Comparison of epicardial adipose tissue radiodensity threshold between contrast and non-contrast enhanced computed tomography scans: A cohort study of derivation and validation

Abstract

Background and aimsEpicardial adipose tissue (EAT) volume derived from contrast enhanced (CE) computed tomography (CT) scans is not well validated. We aim to establish a reliable threshold to accurately quantify EAT volume from CE datasets. MethodsWe analyzed EAT volume on paired non-contrast (NC) and CE datasets from 25 patients to derive appropriate Hounsfield (HU) cutpoints to equalize two EAT volume estimates. The gold standard threshold (−190HU, −30HU) was used to assess EAT volume on NC datasets. For CE datasets, EAT volumes were estimated using three previously reported thresholds: (−190HU, −30HU), (−190HU, −15HU), (−175HU, −15HU) and were analyzed by a semi-automated 3D Fat analysis software. Subsequently, we applied a threshold correction to (−190HU, −30HU) based on mean differences in radiodensity between NC and CE images (ΔEATrd = CE radiodensity - NC radiodensity). We then validated our findings on EAT threshold in 21 additional patients with paired CT datasets. ResultsEAT volume from CE datasets using previously published thresholds consistently underestimated EAT volume from NC dataset standard by a magnitude of 8.2%–19.1%. Using our corrected threshold (−190HU, −3HU) in CE datasets yielded statistically identical EAT volume to NC EAT volume in the validation cohort (186.1 ± 80.3 vs. 185.5 ± 80.1 cm3, Δ = 0.6 cm3, 0.3%, p = 0.374). ConclusionsEstimating EAT volume from contrast enhanced CT scans using a corrected threshold of −190HU, −3HU provided excellent agreement with EAT volume from non-contrast CT scans using a standard threshold of −190HU, −30HU.