Associations between fully-automated, 3D-based functional analysis of the left atrium and classification schemes in atrial fibrillation

Figures

• 

  Fig 1. LA segmentation performed by the CNN. A) Example case of LA segmentation performed by the CNN on oblique-axial CINE images. Segmentations are shown at the time point of maximum volume (LAV_max, top), volume before atrial contraction (LAV_preA, center) and minimum volume (LAV_min, bottom). B) Schematic volume-time curve to demonstrate the respective fiducial points for minimum LA Volume (LAV_min), maximum LA Volume (LAV_max), volume before atrial contraction (LAV_preA) and local minimum between LAV_max and LAV_preA (LAV_min2). C) volume-time curve of the case shown in (A) with all fiducial points identified.

• 

  Fig 2. Overview of automatic workflow. The automated pipeline for image processing is shown in the dashed rectangle at the top: a stack of oblique axial CINE MRI series in 4CH orientation was processed by the convolutional neural network (CNN) for segmentation of the left atrium (LA). Based on the resulting time-resolved 3D model and LA volume-time curve, the characteristic time points were identified. The resulting functional LA parameters (LA emptying fractions (LAEF)) were available fully-automatic. Those parameters were investigated further in relation to stroke risk (CHA2DS2VASc), atrial fibrillation (AF) Burden (AF Burden Score) and other established AF risk factors.

• 

  Fig 3. Study cohort flowchart. * We performed an additional comprehensive analysis of all patients independent from rhythm status during MRI (total n = 151, study cohort (n = 102) and patients with AF during MRI (n = 49)). However, assessment of active LA contraction was not possible in this cohort.

• 

  Fig 4. LA functional parameters and stroke risk based on CHA2DS2VASc. LAEF_total, LAEF_active and LAEF_passive in relation to low and increased stroke risk (based on CHA2DS2VASc score of ≤ 1 or ≥ 2). All three parameters were significantly lower in the group with increased stroke risk based on the CHA2DS2VASc. LAEF–left atrial emptying fraction.

• 

  Fig 5. LA functional parameters and AFBS. LAEF_total, LAEF_active and LAEF_passive in relation to the AFBS categories. Both LAEF_total and LAEF_active significantly decreased with AFBS increase while LAEF_passive did not show significant differences. LAEF–left atrial emptying fraction. AFBS–AF Burden score. Post hoc tests: * = p<0.05, ** = p<0.01.