Pith. sign in

REVIEW 2 major objections 2 minor 49 references

Personalizing a global deep learning model on individual wearable ECG data improves five-minute-ahead forecasting of atrial fibrillation.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

2026-06-27 12:17 UTC pith:2K3WZNPF

load-bearing objection The paper reports AUROC gains from patient-specific fine-tuning on wearable ECG for 5-minute AF forecasting, but the evaluation setup likely allows temporal leakage from the same recordings. the 2 major comments →

arxiv 2606.10900 v1 pith:2K3WZNPF submitted 2026-06-09 eess.SP

Personalized Deep Learning for Short-Term Forecasting of Impending Atrial Fibrillation from Continuous Wearable ECG Signals

Jangwon Suh , Soonil Kwon , Jungmin Ko , Yun Kwan Kim , Hee Seok Song , Eue-Keun Choi , Wonjong Rhee This is my paper
classification eess.SP
keywords personalized deep learningatrial fibrillation forecastingwearable ECGshort-term predictionfine-tuningambulatory monitoringECG features
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests whether fine-tuning a neural network on a patient's own ECG recordings from wearable devices boosts its accuracy at predicting an impending AF episode five minutes ahead. A single global model is first trained on a large collection of ECG segments, then adapted separately for each person using varying amounts of that person's data across multiple cohorts. Personalized versions reach higher AUROCs than the untouched global model, and the improvement grows as more patient-specific segments are added for fine-tuning. The work also shows that the adapted models rely on different temporal patterns and highlight clinical precursors such as premature atrial complexes.

Core claim

Adapting a global deep learning model via fine-tuning on patient-specific 60-second ECG segments significantly improves AUROC for five-minute-ahead forecasting of atrial fibrillation compared with the non-adapted model across the tested cohorts.

What carries the argument

Patient-specific fine-tuning of the neural network on an individual's own 60-second ECG segments to capture inter-patient variability in AF precursors.

Load-bearing premise

The patient-specific fine-tuning data drawn from the same recording sessions as the test segments is representative of future ECG behavior and does not introduce selection or temporal leakage bias.

What would settle it

A controlled test that uses only ECG data recorded earlier in time for fine-tuning and strictly later independent segments for evaluation, checking whether the reported AUROC gains remain.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • Gains in forecasting accuracy increase with the volume of patient-specific adaptation data.
  • Personalized models exhibit temporal dynamics less dependent on proximity to the AF event than the global model in external cohorts.
  • Feature attributions identify frequent premature atrial complexes and short supraventricular tachycardias as clinically relevant precursors.
  • Pre-AF segments show elevated heart rates and RMSSD values compared with non-AF segments.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same personalization step could be applied to other ambulatory arrhythmia forecasts if similar patient-to-patient ECG differences exist.
  • Continuous monitoring systems might reduce unnecessary alerts by switching to patient-adapted models after an initial calibration period.
  • Testing whether performance holds when fine-tuning uses data from separate earlier visits rather than the same session would clarify real-world deployment value.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 2 minor

Summary. The paper claims that fine-tuning a global deep learning model on patient-specific wearable ECG data significantly improves short-term (5-minute horizon) forecasting of impending atrial fibrillation, with reported AUROC gains from 0.614 to 0.711 on ICENTIA11K and from 0.585 to 0.686 on MobiCARE; benefits scale with adaptation volume, and feature analysis highlights precursors such as elevated heart rate, RMSSD, PACs, and short SVTs.

Significance. If the central empirical comparison holds without leakage or missing methodological details, the result would support personalized models for ambulatory AF surveillance and could inform preventive interventions. The multi-cohort design and volume-dependence analysis are positive features, but the absence of architecture, loss, statistical tests, and explicit temporal blocking limits immediate clinical translation.

major comments (2)
  1. [Methods] Methods (data partitioning and fine-tuning protocol): the description of per-cohort adaptation on ICENTIA11K/IRIDIA-AF/MobiCARE does not specify temporal blocking between fine-tuning segments and held-out test segments drawn from the same continuous recordings; this risks non-prospective leakage of patient-specific state and directly undermines the claim that personalization improves generalization to future ECG behavior.
  2. [Results] Results (AUROC reporting): the abstract and results state AUROC gains without model architecture, loss function, class-imbalance handling, confidence intervals, or statistical tests comparing personalized vs. global models; these omissions make it impossible to assess whether the reported differences (e.g., 0.711 vs. 0.614) are robust or load-bearing for the personalization claim.
minor comments (2)
  1. [Abstract] Abstract: the methods paragraph omits any description of the neural-network architecture, preprocessing pipeline details, or evaluation metrics beyond AUROC.
  2. [Results] Results: the statement that personalization benefits increase with adaptation volume would be strengthened by reporting the exact number of segments or minutes used at each volume level and any associated variance.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive review. We address the two major comments below and will revise the manuscript to improve clarity and completeness.

read point-by-point responses

  1. Referee: [Methods] Methods (data partitioning and fine-tuning protocol): the description of per-cohort adaptation on ICENTIA11K/IRIDIA-AF/MobiCARE does not specify temporal blocking between fine-tuning segments and held-out test segments drawn from the same continuous recordings; this risks non-prospective leakage of patient-specific state and directly undermines the claim that personalization improves generalization to future ECG behavior.

    Authors: We agree that explicit temporal blocking is essential to support the personalization claim. The original manuscript did not provide a sufficiently detailed description of the partitioning protocol. In the revision we will add a dedicated subsection describing how, for each patient, fine-tuning segments were taken exclusively from earlier portions of the continuous recording and test segments from strictly later portions, with a minimum temporal gap to enforce prospective evaluation and eliminate leakage of future state information. revision: yes

  2. Referee: [Results] Results (AUROC reporting): the abstract and results state AUROC gains without model architecture, loss function, class-imbalance handling, confidence intervals, or statistical tests comparing personalized vs. global models; these omissions make it impossible to assess whether the reported differences (e.g., 0.711 vs. 0.614) are robust or load-bearing for the personalization claim.

    Authors: We accept that these implementation and statistical details are required for proper evaluation. The revised manuscript will (i) expand the Methods section with the exact network architecture, loss function, and class-imbalance strategy, (ii) report 95% confidence intervals for all AUROCs, and (iii) include a statistical comparison (DeLong test or bootstrap) between personalized and global models. These additions will be placed in both the main text and a new supplementary table. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical model comparison on held-out segments

full rationale

The paper reports an empirical study comparing a global DL model against patient-specific fine-tuned models using AUROC on held-out 60s ECG segments for 5-min AF forecasting across ICENTIA11K, IRIDIA-AF, and MobiCARE cohorts. No equations, derivations, or self-citations are present that would reduce the reported performance gains to a fitted parameter or input by construction. The central claim rests on standard supervised training and evaluation splits rather than any self-definitional or load-bearing self-citation step.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard supervised learning assumptions plus the domain premise that short ECG segments contain usable pre-AF signatures; no new entities are postulated.

free parameters (2)
  • neural network weights
    Millions of parameters in the base DL model fitted to ICENTIA11K and then fine-tuned per patient.
  • fine-tuning data volume
    Amount of patient-specific segments used for adaptation is varied and affects reported performance.
axioms (2)
  • domain assumption ECG segments contain statistically detectable precursors to AF onset
    Invoked when training and evaluating the forecasting task on 60-second windows.
  • domain assumption Fine-tuning on recent patient data improves generalization to future segments from the same patient
    Core premise of the personalization comparison.

pith-pipeline@v0.9.1-grok · 5855 in / 1306 out tokens · 19475 ms · 2026-06-27T12:17:59.329227+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Personalized Deep Learning for Short-Term Forecasting of Impending Atrial Fibrillation from Continuous Wearable ECG Signals." pith.science (2026). https://pith.science/paper/2K3WZNPF

@misc{pith2026260610900,
  author       = {Pith},
  title        = {Pith review of: Personalized Deep Learning for Short-Term Forecasting of Impending Atrial Fibrillation from Continuous Wearable ECG Signals},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2K3WZNPF}},
  note         = {Machine review of arXiv:2606.10900}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Background and Objective: Continuous wearable electrocardiogram (ECG) monitoring is increasingly used for ambulatory arrhythmia surveillance, yet forecasting impending atrial fibrillation (AF) is challenged by inter-patient ECG variability. This study investigated whether personalizing a global model via fine-tuning on an individual's ECG signals improves short-term forecasting of impending AF. Methods: A global model trained on the ICENTIA11K dataset was compared against personalized models fine-tuned across three cohorts: ICENTIA11K, IRIDIA-AF, and MobiCARE. Following preprocessing, models processed 60-second ECG segments for a five-minute forecast horizon. We evaluated the impact of adaptation data volume and analyzed ECG features, such as heart rate and RMSSD. Results: Personalized models significantly outperformed the global model, achieving AUROCs of 0.711 vs. 0.614 in ICENTIA11K and 0.686 vs. 0.585 in MobiCARE. Personalization benefits increased with the amount of patient-specific fine-tuning data. While the global model's accuracy rose as AF onset approached, personalized models in the two external cohorts exhibited distinct temporal dynamics, which may indicate the capture of patient-specific characteristics less dependent on proximity to the AF event. Pre-AF episodes showed elevated heart rates and RMSSD. Feature attributions highlighted clinically relevant precursors, including frequent premature atrial complexes (PACs) and short supraventricular tachycardias (SVTs). Conclusions: Adapting deep learning models with patient-specific wearable ECG data significantly enhances short-term forecasting of impending AF. This personalized framework supports timely preventive interventions and improved AF management in ambulatory monitoring environments.

Figures

Figures reproduced from arXiv: 2606.10900 by Eue-Keun Choi, Hee Seok Song, Jangwon Suh, Jungmin Ko, Soonil Kwon, Wonjong Rhee, Yun Kwan Kim.

Figure 1
Figure 1. Figure 1: Workflow of personalized AF forecasting. With the advent of ambulatory ECG [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Data labeling and segmentation process for AF forecasting. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the experimental design. (a) Training the global model. A deep learning model was trained on the ICENTIA11K dataset to capture generalizable patterns preceding imminent AF. (b) Personalization and evaluation of patient-specific models. The global model was fine-tuned using a small amount of labeled data from individual patients (patient-specific training set) and evaluated on separate, later se… view at source ↗
Figure 4
Figure 4. Figure 4: Patient inclusion algorithm. (a) Patients in the ICENTIA11K dataset were di￾vided into training, validation, and personalization sets for internal validation. (b) Patients in the IRIDIA-AF and MobiCARE datasets were selected to form personalization sets for external validation. 13 [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Illustration of how 60-second segments containing AF were additionally included [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of global and personalized models on ICENTIA11K. (a) ROC curves. [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of global and personalized models on IRIDIA-AF. (a) ROC curves. [PITH_FULL_IMAGE:figures/full_fig_p020_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Comparison of global and personalized models on MobiCARE. (a) ROC curves. [PITH_FULL_IMAGE:figures/full_fig_p021_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Effect of the number of training episodes on personalization performance. AU [PITH_FULL_IMAGE:figures/full_fig_p022_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Forecasting performance with respect to temporal distance from AF onset. [PITH_FULL_IMAGE:figures/full_fig_p024_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Comparison of heart rate and RMSSD between non-AF and pre-AF episodes [PITH_FULL_IMAGE:figures/full_fig_p026_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Visualization of feature attribution analysis for representative ECG character [PITH_FULL_IMAGE:figures/full_fig_p028_12.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

49 extracted references · 4 canonical work pages

  1. [1]

    D. Linz, M. Gawalko, K. Betz, J. M. Hendriks, G. Y. Lip, N. Vinter, Y. Guo, S. Johnsen, Atrial fibrillation: epidemiology, screening and dig- ital health, The Lancet Regional Health–Europe 37 (2024)

    2024

  2. [2]

    I. C. Van Gelder, M. Rienstra, K. V. Bunting, R. Casado-Arroyo, V. Caso, H. J. Crijns, T. J. De Potter, J. Dwight, L. Guasti, T. Hanke, et al., 2024 esc guidelines for the management of atrial fibrillation devel- oped in collaboration with the european association for cardio-thoracic surgery (eacts) developed by the task force for the management of atrial...

    2024

  3. [3]

    Kirchhof, A

    P. Kirchhof, A. J. Camm, A. Goette, A. Brandes, L. Eckardt, A. Elvan, T. Fetsch, I. C. van Gelder, D. Haase, L. M. Haegeli, et al., Early rhythm- control therapy in patients with atrial fibrillation, New England Journal of Medicine 383 (2020) 1305–1316

    2020

  4. [4]

    R. B. Schnabel, E. A. Marinelli, E. Arbelo, G. Boriani, S. Boveda, C. M. Buckley, A. J. Camm, B. Casadei, W. Chua, N. Dagres, et al., Early diagnosis and better rhythm management to improve outcomes in pa- tients with atrial fibrillation: the 8th afnet/ehra consensus conference, Europace 25 (2023) 6–27

    2023

  5. [5]

    Svennberg, F

    E. Svennberg, F. Tjong, A. Goette, N. Akoum, L. Di Biase, P. Bor- dachar, G. Boriani, H. Burri, G. Conte, J. C. Deharo, et al., How to use digital devices to detect and manage arrhythmias: an ehra practical guide, Europace 24 (2022) 979–1005

    2022

  6. [6]

    S. R. Steinhubl, J. Waalen, A. M. Edwards, L. M. Ariniello, R. R. Mehta, G. S. Ebner, C. Carter, K. Baca-Motes, E. Felicione, T. Sarich, et al., Effect of a home-based wearable continuous ecg monitoring patch on de- tection of undiagnosed atrial fibrillation: the mstops randomized clinical trial, Jama 320 (2018) 146–155

    2018

  7. [7]

    A. C. Ha, S. Verma, C. D. Mazer, A. Quan, B. Yanagawa, D. A. Latter, T. M. Yau, F. Jacques, C. D. Brown, R. K. Singal, et al., Effect of 44 continuous electrocardiogram monitoring on detection of undiagnosed atrial fibrillation after hospitalization for cardiac surgery: a randomized clinical trial, JAMA Network Open 4 (2021) e2121867–e2121867

    2021

  8. [8]

    Kwon, S.-R

    S. Kwon, S.-R. Lee, E.-K. Choi, H.-J. Ahn, H.-S. Song, Y.-S. Lee, S. Oh, Validation of adhesive single-lead ecg device compared with holter mon- itoring among non-atrial fibrillation patients, Sensors 21 (2021) 3122

    2021

  9. [9]

    Kwon, S.-R

    S. Kwon, S.-R. Lee, E.-K. Choi, H.-J. Ahn, H.-S. Song, Y.-S. Lee, S. Oh, G. Y. Lip, Comparison between the 24-hour holter test and 72-hour single-lead electrocardiogram monitoring with an adhesive patch-type device for atrial fibrillation detection: prospective cohort study, Journal of medical Internet research 24 (2022) e37970

    2022

  10. [10]

    Z. I. Attia, P. A. Noseworthy, F. Lopez-Jimenez, S. J. Asirvatham, A. J. Deshmukh, B. J. Gersh, R. E. Carter, X. Yao, A. A. Rabinstein, B. J. Erickson, et al., An artificial intelligence-enabled ecg algorithm for the identification of patients with atrial fibrillation during sinus rhythm: a retrospective analysis of outcome prediction, The Lancet 394 (201...

    2019

  11. [11]

    K. Lee, M. Lee, H. Choi, Y. Lee, Y. Kim, N. Yoon, H. Park, Mobile single-lead ecg atrial fibrillation prediction enhancement integrated by standard ecg algorithm with deep learning model, European Heart Jour- nal 45 (2024) ehae666–3431

    2024

  12. [12]

    Gavidia, H

    M. Gavidia, H. Zhu, A. N. Montanari, J. Fuentes, C. Cheng, S. Dubner, M. Chames, P. Maison-Blanche, M. M. Rahman, R. Sassi, et al., Early warning of atrial fibrillation using deep learning, Patterns 5 (2024)

    2024

  13. [13]

    J. P. Singh, J. Fontanarava, G. de Mass´ e, T. Carbonati, J. Li, C. Henry, L. Fiorina, Short-term prediction of atrial fibrillation from ambulatory monitoring ecg using a deep neural network, European Heart Journal- Digital Health 3 (2022) 208–217

    2022

  14. [14]

    Gadaleta, P

    M. Gadaleta, P. Harrington, E. Barnhill, E. Hytopoulos, M. P. Turakhia, S. R. Steinhubl, G. Quer, Prediction of atrial fibrillation from at-home single-lead ecg signals without arrhythmias, npj Digital Medicine 6 (2023) 229. 45

    2023

  15. [15]

    K. H. Boon, M. Khalil-Hani, M. Malarvili, C. W. Sia, Paroxysmal atrial fibrillation prediction method with shorter hrv sequences, Computer methods and programs in biomedicine 134 (2016) 187–196

    2016

  16. [16]

    Ebrahimzadeh, M

    E. Ebrahimzadeh, M. Kalantari, M. Joulani, R. S. Shahraki, F. Fayaz, F. Ahmadi, Prediction of paroxysmal atrial fibrillation: A machine learn- ing based approach using combined feature vector and mixture of ex- pert classification on hrv signal, Computer methods and programs in biomedicine 165 (2018) 53–67

    2018

  17. [17]

    Gilon, J.-M

    C. Gilon, J.-M. Gr´ egoire, H. Bersini, Forecast of paroxysmal atrial fib- rillation using a deep neural network, in: 2020 International Joint Con- ference on Neural Networks (IJCNN), IEEE, 2020, pp. 1–7

    2020

  18. [18]

    Tzou, S.-F

    H.-A. Tzou, S.-F. Lin, P.-S. Chen, Paroxysmal atrial fibrillation predic- tion based on morphological variant p-wave analysis with wideband ecg and deep learning, Computer Methods and Programs in Biomedicine 211 (2021) 106396

    2021

  19. [19]

    C. Ding, T. Yao, C. Wu, J. Ni, Deep learning for personalized electrocar- diogram diagnosis: A review, arXiv preprint arXiv:2409.07975 (2024)

    work page arXiv 2024

  20. [20]

    C. Ding, T. Yao, C. Wu, J. Ni, Advances in deep learning for personalized ecg diagnostics: A systematic review addressing inter-patient variability and generalization constraints, Biosensors and Bioelectronics 271 (2025) 117073

    2025

  21. [21]

    Jeong, J

    Y. Jeong, J. Lee, M. Shin, Enhancing inter-patient performance for ar- rhythmia classification with adversarial learning using beat-score maps, Applied Sciences 14 (2024) 7227

    2024

  22. [22]

    Zhang, X

    Y. Zhang, X. Chen, et al., Explainable recommendation: A survey and new perspectives, Foundations and Trends®in Information Retrieval 14 (2020) 1–101

    2020

  23. [23]

    Sarikaya, The technology behind personal digital assistants: An overview of the system architecture and key components, IEEE Sig- nal Processing Magazine 34 (2017) 67–81

    R. Sarikaya, The technology behind personal digital assistants: An overview of the system architecture and key components, IEEE Sig- nal Processing Magazine 34 (2017) 67–81

    2017

  24. [24]

    J. Hong, J. Gao, Q. Liu, Y. Zhang, Y. Zheng, Deep learning model with individualized fine-tuning for dynamic and beat-to-beat blood pressure 46 estimation, in: 2021 IEEE 17th International Conference on Wearable and Implantable Body Sensor Networks (BSN), IEEE, 2021, pp. 1–4

    2021

  25. [25]

    S. Hu, Y. Wang, J. Liu, C. Yang, Personalized transfer learning for single-lead ecg-based sleep apnea detection: exploring the label mapping length and transfer strategy using hybrid transformer model, IEEE Transactions on Instrumentation and Measurement 72 (2023) 1–15

    2023

  26. [26]

    Ng, M.-T

    Y. Ng, M.-T. Liao, T.-L. Chen, C.-K. Lee, C.-Y. Chou, W. Wang, Few- shot transfer learning for personalized atrial fibrillation detection using patient-based siamese network with single-lead ecg records, Artificial Intelligence in Medicine 144 (2023) 102644

    2023

  27. [27]

    Choi, S.-H

    J.-H. Choi, S.-H. Song, H. Kim, J. Kim, H. Park, J. Jeon, J. Hong, H. B. Gwag, S. H. Lee, J. Lee, et al., Machine learning algorithm to predict atrial fibrillation using serial 12-lead ecgs based on left atrial remodeling, Journal of the American Heart Association 13 (2024) e034154

    2024

  28. [28]

    S. Tan, G. Androz, A. Chamseddine, P. Fecteau, A. Courville, Y. Ben- gio, J. P. Cohen, Icentia11k: An unsupervised representation learning dataset for arrhythmia subtype discovery, 2021 Computing in Cardiol- ogy (CinC), 2021

    2021

  29. [29]

    S. Tan, G. Androz, A. Chamseddine, P. Fecteau, A. Courville, Y. Bengio, J. P. Cohen, Icentia11k single lead continuous raw electrocardiogram dataset (version 1.0), 2022.PhysioNet. https://doi.org/10.13026/kk0v- r952

    work page doi:10.13026/kk0v- 2022

  30. [30]

    Gilon, J.-M

    C. Gilon, J.-M. Gr´ egoire, M. Mathieu, S. Carlier, H. Bersini, Iridia-af, a large paroxysmal atrial fibrillation long-term electrocardiogram mon- itoring database, Scientific data 10 (2023) 714

    2023

  31. [31]

    Ahn, E.-K

    H.-J. Ahn, E.-K. Choi, S.-R. Lee, S. Kwon, H.-S. Song, Y.-S. Lee, S. Oh, Three-day monitoring of adhesive single-lead electrocardiogram patch for premature ventricular complex: Prospective study for diagnosis vali- dation and evaluation of burden fluctuation, Journal of Medical Internet Research 26 (2024) e46098

    2024

  32. [32]

    J. Suh, J. Kim, E. Lee, J. Kim, D. Hwang, J. Park, J. Lee, J. Park, S.-Y. Moon, Y. Kim, et al., Learning ecg representations for multi-label 47 classification of cardiac abnormalities, in: 2021 computing in cardiology (CinC), volume 48, IEEE, 2021, pp. 1–4

    2021

  33. [33]

    K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778

    2016

  34. [34]

    W. J. Youden, Index for rating diagnostic tests, Cancer 3 (1950) 32–35

    1950

  35. [35]

    P.-S. Chen, L. S. Chen, M. C. Fishbein, S.-F. Lin, S. Nattel, Role of the autonomic nervous system in atrial fibrillation: pathophysiology and therapy, Circulation research 114 (2014) 1500–1515

    2014

  36. [36]

    Oh, Neuromodulation for atrial fibrillation control, Korean circulation journal 54 (2024) 223–232

    S. Oh, Neuromodulation for atrial fibrillation control, Korean circulation journal 54 (2024) 223–232

    2024

  37. [37]

    NeuroKit2: A Python toolbox for neurophysiological signal processing,

    D. Makowski, T. Pham, Z. J. Lau, J. C. Brammer, F. Lespinasse, H. Pham, C. Sch¨ olzel, S. H. A. Chen, NeuroKit2: A python tool- box for neurophysiological signal processing, Behavior Research Methods 53 (2021) 1689–1696. URL:https://doi.org/10.3758% 2Fs13428-020-01516-y. doi:10.3758/s13428-020-01516-y

    work page doi:10.3758/s13428-020-01516-y 2021

  38. [38]

    Jahmunah, E

    V. Jahmunah, E. Y. Ng, R.-S. Tan, S. L. Oh, U. R. Acharya, Explainable detection of myocardial infarction using deep learning models with grad- cam technique on ecg signals, Computers in Biology and Medicine 146 (2022) 105550

    2022

  39. [39]

    Singh, A

    P. Singh, A. Sharma, Interpretation and classification of arrhythmia us- ing deep convolutional network, IEEE Transactions on Instrumentation and Measurement 71 (2022) 1–12

    2022

  40. [40]

    S. Kwon, J. Suh, E.-K. Choi, J. Kim, H. Ju, H.-J. Ahn, S. Kim, S.-R. Lee, S. Oh, W. Rhee, Classification of underlying paroxysmal supraven- tricular tachycardia types using deep learning of sinus rhythm electro- cardiograms, Digital Health 10 (2024) 20552076241281200

    2024

  41. [41]

    R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, D. Batra, Grad-cam: Visual explanations from deep networks via gradient-based localization, in: Proceedings of the IEEE international conference on computer vision, 2017, pp. 618–626. 48

    2017

  42. [42]

    J. Suh, J. Kim, S. Kwon, E. Jung, H.-J. Ahn, K.-Y. Lee, E.-K. Choi, W. Rhee, Visual interpretation of deep learning model in ecg classi- fication: A comprehensive evaluation of feature attribution methods, Computers in Biology and Medicine 182 (2024) 109088

    2024

  43. [43]

    Hinrichs, T

    N. Hinrichs, T. Roeschl, P. Lanmueller, F. Balzer, C. Eickhoff, B. O’Brien, V. Falk, A. Meyer, Short-term vital parameter forecast- ing in the intensive care unit: A benchmark study leveraging data from patients after cardiothoracic surgery, PLOS Digital Health 3 (2024) e0000598

    2024

  44. [44]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, I. Polosukhin, Attention is all you need, Advances in neural information processing systems 30 (2017)

    2017

  45. [45]

    E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, W. Chen, et al., Lora: Low-rank adaptation of large language models., ICLR 1 (2022) 3

    2022

  46. [46]

    Kumar, S

    D. Kumar, S. Puthusserypady, H. Dominguez, K. Sharma, J. E. Bardram, Cachet-cadb: A contextualized ambulatory electrocardiogra- phy arrhythmia dataset, Frontiers in Cardiovascular Medicine 9 (2022) 893090

    2022

  47. [47]

    Machorro-Cano, J

    I. Machorro-Cano, J. O. Olmedo-Aguirre, G. Alor-Hern´ andez, L. Rodr´ ıguez-Mazahua, L. N. S´ anchez-Morales, N. P´ erez-Castro, Cloud- based platforms for health monitoring: a review, in: Informatics, vol- ume 11, MDPI, 2023, p. 2

    2023

  48. [48]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, et al., Pytorch: An imper- ative style, high-performance deep learning library, Advances in neural information processing systems 32 (2019)

    2019

  49. [49]

    Kokhlikyan, V

    N. Kokhlikyan, V. Miglani, M. Martin, E. Wang, B. Alsallakh, J. Reynolds, A. Melnikov, N. Kliushkina, C. Araya, S. Yan, et al., Cap- tum: A unified and generic model interpretability library for pytorch, arXiv preprint arXiv:2009.07896 (2020). 49

    work page arXiv 2009