MétaCan
Menu
Back to cohort
Record W4407421485 · doi:10.2196/59380

Predicting Atrial Fibrillation Relapse Using Bayesian Networks: Explainable AI Approach

2025· article· en· W4407421485 on OpenAlexvenueno aff
João Miguel Alves, Daniel Matos, Tiago Silva Martins, Diogo Cavaco, Pedro Carmo, Pedro Galvão, Francisco Moscoso Costa, Francisco Morgado, A Ferreira, P Freitas, Cláudia Camila Dias, Pedro Pereira Rodrigues, Pedro Adragão

Bibliographic record

VenueJMIR Cardio · 2025
Typearticle
Languageen
FieldMedicine
TopicAtrial Fibrillation Management and Outcomes
Canadian institutionsnot available
FundersUniversidade do Porto
KeywordsAtrial fibrillationReceiver operating characteristicBayesian networkMedicineInternal medicineMachine learningArtificial intelligenceComputer science

Abstract

fetched live from OpenAlex

Background: Atrial fibrillation (AF) is a prevalent arrhythmia associated with significant morbidity and mortality. Despite advancements in ablation techniques, predicting recurrence of AF remains a challenge, necessitating reliable models to identify patients at risk of relapse. Traditional scoring systems often lack applicability in diverse clinical settings and may not incorporate the latest evidence-based factors influencing AF outcomes. This study aims to develop an explainable artificial intelligence model using Bayesian networks to predict AF relapse postablation, leveraging on easily obtainable clinical variables. Objective: This study aims to investigate the effectiveness of Bayesian networks as a predictive tool for AF relapse following a percutaneous pulmonary vein isolation (PVI) procedure. The objectives include evaluating the model's performance using various clinical predictors, assessing its adaptability to incorporate new risk factors, and determining its potential to enhance clinical decision-making in the management of AF. Methods: This study analyzed data from 480 patients with symptomatic drug-refractory AF who underwent percutaneous PVI. To predict AF relapse following the procedure, an explainable artificial intelligence model based on Bayesian networks was developed. The model used a variable number of clinical predictors, including age, sex, smoking status, preablation AF type, left atrial volume, epicardial fat, obstructive sleep apnea, and BMI. The predictive performance of the model was evaluated using the area under the receiver operating characteristic curve (AUC-ROC) metrics across different configurations of predictors (5, 6, and 7 variables). Validation was conducted through four distinct sampling techniques to ensure robustness and reliability of the predictions. Results: The Bayesian network model demonstrated promising predictive performance for AF relapse. Using 5 predictors (age, sex, smoking, preablation AF type, and obstructive sleep apnea), the model achieved an AUC-ROC of 0.661 (95% CI 0.603-0.718). Incorporating additional predictors improved performance, with a 6-predictor model (adding BMI) achieving an AUC-ROC of 0.703 (95% CI 0.652-0.753) and a 7-predictor model (adding left atrial volume and epicardial fat) achieving an AUC-ROC of 0.752 (95% CI 0.701-0.800). These results indicate that the model can effectively estimate the risk of AF relapse using readily available clinical variables. Notably, the model maintained acceptable diagnostic accuracy even in scenarios where some predictive features were missing, highlighting its adaptability and potential use in real-world clinical settings. Conclusions: The developed Bayesian network model provides a reliable and interpretable tool for predicting AF relapse in patients undergoing percutaneous PVI. By using easily accessible clinical variables, presenting acceptable diagnostic accuracy, and showing adaptability to incorporate new medical knowledge over time, the model demonstrates a flexibility and robustness that makes it suitable for real-world clinical scenarios.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.000
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: none
Teacher disagreement score0.749
Threshold uncertainty score0.666

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0000.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0000.001
Science and technology studies0.0000.000
Scholarly communication0.0000.000
Open science0.0000.000
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.034
GPT teacher head0.326
Teacher spread0.292 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one teacher head, not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designSimulation or modeling
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations1
Published2025
Admission routes1
Has abstractyes

Explore more

Same venueJMIR CardioSame topicAtrial Fibrillation Management and OutcomesFrench-language works237,207