The purpose of this study was to evaluate intravenous (IV) contrast-enhanced single-source rapid kilovolt (peak)–switching dual-energy (RSDE) multidetector computed tomography (CT) material density assessment of hepatic steatosis compared to conventional unenhanced (CU) Multidetector computed tomography (MDCT).
This is an institutional review board–approved intrapatient study of 363 consecutive adults (189 men, 174 women; mean age, 59 years) evaluated with multiphasic IV abdominal RSDE. Material density virtual unenhanced water and fat hepatic parenchymal values were measured and correlated to Hounsfield units (HUs) on CU CT using linear regression. Study population was dichotomized into steatotic or nonsteatotic liver parenchyma on the basis of CU liver-spleen (L-S) difference. The RSDE fat(-iodine) values (in milligram per milliliter) were compared (t test), correlated to the L-S difference in HU, and a milligram-per-milliliter fat threshold for clinically significant steatosis was calculated using receiver operator curve (ROC) analysis.
Regression analysis revealed r value of 0.86 for mg/mL water (P < 0.001) and 0.87 for milligram-per-milliliter fat (P < 0.001). Twenty-seven participants were excluded from the L-S analysis (splenectomy). A total of 107 (32%) had steatosis (mean L-S, − 6.3; mean fat(-iodine) milligram per milliliter, 1018.4); 229 (68%) had no steatosis (mean L-S, 9.4; milligram per milliliter, 1028.4 [P < 0.001]). The RSDE fat material density measurement correlated to L-S less than 1 with r value of 0.74 (P < 0.001), with an area under receiver operator curve of 0.847. A threshold of 1023-mg/mL fat had 71% sensitivity and 80% specificity, and a threshold of 1027-mg/mL fat had 90% sensitivity and 61% specificity for steatosis.
The RSDE milligram-per-milliliter fat values correlate well with hepatic steatosis defined by the L-S difference less than 1 on conventional MDCT. A threshold of 1027 mg/mL can identify 90% of steatotic livers when post-IV contrast RSDE is used, without obtaining additional CU scans. However, regression equations were not helpful to convert an individual participant’s milligram-per-milliliter fat or milligram-per-milliliter water-derived from RSDE material density images to CU MDCT HU for the estimation of liver fat content.
From the *Department of Radiology, Duke University Medical Center, Durham, NC; †Department of Medicine, UCLA Medical Center, Los Angeles, CA; Departments of ‡Radiology, and §Biostatistics, University of Alabama at Birmingham, Birmingham, AL.
Received for publication August 8, 2013; accepted October 4, 2013.
Reprints: Desiree E. Morgan, MD, Department of Radiology, 619 19th St S, Birmingham, AL 35294 (e-mail: email@example.com).
The authors declare no conflict of interest.