig. 6: Downstream diagnostic classification (macro and superdiagnostic AUROC) on PTB-XL using ResNet1D-Wang, comparing denoised signals against clean and noisy baselines (strong noise setting).

Classifier trained on	Test input	AUC
Clean	Clean	$0.925\ [0.916,0.932]$
Clean	Noisy	$0.879\ [0.865,0.892]$
Noisy	Noisy	$0.923\ [0.913,0.930]$
Clean	Denoised (Mamba1-3B Lead Aware)	$0.909\ [0.901,0.918]$

TABLE I: Downstream classifier AUROC under strong noise.

	CD	HYP	MI	NORM	STTC	Overall
	AUROC					Sensitivity	Specificity	F1
UNet-Mamba1-3B	.925 $\pm$ .01	.883 $\pm$ .02	.909 $\pm$ .02	.926 $\pm$ .01	.880 $\pm$ .02	.789 $\pm$ .02	.840 $\pm$ .01	.703 $\pm$ .02
DR-net-Mamba1-3B	.921 $\pm$ .01	.887 $\pm$ .02	.909 $\pm$ .02	.926 $\pm$ .01	.881 $\pm$ .02	.794 $\pm$ .02	.841 $\pm$ .01	.706 $\pm$ .02
UNet-Mamba1-3B (LS)	.920 $\pm$ .01	.887 $\pm$ .02	.910 $\pm$ .01	.928 $\pm$ .01	.887 $\pm$ .02	.808 $\pm$ .02	.839 $\pm$ .01	.715 $\pm$ .02
DRNN	.900 $\pm$ .02	.885 $\pm$ .02	.895 $\pm$ .02	.910 $\pm$ .02	.824 $\pm$ .02	.783 $\pm$ .02	.780 $\pm$ .01	.651 $\pm$ .02
DAE	.860 $\pm$ .02	.854 $\pm$ .03	.874 $\pm$ .01	.907 $\pm$ .01	.848 $\pm$ .02	.709 $\pm$ .02	.819 $\pm$ .01	.638 $\pm$ .02

TABLE II: Per-superdiagnostic-class AUROC and overall macro-averaged sensitivity, specificity, and F1 on PTB-XL

Fig. 7: Per-diagnostic-class downstream AUROC (Inception1D). (a) AUROC delta when adding a Mamba bottleneck to the UNet, all other design choices held constant. (b) Absolute AUROC of the lead-specific UNet-Mamba; color indicates improvement over the noisy baseline. In both panels, diagnostic classes with fewer than five segments are omitted. See Supplementary Material C-L for abbreviations.

Appendix figures & tables29 assets

Supplementary material from the paper’s appendix.

Appendix

$l$	Layer Type	$s_{l}$	$k_{l}$	$\displaystyle\prod_{i=1}^{l-1}s_{i}$	$\displaystyle(k_{l}-1)\prod_{i=1}^{l-1}s_{i}$
1	Conv	1	25	1	24
2	Conv	1	25	1	24
3	Conv	1	25	1	24
4	AvgPool	5	5	1	4
5	Conv	1	15	5	70
6	Conv	1	15	5	70

Appendix

TABLE S1: Per-layer receptive field computation for the UNet encoder.

Fig. S1: Test RMSE over training steps for UNet-Mamba with and without the learnable log-compression layer. Removing compression yields faster initial convergence but a worse final optimum.

Fig. S2: Peak GPU memory allocation as a function of ECG signal length (1–5 s at 360 Hz, batch size 2). A : inference (forward pass). B : training (forward + backward). MECGE training requires substantially more memory than UNet-Mamba due to STFT-domain processing.

	UNet / IMUNet / DAE /	MECGE	UNet-Mamba
	DRNN / DR-net-S2		(ours)
Optimizer	Adam	AdamW	Adam
Learning rate	$10^{-3}$	$10^{-4}$	$8\times 10^{-4}$
LR scheduler	ReduceLROnPlateau	Exp. decay	Cosine + warmup
Warmup epochs	—	—	10
$T_{\max}$ (cosine)	—	—	400

Appendix

TABLE S2: Training hyperparameters for baseline and proposed models.

Fig. S3: Training loss (top row) and test SNR (bottom row) over training steps across all three datasets: (A) European ST-T, (B) synthetic, (C) PTB-XL. Stage 1 models on the left of each panel; Stage 2 (DR-net) on the right.

	P	Q	R	S	T
$c_{a_{i}}$	1.2	$-5.0$	30.0	$-7.5$	0.75
$\delta_{a_{i}}$	0.6	0.2	0.0	1.0	0.35
$c_{b_{i}}$	0.25	0.1	0.1	0.1	0.4
$\delta_{b_{i}}$	0.1	0.1	0.0	0.0	0.0

Appendix

TABLE S3: Synthetic ECG simulation parameters for the PQRST morphology.

Config	BW	MA	EM	AWGN	Combined SNR
Light	5.0	10.0	15.0	25.0	3.46
Medium	2.5	7.5	12.5	22.5	0.96
Strong	0.0	5.0	10.0	20.0	$-1.54$

Appendix

TABLE S4: Noise configurations. Per-type SNR values (dB) are drawn from ranges recommended by [ 58 ] .

Fig. S4: Empirical vs. theoretical SNR (left) and RMSE (right) across three noise configurations. Boxplots show per-segment distributions; dashed lines indicate theoretical expected values.

Method	Split Length	Epochs	Runtime
UNet-Mamba1-3B	14400	180	41m 45s
UNet-Mamba1-3B	7200	256	1h 4m 52s
UNet-Mamba1-3B	3600	250	1h 50m 39s
UNet-Mamba1-3B	1800	286	7h 18m 38s
UNet	14400	102	12m 33s
UNet	7200	105	21m 22s

Appendix

TABLE S5: Training details for the synthetic dataset across split lengths. Trained on a single NVIDIA RTX A5000.

Method	Split Length	Epochs	Runtime
UNet-Mamba1-3B	14400	263	32m 13s
UNet-Mamba1-3B	7200	285	39m 44s
UNet-Mamba1-3B	3600	274	1h 7m 7s
UNet-Mamba1-3B	1800	113	1h 34m 27s
UNet	14400	257	17m 13s
UNet	7200	206	21m 56s

Appendix

TABLE S6: Training details for the European ST-T dataset across split lengths. Trained on a single NVIDIA RTX A5000.

Fig. S5: Train loss curves for the recording-length experiment on the synthetic dataset.

Fig. S6: Train loss curves for the recording-length experiment on the European ST-T dataset.

Fig. S7: Reconstruction SNR (top) and RMSE (bottom) on PTB-XL as a function of the number of training folds.

Fig. S8: Reconstruction RMSE on synthetic (A), PTB-XL (B), and European ST-T (C) datasets. Empirical noisy SNR = 0.96 dB (medium noise setting, see Supplementary Material B-C ). This figure supplements Figure 3 .

Baseline	UNet Mamba1-3B	DR-net Mamba1-3B
DRNN	$+11.922\ [+11.761,+12.098]$	$+12.546\ [+12.355,+12.720]$
DAE	$+6.657\ [+6.492,+6.811]$	$+7.281\ [+7.116,+7.440]$
UNet	$+2.551\ [+2.414,+2.699]$	$+3.175\ [+3.025,+3.332]$
DR-net-UNet	$+2.538\ [+2.386,+2.682]$	$+3.162\ [+3.015,+3.316]$
IMUNet	$+1.172\ [+1.019,+1.309]$	$+1.796\ [+1.646,+1.954]$
DR-net-IMUNet	$+0.294\ [+0.148,+0.440]$	$+0.918\ [+0.785,+1.060]$

Appendix

TABLE S7: Pairwise comparison with 95% confidence intervals for the synthetic test set (

n=256

source recordings). DR-net Mamba1-3B outperforms every baseline (smallest gain

+0.92

dB SNR over DR-net-IMUNet, CI

[+0.79,+1.06]

). See Figure 3 a.

Baseline	UNet Mamba1-3B	DR-net Mamba1-3B
DRNN	$+3.955\ [+3.889,+4.019]$	$+4.854\ [+4.785,+4.926]$
DAE	$+4.385\ [+4.297,+4.473]$	$+5.284\ [+5.193,+5.382]$
UNet	$+2.409\ [+2.359,+2.460]$	$+3.308\ [+3.244,+3.366]$
DR-net-UNet	$+0.206\ [+0.162,+0.252]$	$+1.105\ [+1.062,+1.148]$
IMUNet	$+1.063\ [+1.020,+1.104]$	$+1.962\ [+1.911,+2.016]$
DR-net-IMUNet	$-0.070\ [-0.116,-0.022]$	$+0.828\ [+0.778,+0.882]$

Appendix

TABLE S8: Pairwise comparison with 95% confidence intervals for PTB-XL (

n=1{,}639

recordings, 10 s each at 360 Hz). DR-net Mamba1-3B outperforms every baseline (smallest gain

+0.83

dB SNR over DR-net-IMUNet, CI

[+0.78,+0.88]

). See Figure 3 b.

Baseline	UNet Mamba1-3B	DR-net Mamba1-3B
DRNN	$+3.650\ [+3.489,+3.813]$	$+4.878\ [+4.712,+5.055]$
DAE	$+4.966\ [+4.801,+5.113]$	$+6.194\ [+6.014,+6.380]$
UNet	$+2.078\ [+1.940,+2.226]$	$+3.307\ [+3.157,+3.453]$
DR-net-UNet	$-0.784\ [-0.930,-0.622]$	$+0.445\ [+0.315,+0.592]$
IMUNet	$+1.391\ [+1.232,+1.550]$	$+2.619\ [+2.428,+2.815]$
DR-net-IMUNet	$-0.887\ [-1.045,-0.713]$	$+0.342\ [+0.198,+0.493]$

Appendix

TABLE S9: Pairwise comparison with 95% confidence intervals for the ST-T database (

n=256

recordings, 40 s). DR-net Mamba1-3B outperforms every baseline (smallest gain

+0.34

dB SNR over DR-net-IMUNet, CI

[+0.20,+0.49]

). See Figure 3 c.

Baseline	UNet	DR-net- UNet	UNet Mamba1-3B (No Comp)	DR-net Mamba1-3B (No Comp)	UNet Mamba1-3B	DR-net Mamba1-3B
UNet	—	$+2.9$ $[+2.7,+3.0]$	$+1.2$ $[+1.1,+1.3]$	$+3.3$ $[+3.2,+3.4]$	$+2.1$ $[+1.9,+2.2]$	$+3.3$ $[+3.2,+3.5]$
DR-net- UNet		—	$-1.6$ $[-1.8,-1.5]$	$+0.5$ $[+0.3,+0.6]$	$-0.8$ $[-0.9,-0.6]$	$+0.4$ $[+0.3,+0.6]$
UNet Mamba1-3B (No Comp)			—	$+2.1$ $[+2.0,+2.2]$	$+0.9$ $[+0.7,+1.0]$	$+2.1$ $[+1.9,+2.2]$
DR-net Mamba1-3B (No Comp)				—	$-1.2$ $[-1.4,-1.1]$	$-0.0$ $[-0.1,+0.1]$
UNet Mamba1-3B					—	$+1.2$ $[+1.1,+1.3]$

Appendix

TABLE S10: Ablation study separating the effects of Mamba, DR-net, and log compression. Mean differences and 95% confidence intervals are reported from paired bootstrap analysis with

B=1000

resamples with replacement.

BW	MA	EM	AWGN
2.5	7.5	12.5	22.5

Appendix

TABLE S11: Per-type SNR (dB) for the lower (medium) downstream noise setting.

Fig. S9: Downstream diagnostic classification (macro and superdiagnostic AUROC) on PTB-XL under the lower noise setting, using Inception1D and ResNet1D-Wang.

	CD	HYP	MI	NORM	STTC
UNet Mamba1-3B	.926 $\pm$ .02	.888 $\pm$ .02	.916 $\pm$ .01	.931 $\pm$ .01	.888 $\pm$ .02
DRNET Mamba1-3B	.927 $\pm$ .02	.895 $\pm$ .02	.915 $\pm$ .02	.933 $\pm$ .01	.889 $\pm$ .02
UNet Mamba1-3B (LS)	.924 $\pm$ .01	.894 $\pm$ .02	.918 $\pm$ .02	.935 $\pm$ .01	.899 $\pm$ .02
DRNN	.911 $\pm$ .01	.887 $\pm$ .02	.903 $\pm$ .01	.919 $\pm$ .01	.857 $\pm$ .02
DAE	.863 $\pm$ .02	.853 $\pm$ .02	.886 $\pm$ .02	.909 $\pm$ .01	.859 $\pm$ .02
UNet	.919 $\pm$ .01	.877 $\pm$ .02	.913 $\pm$ .01	.926 $\pm$ .01	.881 $\pm$ .02

Appendix

TABLE S12: Per-superdiagnostic-class AUROC on PTB-XL (lower noise, Inception1D). Bold/underlined: best/second-best.

	Sensitivity	Specificity	F1
UNet Mamba1-3B (ours)	.785 $\pm$ .02	.862 $\pm$ .01	.723 $\pm$ .01
DRNET Mamba1-3B (ours)	.788 $\pm$ .02	.865 $\pm$ .01	.727 $\pm$ .01
UNet Mamba1-3B (Lead aware) (ours)	.791 $\pm$ .02	.861 $\pm$ .01	.725 $\pm$ .01
DAE	.698 $\pm$ .02	.846 $\pm$ .01	.653 $\pm$ .02
DRNN	.784 $\pm$ .02	.819 $\pm$ .01	.682 $\pm$ .02
MECGE	.760 $\pm$ .02	.846 $\pm$ .01	.693 $\pm$ .02

Appendix

TABLE S13: Macro-averaged sensitivity, specificity, and F1 on PTB-XL (lower noise, Inception1D). Bold/underlined: best/second-best.

	Sensitivity	Specificity	F1
UNet Mamba1-3B (ours)	.797 $\pm$ .02	.853 $\pm$ .01	.722 $\pm$ .01
DRNET Mamba1-3B (ours)	.800 $\pm$ .02	.858 $\pm$ .01	.727 $\pm$ .02
UNet Mamba1-3B (Lead Specific) (ours)	.798 $\pm$ .02	.858 $\pm$ .01	.728 $\pm$ .02
DAE	.733 $\pm$ .02	.831 $\pm$ .01	.668 $\pm$ .02
DRNN	.787 $\pm$ .02	.808 $\pm$ .01	.679 $\pm$ .02
MECGE	.786 $\pm$ .02	.826 $\pm$ .01	.691 $\pm$ .02

Appendix

TABLE S14: Macro-averaged sensitivity, specificity, and F1 on PTB-XL (strong noise, ResNet1D-Wang). Bold/underlined: best/second-best.

Fig. S10: Downstream diagnostic classification (macro and superdiagnostic AUROC) on PTB-XL using Inception1D, comparing denoised signals against clean and noisy baselines (strong noise setting).

	CD	HYP	MI	NORM	STTC
Sensitivity
UNet-Mamba1-3B	.822 $\pm$ .03	.526 $\pm$ .05	.865 $\pm$ .03	.918 $\pm$ .02	.792 $\pm$ .03
DR-net-Mamba1-3B	.834 $\pm$ .04	.534 $\pm$ .05	.854 $\pm$ .04	.917 $\pm$ .02	.801 $\pm$ .04
UNet-Mamba1-3B (LS)	.816 $\pm$ .03	.556 $\pm$ .05	.815 $\pm$ .03	.943 $\pm$ .01	.824 $\pm$ .04
DAE	.640 $\pm$ .05	.299 $\pm$ .07	.889 $\pm$ .03	.891 $\pm$ .02	.769 $\pm$ .04
DRNN	.822 $\pm$ .03	.489 $\pm$ .06	.887 $\pm$ .03	.863 $\pm$ .03	.860 $\pm$ .03

Appendix

TABLE S15: Per-superdiagnostic-class sensitivity, specificity, and F1 on PTB-XL (strong noise, Inception1D). Bold/underlined: best/second-best.

	CD	HYP	MI	NORM	STTC
Sensitivity
UNet-Mamba1-3B	.784 $\pm$ .03	.515 $\pm$ .06	.898 $\pm$ .02	.942 $\pm$ .02	.786 $\pm$ .03
DR-net-Mamba1-3B	.784 $\pm$ .03	.522 $\pm$ .07	.870 $\pm$ .03	.931 $\pm$ .02	.794 $\pm$ .04
UNet-Mamba1-3B (LS)	.780 $\pm$ .03	.541 $\pm$ .06	.841 $\pm$ .03	.961 $\pm$ .01	.803 $\pm$ .04
DAE	.667 $\pm$ .04	.347 $\pm$ .06	.894 $\pm$ .03	.915 $\pm$ .02	.756 $\pm$ .04
DRNN	.762 $\pm$ .03	.507 $\pm$ .06	.902 $\pm$ .02	.898 $\pm$ .03	.850 $\pm$ .03

Appendix

TABLE S16: Per-superdiagnostic-class sensitivity, specificity, and F1 on PTB-XL (strong noise, ResNet1D-Wang). Bold/underlined: best/second-best.

Fig. S11: Per-diagnostic-class downstream AUROC (ResNet1D-Wang). (A) AUROC delta when adding a Mamba bottleneck to the UNet, all other design choices held constant. (B) Absolute AUROC of the lead-specific UNet-Mamba; color indicates improvement over the noisy baseline. In both panels, diagnostic classes with fewer than five segments are omitted. See Table S17 for abbreviations.

Abbreviation	Description
NDT	non-diagnostic T abnormalities
NST_	non-specific ST changes
DIG	digitalis-effect
LNGQT	long QT-interval
NORM	normal ECG
IMI	inferior myocardial infarction

Appendix

TABLE S17: ECG Form Statement Labels

	STTC	CD
Temporal scale	Broad (200–400 ms)	Narrow (R-peak time 45–60 ms)
Bottleneck steps spanned	${\sim}4$ –7	${\sim}1$
Frequency content	Low	High
Context needed	Global (baseline-relative)	Local (QRS-internal)
Mamba effect on denoising	Better noise/signal separation	Over-smoothing of sharp features

Appendix

TABLE S18: Comparison of STTC and CD diagnostic morphologies and their interaction with the Mamba bottleneck.

Explore similar work

May 18, 2026cs.CV

HexagonalWarriorMamba: Superior Threshold-Dependent Multi-label Classification of 12-Lead ECG Cardiac Abnormalities

The accurate automated diagnosis of cardiac abnormalities from 12-lead electrocardiograms (ECGs) is critical for managing cardiovascular disease. However, detecting concurrent conditions remains a challenge for traditional deep learning models, which often have limited ability to model the long-range dependencies inherent in ECG signals. This manuscript proposes HexagonalWarriorMamba (HWMamba), a framework built on the Mamba architecture that processes 12-lead ECGs as single-channel 2D images rather than conventional 1D time series. By integrating a hierarchical architecture with a 2D Selective Scan mechanism, HWMamba is designed to model global context and complex spatial relationships within the data. The model is evaluated on the PhysioNet/Computing in Cardiology Challenge 2021 dataset, which includes 26 diagnostic labels and comprises recordings collected from seven institutions across four countries and three continents. Results demonstrate that HWMamba outperforms current state-of-the-art (SOTA) methods across five key threshold-dependent metrics, including Challenge Score and Subset Accuracy. These improvements provide a balance between strong discriminative capability and effective threshold selection derived from the training data, while maintaining near-SOTA performance in Macro AUROC. This Hexagonal Warrior performance, reflecting consistent performance across multiple evaluation dimensions, positions HWMamba as a robust and versatile approach for multi-label ECG classification.

Huawei Jiang, Husna Mutahira, Shibo Wei +6

Sungkyunkwan University, Department of Computer Science and Engineering, Suwon, 16419, South Korea · Sogang University, Department of Computer Science and Engineering, Seoul, 04104, South Korea · Gwangju Institute of Science and Technology, Department of Biomedical Science and Engineering, Gwangju, 61005, South Korea +4

Electrocardiogram Classification Adult-To-Pediatric Electrocardiogram Transfer Cardiac Mamba-Based Models Topomamba Multi-Label Classification

Ask PDF

Apr 17, 2026eess.SP

MedMamba: Recasting Mamba for Medical Time Series Classification

Medical time series, such as electrocardiograms (ECG) and electroencephalograms (EEG), exhibit complex temporal dynamics and structured cross-channel dependencies, posing fundamental challenges for automated analysis. Conventional convolutional and recurrent models struggle to capture long-range dependencies, while Transformer-based approaches incur quadratic complexity and often introduce redundant interactions that are misaligned with the intrinsic structure of physiological signals. To address these limitations, we propose MedMamba, a principle-driven multi-scale bidirectional state space architecture tailored for medical time series classification. Our design is guided by three key inductive biases of physiological signals: spatial centralization, multi-timescale temporal composition, and non-causal contextual dependency. These principles are instantiated through a lightweight channel-mixing module for cross-channel reparameterization, multi-scale convolutional tokenization for temporal decomposition, and bidirectional Mamba blocks for efficient global context modeling with linear complexity. Extensive experiments on six benchmark datasets spanning EEG, ECG, and human activity signals demonstrate that MedMamba consistently outperforms state-of-the-art methods across diverse modalities. Notably, it achieves 85.97% accuracy on PTB and establishes new state-of-the-art performance on the challenging ADFTD dataset (54.72% accuracy and 52.01% F1-score). Strong results on long-sequence benchmarks, such as SleepEDF, further validate its capability in modeling long-range dependencies. Moreover, MedMamba achieves a speedup of 4.6x in inference, highlighting its practicality for real-time clinical deployment. These results suggest that principle-guided state space modeling offers an effective and scalable alternative to Transformer-based approaches for medical time series analysis.

ZhengXiao He, Huayu Li, Xiwen Chen +4

Department of Electrical & Computer Engineering at the University of Arizona, Tucson, AZ, USA

Time-Series Classification Time Series State Space Models Mamba Topomamba Recurrent Model+1

Ask PDF

Sep 24, 2026cs.AI

Wearable ECG Quality Assessment: A Deep Learning and Ambulatory Context-Awareness Approach

This paper presents and evaluates a Deep Learning-based (DL-based) Signal Quality Assessment (SQA) model to distinguish between clean and noisy ambulatory Electrocardiograms (ECG). The model is trained on Copenhagen Center for Health Technology-Contextualized Arrhythmia Database (CACHET-CADB), which, to the best of our knowledge, is the first ambulatory ECG database with both physical and patient-reported contextual data. The model shows stable performance on different databases such as MIT-databases and the latest PyhsioNet/Cinc Challenge 2021 databases. Subsequently, the paper demonstrates how complicated ECG noise can be investigated by the SQA model and the physical contextual data.

Xiaopeng Mao, Marike Weisbjerg, Sadasivan Puthusserypady

Health Technology department Technical University of Denmark Kongens Lyngby, Denmark

Arrhythmia Classification Deep Learning Signal-To-Noise Ratio Context-Awareness Patients

Ask PDF