Reduced left atrial strain from 4-dimensional computed tomography predicts atrial fibrillation recurrence after...

Existing models that use clinical history and cardiac imaging data remain inadequate for accurate prediction of the success of catheter ablation for atrial fibrillation (AF). Local strain in the left atrium (LA) from 4-dimensional computed tomography (4DCT) has potential to add value to models of ablation outcome.

This study aimed to investigate whether local LA strain from 4DCT adds predictive value to models of AF recurrence.

This study retrospectively analyzed preablation 4DCT data from 261 AF patients at a single center. An image-processing pipeline automatically computed volumetric LA segmentations and endocardial strain in 24 segments of the LA wall for 1 cardiac cycle. Machine learning models were trained on LA strain, demographic, and clinical features to predict AF recurrence within 12 months after ablation. Spatial strain maps were used to perform comparisons of strain distribution between patient subgroups.

Random forest models produced the best results. The addition of a 24-segment regional strain to demographic and clinical features increased the area under the receiver operator characteristic curve to predict AF recurrence from 0.738 to 0.813. Low strain in the inferior wall was among the features most significantly correlated with a poor ablation outcome (P < .01). Subjects with persistent AF had reduced regional strain compared with paroxysmal AF, as did subjects with a history of previous ablations compared with first-time ablation candidates (P < .01).

Patient subgroups display different spatial patterns of LA strain. Regional LA strain analysis adds independent predictive power to models of AF recurrence after thermal catheter ablation. Weak inferior wall motion preablation is particularly associated with poor ablation outcome.