Identification of atrial myopathy and atrial fibrillation recurrence after ablation using 3D left atrial phasic strain from retrospective gated computed tomography

Figures

• 

  Figure 1. Pipeline for phasic 3D LA motion and strain measurement from RGCT scans. (A): Input data were ten reconstructed RGCT scans at 10% increments of the R-R interval. (B): Patient-specific LA models were created from the end- diastolic (t = 0%) CT frame. LA body regions were standardized using the UAC system. (C): 3D Feature tracking estimated 3D LA motion from the RGCT scans. The tracked 3D LA endocardial outline is shown at ventricular end-diastole (t = 0%), end-systole (t = 40%) and early diastole (t = 70%). The end-diastolic LA contour (in green) is repeated at later CT frames to explicitly outline the change in LA shape. White arrows outline LA enlargement at end-systole. (D): Global and regional phasic strain analysis from the 3D motion. Phases were defined using the global strain transient (black line). Ant, anterior; inf, inferior; lat, lateral; post, posterior; sept, septum.

• 

  Figure 2. Comparison of the global phasic strains and peak SRs from 3D LA motion between AF recurrence vs. no recurrence. Contractile parameters displayed the most significant differences in patients with successful AF ablation endpoint. P-value legend, ns: not significant; *: P < 0.05, **: P < 0.01.

• 

  Figure 3. ROC analysis compares predictive ability of conventional and RGCT-derived 3D LA biomechanics features to identify (A) non-paroxysmal AF and (B) AF recurrence. Mean and 95% CIs for AUC values are presented for the most predictive global and regional features from RGCT-derived 3D LA biomechanics and conventional features. Passive and contractile features separated non-paroxysmal AF and recurrent AF, respectively.

• 

  Figure 4. Results of K-means clustering using the RGCT-derived 3D strain and SR parameters. (A): The two identified clusters shown in the first two principal components’ space with 95% CI ellipses. No clinical variables were used to identify the clusters. (B): Mean strain (top) and SR (bottom) curves in both clusters. (C): Reservoir strain maps from representative cases within the clusters. Extent of 3D LA motion at ventricular end-systole is shown by the semi-transparent shell. Blue and orange represent smaller and larger LA stretch and hence reservoir strain, respectively.