What does this research mean for the field?

A machine learning-assisted nomogram achieves an AUC of 0.88 for predicting left atrial appendage thrombosis and severe spontaneous echo contrast in patients with non-paroxysmal atrial fibrillation, significantly outperforming the CHA2DS2-VASc score. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

The aim is to develop a nomogram using machine learning for predicting LAA thrombosis and severe SEC in non-paroxysmal NVAF patients.

April 12, 2026BMC Cardiovascular DisordersOpen Access

Machine learning nomogram achieves 0.88 AUC for predicting LAA thrombosis and severe SEC, outperforming CHA2DS2-VASc.

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Non-paroxysmal non-valvular atrial fibrillation (n=327). Intervention: Machine learning-assisted nomogram vs. CHA2DS2-VASc score. Primary outcome: Presence of LAA thrombosis and severe SEC (grade 4) detected by TEE (95% CI 0.83-0.93, p=<0.001).

View Full Paper

Ask AI

Bookmark

Why the study?

Detection of left atrial appendage thrombosis and severe spontaneous echo contrast is crucial for risk stratification and management in patients with non-valvular atrial fibrillation.

Does a machine learning-derived nomogram improve the prediction of LAA thrombosis or severe SEC compared to the CHA2DS2-VASc score in patients with non-paroxysmal NVAF?

Population

327 patients with non-paroxysmal NVAF

Comparison

Machine learning-based nomogram vs CHA2DS2-VASc score

Design

Retrospective observational study

Key result

A machine learning-assisted nomogram achieved an AUC of 0.88 for predicting left atrial appendage thrombosis and severe spontaneous echo contrast, significantly outperforming the CHA2DS2-VASc score.

Authors

HWHao Wang JFJunyu Fan BZBingyuan Zhou

Discussion

Member takes

Overview

May improve LAA thrombosis risk stratification in non-paroxysmal AF; extends CHA2DS2-VASc but hypothesis-generating pending external validation.

Key Points

The aim is to develop a nomogram using machine learning for predicting LAA thrombosis and severe SEC in non-paroxysmal NVAF patients.
Retrospective enrollment of 327 patients with non-paroxysmal NVAF
Collection of 34 clinical and echocardiographic variables
Application of three machine learning approaches (SVM-RFE, Boruta, LASSO) for feature selection
Evaluation of model performance using ROC curves, calibration curves, DCA, and CIC
Reclassification analysis performed using NRI and IDI
LAA thrombosis or severe SEC detected in 12.2% of patients
AUC values for machine learning algorithms: 0.88 (SVM-RFE), 0.89 (LASSO), 0.89 (Boruta)
Final nomogram AUC of 0.88, significantly higher than CHA2DS2-VASc score (AUC = 0.68)
High net benefits shown in DCA and CIC analysis across various threshold probabilities
NRI of 0.957 and IDI of 0.254 compared to CHA2DS2-VASc score

Study Design

Type

Observational (n=327)

Multicenter

Structured PICO

Does a machine learning-derived nomogram improve the prediction of LAA thrombosis or severe SEC compared to the CHA2DS2-VASc score in patients with non-paroxysmal NVAF?

Population

327 patients with non-paroxysmal non-valvular atrial fibrillation (NVAF)

Intervention

Nomogram constructed using machine learning-assisted feature selection (SVM-RFE, Boruta, and LASSO)

Comparator

CHA2DS2-VASc score

Outcome

Detection of left atrial appendage (LAA) thrombosis or severe spontaneous echo contrast (SEC)surrogate

Main Result

Absolute Event Rate: 0.88% vs 0.68%

p-value: p=<0.001

A novel machine learning-derived nomogram significantly improves the prediction of LAA thrombosis and severe SEC compared to the standard CHA2DS2-VASc score in patients with non-paroxysmal NVAF.

Limitations

Small sample size with only 40 patients diagnosed with LAA thrombosis/severe SEC
Exploratory study
Clinical reliability requires further validation in larger, independent, prospective cohorts

Cite This Study

Wang et al. (2026) conducted an observational in Non-paroxysmal non-valvular atrial fibrillation (n=327). Machine learning-assisted nomogram vs. CHA2DS2-VASc score was evaluated on Presence of LAA thrombosis and severe SEC (grade 4) detected by TEE (95% CI 0.83-0.93, p=<0.001). A machine learning-assisted nomogram achieved an AUC of 0.88 for predicting left atrial appendage thrombosis and severe spontaneous echo contrast, significantly outperforming the CHA2DS2-VASc score.

synapsesocial.com/papers/69db375f4fe01fead37c54ec https://doi.org/10.1186/s12872-026-05805-w

View Full Paper

Ask AI

Bookmark

April 12, 2026BMC Cardiovascular DisordersOpen Access

Machine learning nomogram achieves 0.88 AUC for predicting LAA thrombosis and severe SEC, outperforming CHA2DS2-VASc.

View Full Paper

Ask AI

Bookmark

Why the study?

Detection of left atrial appendage thrombosis and severe spontaneous echo contrast is crucial for risk stratification and management in patients with non-valvular atrial fibrillation.

Does a machine learning-derived nomogram improve the prediction of LAA thrombosis or severe SEC compared to the CHA2DS2-VASc score in patients with non-paroxysmal NVAF?

Population

327 patients with non-paroxysmal NVAF

Comparison

Machine learning-based nomogram vs CHA2DS2-VASc score

Design

Retrospective observational study

Key result

A machine learning-assisted nomogram achieved an AUC of 0.88 for predicting left atrial appendage thrombosis and severe spontaneous echo contrast, significantly outperforming the CHA2DS2-VASc score.

Authors

HWHao Wang JFJunyu Fan BZBingyuan Zhou

Discussion

Member takes

Overview

May improve LAA thrombosis risk stratification in non-paroxysmal AF; extends CHA2DS2-VASc but hypothesis-generating pending external validation.

Key Points

The aim is to develop a nomogram using machine learning for predicting LAA thrombosis and severe SEC in non-paroxysmal NVAF patients.
Retrospective enrollment of 327 patients with non-paroxysmal NVAF
Collection of 34 clinical and echocardiographic variables
Application of three machine learning approaches (SVM-RFE, Boruta, LASSO) for feature selection
Evaluation of model performance using ROC curves, calibration curves, DCA, and CIC
Reclassification analysis performed using NRI and IDI
LAA thrombosis or severe SEC detected in 12.2% of patients
AUC values for machine learning algorithms: 0.88 (SVM-RFE), 0.89 (LASSO), 0.89 (Boruta)
Final nomogram AUC of 0.88, significantly higher than CHA2DS2-VASc score (AUC = 0.68)
High net benefits shown in DCA and CIC analysis across various threshold probabilities
NRI of 0.957 and IDI of 0.254 compared to CHA2DS2-VASc score

Study Design

Type

Observational (n=327)

Multicenter

Structured PICO

Does a machine learning-derived nomogram improve the prediction of LAA thrombosis or severe SEC compared to the CHA2DS2-VASc score in patients with non-paroxysmal NVAF?

Population

327 patients with non-paroxysmal non-valvular atrial fibrillation (NVAF)

Intervention

Nomogram constructed using machine learning-assisted feature selection (SVM-RFE, Boruta, and LASSO)

Comparator

CHA2DS2-VASc score

Outcome

Detection of left atrial appendage (LAA) thrombosis or severe spontaneous echo contrast (SEC)surrogate

Main Result

Absolute Event Rate: 0.88% vs 0.68%

p-value: p=<0.001

A novel machine learning-derived nomogram significantly improves the prediction of LAA thrombosis and severe SEC compared to the standard CHA2DS2-VASc score in patients with non-paroxysmal NVAF.

Limitations

Small sample size with only 40 patients diagnosed with LAA thrombosis/severe SEC
Exploratory study
Clinical reliability requires further validation in larger, independent, prospective cohorts

Cite This Study

synapsesocial.com/papers/69db375f4fe01fead37c54ec https://doi.org/10.1186/s12872-026-05805-w

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

View Full Paper

Ask AI

Bookmark