Pith. sign in

REVIEW 4 major objections 5 minor 57 references

Early Diagnosis of Atrial Fibrillation Recurrence: A Large Tabular Model Approach with Structured and Unstructured Clinical Data

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read A large tabular model trained on merged structured and free-text EHR data predicts atrial fibrillation recurrence one to two years after onset better than clinical scores (CHADS2-VASc, HATCH, APPLE) and a feature-engineered SVM, reaching…

desk verdict Solid applied study with a reusable NLP-plus-EHR pipeline and an honest small effect; the 'surpassing' claim overreaches the evidence, but the paper deserves peer review with required revisions. read the letter →

arxiv 2505.14643 v1 pith:BHOIIA7A submitted 2025-05-20 cs.LG

classification cs.LG
keywords AtrialfibrillationrecurrenceElectronichealthrecordsLargetabularmodelsNaturallanguageprocessingClinicalscoresSilverlabelsTabPFNEarlydiagnosis
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to show that a large tabular model, TabPFN, can predict whether atrial fibrillation will recur between one month and two years after a first documented episode, and that it beats the clinical scores currently used for this task. To do this, the authors build a tabular dataset for 1,508 patients by merging codified electronic health record variables with features automatically extracted from Spanish discharge reports using a regex and language-model pipeline. The training labels are generated automatically ("silver") and checked against manually annotated test labels, with 83% agreement. On the manually annotated test set of 278 patients, TabPFN reaches a Matthews correlation coefficient of 0.1886, versus 0.1626 for SVM and 0.0832 for the best clinical score, HATCH. If true, this would give clinicians a data-driven, minimally annotated tool for earlier rhythm-control decisions.

What carries the argument

The load-bearing object is TabPFN, a large tabular model pretrained on millions of synthetic tables that is designed to perform well on small datasets with missing values and many features, requiring no hyperparameter tuning or feature preprocessing. The second component is the three-step text-to-tabular pipeline: Report2Vector converts each Spanish discharge report into an 86-variable vector using section identification, medical entity recognition with negation detection, and cardiologist-designed regular expressions; Structured2Vector extracts codified variables from the hospital's business intelligence system; VectorMerger merges them into one row per patient with time-window constraints, and also assigns the silver recurrence labels by looking for AF evidence in follow-up reports one month to two years after onset. The claim depends on this pipeline producing both the predictive features and the training labels.

What would settle it

Have clinicians re-annotate the 978-patient training set manually, retrain TabPFN on those gold labels, and evaluate on the same 278-patient test set; if the advantage over HATCH largely disappears, or if the original model performs at chance on test cases where the silver and gold labels disagree, the reported MCC was an artifact of learning the automatic labelling pipeline.

Watch

Extended reading notes

Core claim

The paper's central claim is that a tabular foundation model, TabPFN, predicts atrial fibrillation recurrence between one month and two years after a first documented episode with higher accuracy than the commonly used clinical scores CHADS2-VASc, HATCH, and APPLE, and also than a support vector machine that requires extensive preprocessing. The model is trained on a dataset of 978 automatically labeled patients and evaluated on 278 manually annotated patients; the best variant, TabPFN-pre, achieves a Matthews correlation coefficient of 0.1886 on the held-out set, compared with 0.1626 for SVM and 0.0832 for HATCH, with the difference from SVM reported as statistically significant (paired t-test p=0.0118). The study also claims that the underlying data-generation methodology—merging structured codified EHR data with features extracted from free-text discharge reports through a regex and language-model pipeline—reduces annotation effort and mitigates coding errors, and that this methodology is generalizable to other diseases. In subgroup analyses, the paper finds that models and scores perform worse on male patients and that for patients under 75 TabPFN reaches MCC around 0.27, roughly double the HATCH score.

Load-bearing premise

The automatic silver labels for the training set are generated by the same regular-expression and language-model pipeline that extracts the predictive features, and they agree with expert manual labels in only 83% of cases, so systematic labelling errors could make the model learn the pipeline's output rather than true recurrence.

Editorial extensions

If this is right

  • If the result holds, cardiologists could obtain a recurrence risk estimate for the one-month-to-two-year window from a model that needs no manual feature engineering, only tabular data plus discharge reports.
  • The comparison quantifies how much predictive information the three clinical scores leave unused: HATCH, the best score, reaches only MCC 0.0832, close to chance-level discrimination for this outcome.
  • The silver-labelling approach would allow other health systems to build predictive models for outcomes that are not explicitly coded, provided they have discharge reports and a small set of manual labels for validation.
  • For patients under 75, where rhythm-control decisions matter most, TabPFN's MCC roughly doubles that of HATCH, suggesting the model could support treatment personalization precisely in the group where clinical guidance is least settled.
  • The gender-stratified results imply that a model acceptable on the full cohort may underperform on male patients; age differences between male and female cohorts account for part of the gap.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A direct corollary the authors do not stress: since features and labels are produced by the same regexes, the model's superiority over HATCH may in part reflect learning the annotator's decision rules; an independent gold-label retraining is needed to know how much real clinical signal remains.
  • The 83% silver-gold agreement effectively caps the achievable gold MCC under label noise; if the noise is non-random, even a perfect model could not exceed the observed 0.19 on this test set.
  • The claimed generalizability to other diseases is stronger for the vector-merging procedure than for the label generator, since the regular expressions and AF-specific flags would have to be redesigned per pathology.
  • A testable extension: run the same TabPFN-versus-scores comparison using only structured data, isolating how much of the performance gain comes from the NLP-extracted features rather than from the model architecture.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript proposes a method for predicting atrial fibrillation (AF) recurrence between one month and two years after the first documented AF onset, using a tabular dataset generated by merging structured EHR data from the Osakidetza Business Intelligence system with information extracted from Spanish discharge reports. A silver training set of 978 patients is labeled automatically by a three-step NLP/regex pipeline, while a gold test set of 278 patients is manually annotated by the cardiology department. The study compares clinical scores (CHADS2-VASc, HATCH, APPLE), classical ML models (SVM, LR, XGB), and the TabPFN large tabular model in two variants (preprocessed and raw). The central claim is that TabPFN-pre achieves the highest predictive performance, with MCC 0.1886 versus 0.1626 for SVM and 0.0832 for HATCH; additional experiments examine gender and age subgroups. The paper positions the dataset-generation methodology and the two-year prediction horizon as the main contributions, and it makes code publicly available.

Significance. If the results are validated, the paper would make a useful contribution to clinically oriented ML for AF recurrence: it addresses a less-studied time window, combines structured and free-text EHR sources, and uses a manually annotated gold test set while reducing annotation effort through silver labels. The public code, the explicit comparison with established clinical scores, and the gender/age subgroup analyses are all strengths. However, the central superiority claim rests on a small MCC gap with an unspecified significance test, and the silver-labeling procedure creates a potential feedback loop between feature extraction and label assignment. These issues affect the main scientific conclusion and need to be addressed before the claims can be accepted.

major comments (4)
  1. [Section 5.2 and Section 6] The central claim in the abstract that the LTM 'surpassing both traditional clinical scores and ML models' is not robustly supported by Table 3. The difference between TabPFN-pre (MCC 0.1886) and SVM (MCC 0.1626) is only 0.026 on a 278-patient test set, and the reported paired t-test (footnote 9, p=0.0118) does not specify what quantity was tested; a paired t-test on per-patient predicted probabilities does not establish a statistically significant difference in MCC, and no multiple-comparison correction is reported. In addition, Section 6 concludes that the LTM performed 'a la par' with the SVM-based model, directly contradicting the abstract's stronger claim. Please provide bootstrap or permutation confidence intervals for MCC itself, and reconcile the wording between the abstract and the conclusion.
  2. [Section 3.2.2 and Appendix C] The silver training labels are produced by the same Report2Vector and VectorMerger pipeline that generates the predictive features; the recurrence label is a downstream function of the same discharge-report sections and AF flags that enter the feature vector. Because the silver labels agree with expert manual labels in only 83% of cases (Section 3.2.2), systematic pipeline errors could be learned by the model, making the reported MCC partly a measure of agreement with the pipeline rather than with true recurrence. The paper should quantify this risk, for example by retraining on the gold-labeled patients only, by examining feature distributions for concordant versus discordant cases, or by a label-noise robustness analysis.
  3. [Section 3.2.1 and Section 5.2] The paper's methodological claim that integrating unstructured free text with structured codified data enhances dataset quality and predictive performance is not tested by an ablation. All models in Table 3 are evaluated on the enriched dataset, so the reader cannot determine whether the NLP-extracted variables add any predictive value beyond the structured OBI variables alone. Please report results for models trained on structured-only data and on the enriched data, using the same gold test set and the same model architectures.
  4. [Section 5.3 and Section 5.4] The gender and age subgroup conclusions are based on the same unspecified paired t-test (footnotes 10, 11, and 13) and on small subgroups. For example, the TabPFN-pre MCC difference between the female and male subsets in Table 5 is 0.1747 versus 0.0830, but no confidence intervals or interaction test are provided, and the reported p-values are not adjusted for multiple comparisons. The disparity claims should be supported by an appropriate subgroup analysis, ideally with a formal interaction term or bootstrap intervals.
minor comments (5)
  1. [Section 3.3 and Table 1] Table 1 reports a training set of 1,200 patients, while Section 3.3 states that the final training set after exclusions consisted of 978 patients; the table should be updated or the relationship between the two numbers clarified.
  2. [Section 4.4] The note that 'ROC-AUC is generated including predictions that obtained insufficient sensitivity and specificity' is unclear and should be rephrased to describe the metric's threshold-free property more precisely.
  3. [Table 4] The value '68,49' uses a comma as a decimal separator while other values use periods; please standardize the number formatting throughout.
  4. [Section 6] The phrase 'a la par' should be replaced with a precise comparative statement, since it directly conflicts with the abstract's claim of surpassing ML models.
  5. [Appendix A] The 'Type' column for 'Atrial Fibrillation' lists 'Categorical (c = 4)' but the percentages show only three non-zero values; please check the consistency of the reported categories.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central TabPFN-vs-baselines comparison is evaluated on a manually annotated gold test set, and the self-cited NLP components are auxiliary, not load-bearing.

full rationale

The manuscript's central claim is an empirical performance comparison on a held-out test set manually annotated by the BioBizkaia cardiology department (Section 3.2.2 and Table 3), so the superiority claim is not defined in terms of the fitted model's own outputs. The automatic silver labels for the 978-patient training set are produced by the same Report2Vector / regular-expression pipeline used to build the tabular features (Sections 3.2.1-3.2.2 and Appendix C); this is a real data-quality and possible label-noise concern, but the paper reports only 83% agreement with expert annotation and does not equate the silver label with the gold label, nor does any equation reduce a predicted metric to a fitted value. Self-citations (EriBERTa in de la Iglesia et al. 2023 and the section-identification tool in Goenaga et al. 2021) support auxiliary text-processing components with their own validation and are not used to forbid alternatives. The statistical-significance footnote (paired t-test, p=0.0118) is under-specified and the abstract's 'surpassing ML models' is in tension with the conclusion's 'a la par' phrasing, but these are correctness/reporting issues, not circularity. Therefore the derivation chain is self-contained for circularity purposes.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper relies on three resources: manual cardiology annotation as ground truth, a regex and LM pipeline's silver labels as training signal, and TabPFN's pretrained prior. None are derived here, and the silver-label agreement of 83% is the largest unmeasured source of uncertainty. Clinical score thresholds from the literature are used as fixed cut-offs rather than fitted, but they influence the comparison.

free parameters (3)
  • Clinical score thresholds = CHA2DS2-VASc >=2, HATCH >=2, APPLE >=2
    Chosen by the collaborating cardiologist based on literature; they convert continuous scores into binary recurrence labels for comparison, and the thresholds affect the reported MCC.
  • LSFM feature retention fraction = top 25% features
    Hand-chosen in preprocessing for XGBoost; affects XGB's reported MCC of 0.14.
  • Undersampling target balance = 50-50 recurrence to non-recurrence
    Chosen to address class imbalance; affects the classification threshold and reported metrics for the classic ML models.
assumptions (4)
  • domain assumption Manual annotation by the cardiology department is the ground truth for recurrence.
    Test label quality is assumed perfect and is used to evaluate all models and to estimate the 83% silver-label accuracy in Section 3.2.2.
  • domain assumption The automatic NLP and regex pipeline's silver labels are sufficiently accurate for training.
    The 978-patient training set is labelled by the same regex and LM tools that extract features; the paper relies on 83% agreement with manual labels but does not model the effect of label noise on training.
  • domain assumption The recurrence definition (a clinical AF event between one month and two years after onset) is the clinically relevant target.
    The entire label and feature design follows this definition set by the cardiology unit in Section 3.2.2; changing the window would change all reported results.
  • domain assumption TabPFN's pretraining on synthetic tabular datasets transfers to this clinical table.
    The paper argues TabPFN is suited to small, high-dimensional data without fine-tuning (Section 4.3); if the prior-data distribution mismatches EHR data, the reported performance may not transfer.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Early Diagnosis of Atrial Fibrillation Recurrence: A Large Tabular Model Approach with Structured and Unstructured Clinical Data." pith.science (2026). https://pith.science/paper/BHOIIA7A

@misc{pith2026250514643,
  author       = {Pith},
  title        = {Pith review of: Early Diagnosis of Atrial Fibrillation Recurrence: A Large Tabular Model Approach with Structured and Unstructured Clinical Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BHOIIA7A}},
  note         = {Machine review of arXiv:2505.14643}
}
read the original abstract

BACKGROUND: Atrial fibrillation (AF), the most common arrhythmia, is linked to high morbidity and mortality. In a fast-evolving AF rhythm control treatment era, predicting AF recurrence after its onset may be crucial to achieve the optimal therapeutic approach, yet traditional scores like CHADS2-VASc, HATCH, and APPLE show limited predictive accuracy. Moreover, early diagnosis studies often rely on codified electronic health record (EHR) data, which may contain errors and missing information. OBJECTIVE: This study aims to predict AF recurrence between one month and two years after onset by evaluating traditional clinical scores, ML models, and our LTM approach. Moreover, another objective is to develop a methodology for integrating structured and unstructured data to enhance tabular dataset quality. METHODS: A tabular dataset was generated by combining structured clinical data with free-text discharge reports processed through natural language processing techniques, reducing errors and annotation effort. A total of 1,508 patients with documented AF onset were identified, and models were evaluated on a manually annotated test set. The proposed approach includes a LTM compared against traditional clinical scores and ML models. RESULTS: The proposed LTM approach achieved the highest predictive performance, surpassing both traditional clinical scores and ML models. Additionally, the gender and age bias analyses revealed demographic disparities. CONCLUSION: The integration of structured data and free-text sources resulted in a high-quality dataset. The findings emphasize the limitations of traditional clinical scores in predicting AF recurrence and highlight the potential of ML-based approaches, particularly our LTM model.

Figures

Figures reproduced from arXiv: 2505.14643 by the authors.

Figure 1
Figure 1. Overview of the generated dataset and the experimental setup for AF recurrence prediction. A.G. Domingo-Aldama et al.: Preprint submitted to Elsevier Page 3 of 20 [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Summary of the AF recurrence cohort generation process. First, patients with an AF onset are identified, followed by the assignment of the AF recurrence label. First, we introduce details on the detection of AF onset episodes, including the patient inclusion criteria (subsec￾tion 3.1). Second, we describe the generation of a silver dataset (subsection 3.2), starting with the feature vector generation process for eac… view at source ↗
Figure 3
Figure 3. Summary of the vector generation process. Information from both free-text discharge reports and structured codified data from OBI system are merged in a single vector per patient describing its health status around the time of AF onset. Additional details on each tool can be found in Ap￾pendix C. We must stress that the vector generation method￾ology proposed in this study can be adapted to other health￾care systems… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Age distribution comparison by gender, with females in orange and males in blue. 5.4. Results for age experiments Given the observed differences in age distribution, an additional experiment was conducted on patients under the age of 75 to assess the performance of bot…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 53 canonical work pages

  1. [1]

    author Arik, S. \"O . , author Pfister, T. , year 2021 . title Tabnet: Attentive interpretable tabular learning , in: booktitle Proceedings of the AAAI conference on artificial intelligence , pp. pages 6679--6687

  2. [2]

    , author Barr \'e , V

    author Benali, K. , author Barr \'e , V. , author Hermida, A. , author Galand, V. , author Milhem, A. , author Philibert, S. , author Boveda, S. , author Bars, C. , author Anselme, F. , author Maille, B. , et al., year 2023 . title Recurrences of atrial fibrillation despite durable pulmonary vein isolation: the party-pvi study . journal Circulation: Arrhy...

  3. [3]

    , author Patel, P

    author Bhatt, C.M. , author Patel, P. , author Ghetia, T. , author Mazzeo, P.L. , year 2023 . title Effective heart disease prediction using machine learning techniques . journal Algorithms volume 16 , pages 88

  4. [4]

    , author Hartvigsen, G

    author Botsis, T. , author Hartvigsen, G. , author Chen, F. , author Weng, C. , year 2010 . title Secondary use of ehr: data quality issues and informatics opportunities . journal Summit on translational bioinformatics volume 2010 , pages 1

  5. [5]

    , author Zou, F

    author Brahier, M.S. , author Zou, F. , author Abdulkareem, M. , author Kochi, S. , author Migliarese, F. , author Thomaides, A. , author Ma, X. , author Wu, C. , author Sandfort, V. , author Bergquist, P.J. , et al., year 2023 . title Using machine learning to enhance prediction of atrial fibrillation recurrence after catheter ablation . journal Journal ...

  6. [6]

    , author van der Schaar, M

    author van Breugel, B. , author van der Schaar, M. , year 2024 . title Why tabular foundation models should be a research priority . journal Proceedings of the 41st International Conference on Machine Learning volume 235 , pages 48976--48993 . https://proceedings.mlr.press/v235/van-breugel24a.html

  7. [7]

    , author Kaczmarek-Majer, K

    author Budzianowski, J. , author Kaczmarek-Majer, K. , author Rze \'z niczak, J. , author S omczy \'n ski, M. , author Wichrowski, F. , author Hiczkiewicz, D. , author Musielak, B. , author Grydz, . , author Hiczkiewicz, J. , author Burchardt, P. , year 2023 . title Machine learning model for predicting late recurrence of atrial fibrillation after cathete...

  8. [8]

    , author Ezeiza, N

    author Casillas, A. , author Ezeiza, N. , author Goenaga, I. , author Perez, A. , author Soto, X. , year 2019 . title Measuring the effect of different types of unsupervised word representations on medical named entity recognition . journal International journal of medical informatics volume 129 , pages 100--106

Show all 57 references
  1. [9]

    , author Chow, C.K

    author Celermajer, D.S. , author Chow, C.K. , author Marijon, E. , author Anstey, N.M. , author Woo, K.S. , year 2012 . title Cardiovascular disease in the developing world: prevalences, patterns, and the potential of early disease detection . journal Journal of the American C...

  2. [10]

    , author Bowyer, K.W

    author Chawla, N.V. , author Bowyer, K.W. , author Hall, L.O. , author Kegelmeyer, W.P. , year 2002 . title Smote: synthetic minority over-sampling technique . journal Journal of artificial intelligence research volume 16 , pages 321--357

  3. [11]

    , author Jurman, G

    author Chicco, D. , author Jurman, G. , year 2023 . title The matthews correlation coefficient (mcc) should replace the roc auc as the standard metric for assessing binary classification . journal BioData Mining volume 16 , pages 4

  4. [12]

    , author Chieng, D

    author Crowley, R. , author Chieng, D. , author Sugumar, H. , author Ling, L.H. , author Segan, L. , author William, J. , author Prabhu, S. , author Voskoboinik, A. , author Wong, G. , author Morton, J.B. , et al., year 2024 . title Catheter ablation for persistent atrial fibr...

  5. [13]

    , author Rodrigues, P.P

    author Cruz-Correia, R.J. , author Rodrigues, P.P. , author Freitas, A. , author Almeida, F.C. , author Chen, R. , author Costa-Pereira, A. , year 2009 . title Data quality and integration issues in electronic health records , in: booktitle Information discovery on electronic ...

  6. [14]

    , author Pisters, R

    author De Vos, C.B. , author Pisters, R. , author Nieuwlaat, R. , author Prins, M.H. , author Tieleman, R.G. , author Coelen, R.J.S. , author van den Heijkant, A.C. , author Allessie, M.A. , author Crijns, H.J. , year 2010 . title Progression from paroxysmal to persistent atri...

  7. [15]

    , author Aljurf, M

    author Elhassan, T. , author Aljurf, M. , et al., year 2016 . title Classification of imbalance data using tomek link (t-link) combined with random under-sampling (rus) as a data reduction method . journal Global J Technol Optim S volume 1 , pages 2016

  8. [16]

    , author Mueller, J

    author Erickson, N. , author Mueller, J. , author Shirkov, A. , author Zhang, H. , author Larroy, P. , author Li, M. , author Smola, A. , year 2020 . title Autogluon-tabular: Robust and accurate automl for structured data . journal arXiv preprint arXiv:2003.06505

  9. [17]

    , author Li, Y

    author Fan, X. , author Li, Y. , author He, Q. , author Wang, M. , author Lan, X. , author Zhang, K. , author Ma, C. , author Zhang, H. , year 2023 . title Predictive value of machine learning for recurrence of atrial fibrillation after catheter ablation: A systematic review a...

  10. [18]

    , year 2018

    author Feder, S.L. , year 2018 . title Data quality in electronic health records research: quality domains and assessment methods . journal Western journal of nursing research volume 40 , pages 753--766

  11. [19]

    , author Ormaetxe Merodio, J

    author Garcia Olea, A. , author Ormaetxe Merodio, J. , author Atutxa Salazar, A. , author Diez Gonzalez, I. , author Fernandez De La Prieta, I. , author Maeztu Rada, M. , author Amuriza De Luis, E. , author Ugedo Alzaga, K. , author Idiazabal Rodriguez, U. , author Pereiro Lil...

  12. [20]

    , author Lahuerta, X

    author Goenaga, I. , author Lahuerta, X. , author Atutxa, A. , author Gojenola, K. , year 2021 . title A section identification tool: towards hl7 cda/ccr standardization in spanish discharge summaries . journal Journal of Biomedical Informatics volume 121 , pages 103875

  13. [21]

    , author Wang, L

    author Goh, K.H. , author Wang, L. , author Yeow, A.Y.K. , author Poh, H. , author Li, K. , author Yeow, J.J.L. , author Tan, G.Y.H. , year 2021 . title Artificial intelligence in sepsis early prediction and diagnosis using unstructured data in healthcare . journal Nature comm...

  14. [22]

    , author Rubachev, I

    author Gorishniy, Y. , author Rubachev, I. , author Khrulkov, V. , author Babenko, A. , year 2021 . title Revisiting deep learning models for tabular data . journal Advances in Neural Information Processing Systems volume 34 , pages 18932--18943

  15. [23]

    u ller, S. , author Purucker, L. , author Krishnakumar, A. , author K \

    author Hollmann, N. , author M \"u ller, S. , author Purucker, L. , author Krishnakumar, A. , author K \"o rfer, M. , author Hoo, S.B. , author Schirrmeister, R.T. , author Hutter, F. , year 2025 . title Accurate predictions on small data with a tabular foundation model . jour...

  16. [24]

    , author Sun, H

    author Huang, W. , author Sun, H. , author Luo, Y. , author Xiong, S. , author Tang, Y. , author Long, Y. , author Zhang, Z. , author Liu, H. , year 2024 . title Better performance of the apple score for the prediction of very early atrial fibrillation recurrence post-ablation...

  17. [25]

    , author Khurshid, S

    author Hulme, O.L. , author Khurshid, S. , author Weng, L.C. , author Anderson, C.D. , author Wang, E.Y. , author Ashburner, J.M. , author Ko, D. , author McManus, D.D. , author Benjamin, E.J. , author Ellinor, P.T. , et al., year 2019 . title Development and validation of a p...

  18. [26]

    , author Atutxa, A

    author de la Iglesia, I. , author Atutxa, A. , author Gojenola, K. , author Barrena, A. , year 2023 . title Eriberta: A bilingual pre-trained language model for clinical natural language processing . arXiv:2306.07373 http://arxiv.org/abs/2306.07373

  19. [27]

    , author Zhang, H

    author Jetley, G. , author Zhang, H. , year 2019 . title Electronic health records in is research: Quality issues, essential thresholds and remedial actions . journal Decision Support Systems volume 126 , pages 113137

  20. [28]

    , author Crowe, B

    author Kanjee, Z. , author Crowe, B. , author Rodman, A. , year 2023 . title Accuracy of a generative artificial intelligence model in a complex diagnostic challenge . journal JAMA volume 330 , pages 78--80 . https://doi.org/10.1001/jama.2023.8288, :10.1001/jama.2023.8288

  21. [29]

    , author Cyriac, J

    author Knecht, S. , author Cyriac, J. , author Badertscher, P. , author Krisai, P. , author Schlageter, V. , author Osswald, S. , author Zellweger, M. , author Kuhne, M. , author Sticherling, C. , year 2024 . title Machine learning for outcome prediction of atrial fibrillation...

  22. [30]

    , author Hindricks, G

    author Kornej, J. , author Hindricks, G. , author Shoemaker, M.B. , author Husser, D. , author Arya, A. , author Sommer, P. , author Rolf, S. , author Saavedra, P. , author Kanagasundram, A. , author Patrick Whalen, S. , et al., year 2015 . title The apple score: a novel and s...

  23. [31]

    , author Koul, A

    author Kumar, Y. , author Koul, A. , author Singla, R. , author Ijaz, M.F. , year 2023 . title Artificial intelligence in disease diagnosis: a systematic literature review, synthesizing framework and future research agenda . journal Journal of Ambient Intelligence and Humanize...

  24. [32]

    , author Weiskopf, N

    author Lewis, A.E. , author Weiskopf, N. , author Abrams, Z.B. , author Foraker, R. , author Lai, A.M. , author Payne, P.R. , author Gupta, A. , year 2023 . title Electronic health record data quality assessment and tools: a systematic review . journal Journal of the American ...

  25. [33]

    , author Nieuwlaat, R

    author Lip, G.Y. , author Nieuwlaat, R. , author Pisters, R. , author Lane, D.A. , author Crijns, H.J. , year 2010 . title Refining clinical risk stratification for predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based approach: the euro ...

  26. [34]

    , author Sanchis-Gomar, F

    author Lippi, G. , author Sanchis-Gomar, F. , author Cervellin, G. , year 2021 . title Global epidemiology of atrial fibrillation: an increasing epidemic and public health challenge . journal International journal of stroke volume 16 , pages 217--221

  27. [35]

    , author Chen, W.S

    author Liu, C.M. , author Chen, W.S. , author Chang, S.L. , author Hsieh, Y.C. , author Hsu, Y.H. , author Chang, H.X. , author Lin, Y.J. , author Lo, L.W. , author Hu, Y.F. , author Chung, F.P. , et al., year 2024 . title Use of artificial intelligence and i-score for predict...

  28. [36]

    , author Vadakken, M.E

    author McIntyre, W.F. , author Vadakken, M.E. , author Connolly, S.J. , author Mendoza, P.A. , author Lengyel, A.P. , author Rai, A.S. , author Latendresse, N.R. , author Grinvalds, A.J. , author Ramasundarahettige, C. , author Acosta, J.G. , et al., year 2023 . title Atrial f...

  29. [37]

    , author Wu, J

    author Nadarajah, R. , author Wu, J. , author Frangi, A.F. , author Hogg, D. , author Cowan, C. , author Gale, C. , year 2021 . title Predicting patient-level new-onset atrial fibrillation from population-based nationwide electronic health records: protocol of find-af for deve...

  30. [38]

    , author Majeed, A.A

    author Naser, M.A. , author Majeed, A.A. , author Alsabah, M. , author Al-Shaikhli, T.R. , author Kaky, K.M. , year 2024 . title A review of machine learning’s role in cardiovascular disease prediction: Recent advances and future challenges . journal Algorithms volume 17 , pages 78

  31. [39]

    , author Kannabhiran, M

    author Njoku, A. , author Kannabhiran, M. , author Arora, R. , author Reddy, P. , author Gopinathannair, R. , author Lakkireddy, D. , author Dominic, P. , year 2018 . title Left atrial volume predicts atrial fibrillation recurrence after radiofrequency ablation: a meta-analysi...

  32. [40]

    , author Mark, D.B

    author Packer, D.L. , author Mark, D.B. , author Robb, R.A. , author Monahan, K.H. , author Bahnson, T.D. , author Poole, J.E. , author Noseworthy, P.A. , author Rosenberg, Y.D. , author Jeffries, N. , author Mitchell, L.B. , et al., year 2019 . title Effect of catheter ablati...

  33. [41]

    , author Morozov, S

    author Popov, S. , author Morozov, S. , author Babenko, A. , year 2019 . title Neural oblivious decision ensembles for deep learning on tabular data . journal arXiv preprint arXiv:1909.06312

  34. [42]

    , author Guo, H

    author Qiu, Y. , author Guo, H. , author Wang, S. , author Yang, S. , author Peng, X. , author Xiayao, D. , author Chen, R. , author Yang, J. , author Liu, J. , author Li, M. , et al., year 2024 . title Deep learning-based multimodal fusion of the surface ecg and clinical feat...

  35. [43]

    , year 2023

    author Ramos, O. , year 2023 . title The use of hatch score for the prediction of post operative atrial fibrillation (poaf) after coronary artery bypass graft: a meta-analysis . journal European Heart Journal volume 44 , pages ehac779--019

  36. [44]

    , author Sim, I

    author Roney, C.H. , author Sim, I. , author Yu, J. , author Beach, M. , author Mehta, A. , author Alonso Solis-Lemus, J. , author Kotadia, I. , author Whitaker, J. , author Corrado, C. , author Razeghi, O. , et al., year 2022 . title Predicting atrial fibrillation recurrence ...

  37. [45]

    , author Park, M.Y

    author Shin, S.H. , author Park, M.Y. , author Oh, W.J. , author Hong, S.J. , author Pak, H.N. , author Song, W.H. , author Lim, D.S. , author Kim, Y.H. , author Shim, W.J. , year 2008 . title Left atrial volume is a predictor of atrial fibrillation recurrence after catheter a...

  38. [46]

    , author Yao, X

    author Siontis, K.C. , author Yao, X. , author Pirruccello, J.P. , author Philippakis, A.A. , author Noseworthy, P.A. , year 2020 . title How will machine learning inform the clinical care of atrial fibrillation? journal Circulation research volume 127 , pages 155--169

  39. [47]

    , author Ansari, U

    author Soni, J. , author Ansari, U. , author Sharma, D. , author Soni, S. , et al., year 2011 . title Predictive data mining for medical diagnosis: An overview of heart disease prediction . journal International Journal of Computer Applications volume 17 , pages 43--48

  40. [48]

    , author Redberg, R

    author Stock, E.O. , author Redberg, R. , year 2012 . title Cardiovascular disease in women . journal Current problems in cardiology volume 37 , pages 450--526

  41. [49]

    , author Sung, K.L

    author Sung, S.F. , author Sung, K.L. , author Pan, R.C. , author Lee, P.J. , author Hu, Y.H. , year 2022 . title Automated risk assessment of newly detected atrial fibrillation poststroke from electronic health record data using machine learning and natural language processin...

  42. [50]

    , author Stewart, M

    author Terry, A.L. , author Stewart, M. , author Cejic, S. , author Marshall, J.N. , author de Lusignan, S. , author Chesworth, B.M. , author Chevendra, V. , author Maddocks, H. , author Shadd, J. , author Burge, F. , et al., year 2019 . title A basic model for assessing prima...

  43. [51]

    , author Colborn, K.L

    author Tiwari, P. , author Colborn, K.L. , author Smith, D.E. , author Xing, F. , author Ghosh, D. , author Rosenberg, M.A. , year 2020 . title Assessment of a machine learning model applied to harmonized electronic health record data for the prediction of incident atrial fibr...

  44. [52]

    , author Noseworthy, P.A

    author Tseng, A.S. , author Noseworthy, P.A. , year 2021 . title Prediction of atrial fibrillation using machine learning: a review . journal Frontiers in Physiology volume 12 , pages 752317

  45. [53]

    , author Serenelli, M

    author Vitali, F. , author Serenelli, M. , author Airaksinen, J. , author Pavasini, R. , author Tomaszuk-Kazberuk, A. , author Mlodawska, E. , author Jaakkola, S. , author Balla, C. , author Falsetti, L. , author Tarquinio, N. , et al., year 2019 . title Cha2ds2-vasc score pre...

  46. [54]

    , author Proietti, M

    author Vitolo, M. , author Proietti, M. , author Imberti, J.F. , author Bonini, N. , author Romiti, G.F. , author Mei, D.A. , author Malavasi, V.L. , author Diemberger, I. , author Fauchier, L. , author Marin, F. , et al., year 2023 . title Factors associated with progression ...

  47. [55]

    , author Chaudhary, M

    author Yelne, S. , author Chaudhary, M. , author Dod, K. , author Sayyad, A. , author Sharma, R. , year 2023 . title Harnessing the power of ai: a comprehensive review of its impact and challenges in nursing science and healthcare . journal Cureus volume 15

  48. [56]

    , author Xu, Z

    author Zhou, S. , author Xu, Z. , author Zhang, M. , author Xu, C. , author Guo, Y. , author Zhan, Z. , author Ding, S. , author Wang, J. , author Xu, K. , author Fang, Y. , et al., year 2024 . title Large language models for disease diagnosis: A scoping review . journal arXiv...

  49. [57]

    , author Nakamura, K

    author Zhou, X. , author Nakamura, K. , author Sahara, N. , author Takagi, T. , author Toyoda, Y. , author Enomoto, Y. , author Hara, H. , author Noro, M. , author Sugi, K. , author Moroi, M. , et al., year 2022 . title Deep learning-based recurrence prediction of atrial fibri...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.