Automatic volumetric estimates of the left and right atrium using dynamic PET

Left (LAV) and right (RAV) atrial volumes are independent markers of cardiovascular risk in heart failure. Simultaneous assessment of myocardial blood flow (MBF) and atrial volumes might improve the clinical utility of cardiac PET. The aim of this study was to develop and validate an automated method for obtaining atrial volumes from dynamic myocardial perfusion PET scans without ECG-gating.

The atria were segmented automatically using first-pass data from [¹⁵O]-water PET at rest, combining voxel-wise images of bolus area-under-curve and arrival time. Data of multiple patient cohorts were analyzed: retrospective method development in 36 subjects with systolic heart failure with prospective validation in 59 subjects with same-day echocardiograms (primary hypertrophic cardiomyopathy (n = 25), suspected or known cardiac amyloidosis (n = 25) and healthy controls (n = 9)). Test-retest repeatability of PET was assessed in clinical chest pain patients scanned twice on the same day (n = 15). Segmentation was successful in all scans (n=125). PET and echocardiography correlated for LAV in the development cohort (r=0.83, p<0.001) and in the validation cohort (LAV: r=0.83, RAV: r=0.77, both p<0.001). In alignment with echocardiography, PET identified statistically significant differences between healthy controls and subjects with hypertrophied hearts for LAV index: 26 (interquartile range: 24-29) versus 41 (32-51) ml/m², p<0.001, and RAV index: 31 (25-43) versus 48 (38-61) ml/m², p=0.003). Test-retest reproducibility was excellent for LAV (intraclass correlation coefficient ICC=0.96, repeatability coefficient RPC=8.6 ml/m²) and for RAV (ICC=0.96, RPC=11.6 ml/m²).

Left and right atrial volumes can be extracted automatically, accurately and reproducibly using dynamic PET.