Hepatic fat quantification in dual-layer computed tomography using a three-material decomposition algorithm

The purpose of this study was to evaluate a three-material decomposition algorithm for hepatic fat quantification using a dual-layer computed tomography (DL-CT) and MRI as reference standard on a large patient cohort. A total of 104 patients were retrospectively included in our study, i.e., each patient had an MRI exam and a DL-CT exam in our institution within a maximum of 31 days. Four regions of interest (ROIs) were positioned blindly and similarly in the liver, by two independent readers on DL-CT and MRI images. For DL-CT exams, all imaging phases were included. Fat fraction agreement between CT and MRI was performed using intraclass correlation coefficients (ICC), determination coefficients R${}^2$, and Bland-Altman plots. Diagnostic performance was determined using sensitivity, specificity, and positive and negative predictive values. The cutoff for steatosis was 5%. Correlation between MRI and CT data was excellent for all perfusion phases with ICC calculated at 0.99 for each phase. Determination coefficients R${}^2$ were also good for all perfusion phases (about 0.95 for all phases). Performance of DL-CT in the diagnosis of hepatic steatosis was good with sensitivity between 89 and 91% and specificity ranging from 75 to 80%, depending on the perfusion phase. The positive predictive value was ranging from 78 to 93% and the negative predictive value from 82 to 86%. Multi-material decomposition in DL-CT allows quantification of hepatic fat fraction with a good correlation to MRI data. The use of DL-CT allows for detection of hepatic steatosis. This is especially interesting as an opportunistic finding CT performed for other reasons, as early detection can help prevent or slowdown the development of liver metabolic disease.