On the Concordance Between Cardiac Magnetic Resonance and Computed Tomography for Left Heart Function Assessment

To assess the agreement between cardiac computed tomography (CCT) and cardiac magnetic resonance (CMR) for measuring left ventricular (LV) and left atrial (LA) function, and to evaluate the influence of segmentation approach and volume calculation method.

This retrospective study included 24 patients (mean age 74.2±10.9 y; 54% male) who underwent CMR and multiphase CCT within 14 days. LV and LA end-diastolic volume (EDV), end-systolic volume (ESV), ejection fraction (EF), and LV mass were measured using CCT voxel-based volumetry with papillary-inclusion (PIS) and exclusion (PES) segmentations and then compared with CMR. Simulations of Simpson's Area-Length (AL) and Disk-Summation (DS) techniques were performed on reformatted CCT images using increasing slice spacings (1 to 32 mm). Agreement was evaluated using correlation coefficients, intraclass correlation coefficients (ICC), percent error, and Bland-Altman analysis.

CCT demonstrated excellent correlation with CMR for LV-EDV, ESV, EF, and mass (r=0.82 to 0.98; ICC=0.72 to 0.94). PES yielded no EF bias, while PIS overestimated EF by 7.0%. LV mass was consistently overestimated by 39 to 52 g ( P <0.01). LA volumes showed moderate-to-strong correlation (r=0.70 to 0.90), but poor-to-moderate agreement (ICC=0.21 to 0.55). Simulated LA Simpson-DS measurements with slice spacings ≤8 mm preserved agreement with voxel-based values (ICC ≥0.99).

CCT can accurately assess LV function when compared with CMR when using voxel-based methods and consistent papillary segmentation approaches. Biplane and measurements with wide slice spacings reduce agreement, warranting standardization for clinical interchangeability.