Key result
Integrated radiomics model predicts post-ablation AF recurrence with ~0.87 AUC, outperforming clinical scores alone.
Why the study?
AF recurrence after RFCA remains a clinical challenge, and previous risk assessments ignored complex structural features related to atrial remodeling.
Does an integrated model combining fractal dimension, radiomics, and clinical scores improve the prediction of atrial fibrillation recurrence after radiofrequency catheter ablation?
Cohort (n=536)
Outcome assessors blinded
No
Does an integrated model combining fractal dimension, radiomics, and clinical scores improve the prediction of atrial fibrillation recurrence after radiofrequency catheter ablation?
Effect estimate: AUC 0.869 (95% CI 0.776-0.941)
Absolute Event Rate: 0.869% vs 0.66%
An integrated model combining fractal dimension, radiomics, and clinical scores accurately predicts atrial fibrillation recurrence after radiofrequency ablation, significantly outperforming traditional clinical scores.
Background Atrial fibrillation (AF) is the most common clinical arrhythmia. Radiofrequency catheter ablation (RFCA) is a primary treatment, but its success rate (50%–80%) varies by AF type and individual differences, making recurrence a key clinical challenge. Previous recurrence assessments ignored atrial remodeling-related complex structural features; fractal dimension (FD) quantifies structural complexity, and radiomics enables high-throughput feature extraction. Integrating these with clinical scores to build a composite model may improve risk stratification accuracy. Methods This single-center retrospective study enrolled 536 AF patients who underwent RFCA between January 2016 and March 2024, with follow-up through March 2025. Preoperative cardiac computed tomography venography (CTV) images and clinical data, such as APPLE and CHA₂DS₂-VASc scores, were collected via the Hospital Information System (HIS). Automated segmentation of the left atrium (LA) and left atrial appendage (LAA) was performed using specialized software, from which radiomics features were extracted and the FD was calculated. Patients were stratified and randomly divided into a training set ( n = 375) and a validation set ( n = 161) in a 7:3 ratio. Following feature selection, a multidimensional recurrence prediction model was constructed, and its parameters were optimized using five-fold cross-validation. Results Among the 536 patients, AF recurrence occurred in 139 cases (25.9%). Specifically, recurrence was observed in 111 cases (25.9%) in the training set and 28 cases (25.9%) in the test set. The integrated model, which combined the FD, radiomics features from both the LA and the LAA, as well as the APPLE and CHA₂DS₂-VASc scores, achieved an area under the curve (AUC) of 0.87 in the validation set. Its performance was significantly superior to that of models based solely on clinical scores (APPLE score AUC = 0.57; CHA₂DS₂-VASc score AUC = 0.66) or radiomics features alone (LAA model AUC = 0.77; LA model AUC = 0.62). Conclusion The integrated model incorporating multidimensional features enables effective and accurate prediction of AF recurrence after ablation, providing a practical tool for clinical individualized management. Its utility requires further validation and broader application through prospective studies.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) conducted a cohort in Atrial Fibrillation (n=536). Integrated multidimensional prediction model (fractal dimension, radiomics, and clinical scores) vs. Clinical scores alone (APPLE, CHA2DS2-VASc) or radiomics alone was evaluated on Prediction of atrial fibrillation recurrence (Area Under the Curve) (AUC 0.869, 95% CI 0.776-0.941). An integrated model incorporating fractal dimension, radiomics, and clinical scores predicted atrial fibrillation recurrence after radiofrequency ablation with an AUC of 0.87, outperforming clinical scores alone.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: