Atrial Fibrillation Recurrence Prediction with Cardiac CT Features

297 patient records with 37 clinical and cardiac computed tomography imaging features were used to develop machine learning models for predicting recurrence after catheter ablation for atrial fibrillation. The dataset, created by Xiaonan Han and published on figshare under CC-BY-4.0, includes distinct models for paroxysmal and persistent AF types. The models achieved AUC scores of 0.831 and 0.917, respectively, with explainable AI techniques applied.

Use Cases

• Predicting recurrence of paroxysmal atrial fibrillation based on right atrial appendage short diameter and other CT features.
• Predicting recurrence of persistent atrial fibrillation based on right atrial volume and other CT features.
• Developing explainable AI clinical tools for personalized prognosis assessment post-ablation.
• Analyzing nonlinear relationships and interaction effects between cardiac structural parameters and recurrence risk.

Strengths

• Includes 37 clinical and cardiac CT imaging features per patient.
• Models were trained on 297 patient records, with 230 in the paroxysmal AF group and 67 in the persistent AF group.
• Model performance is quantified with AUC scores of 0.831 and 0.917.
• Employed SHAP framework for model interpretability.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is very small at 45.6 KB, indicating limited scope.

Provenance

Source

figshare

Collection Method

Retrospective collection of clinical and cardiac CT imaging features from patients undergoing radiofrequency catheter ablation.

Freshness

Last updated 2026-04-29 05:57:12; freshness should be verified.