Virtual Unenhanced Images of the Abdomen With Second-Generation Dual-Source Dual-Energy Computed Tomography Image Quality and Liver Lesion Detection

Purposes: We aimed to analyze the image quality, CT number, artifacts, radiation dose reduction, and coverage in abdominal virtual unenhanced (VU) and conventional unenhanced (CU) data sets obtained with a second-generation dual-energy computed tomography and to compare the sensitivity of VU and CU data sets for liver lesion detection. Materials and Methods: One hundred eleven patients underwent triphasic abdominal CT examination that included single-energy CU and dual-energy arterial and portal phases. Virtual unenhanced images were generated from arterial (AVU) and portal (PVU) phases. Two abdominal radiologists independently (a) analyzed the image quality using a 5-point scale, CT number, and noise of the abdominal organs and (b) identified and characterized liver lesions in CU, AVU, and PVU. The triphasic abdominal examination was considered the reference standard for liver lesion detection and characterization. Results: The quality of VU images was mostly excellent but not as good as CU images (P < 0.05). The mean (SD) image quality classified by readers 1 and 2 was 4.9 (0.2; range, 4.7-5.0) and 4.8 (0.5; range, 4.4-4.9) for CU, 4.7 (0.4; range, 4.3-4.9) and 4.6 (0.4; range, 4.2-4.8) for AVU, and 4.7 (0.6; range, 4.1-4.8) and 4.6 (0.4; range, 4.2-4.8) for PVU, respectively. The CT number of the liver, the spleen, the pancreas, the renal cortex and medulla, the aorta, and the retroperitoneal fat was higher in AVU and PVU than in CU images. A total of 270 liver lesions were found in 76 patients. Portal virtual unenhanced data set was more sensitive than AVU and CU were for hypodense lesions smaller than 1 cm. Reader 1 correctly detected 72/144 (50.0%), 61/144 (42.4%), and 55/144 (38.2%) hypodense lesions with PVU, AVU, and CU, respectively; and reader 2 correctly diagnosed 70/144 (48.7%), 62/144 (43.0%), and 53/144 (36.8%) lesions with PVU, AVU, and CU, respectively. Conventional unenhanced data set was more sensitive than AVU or PVU was for small calcified lesions. Reader 1 detected 24/40 (60.0%), 24/40 (60.0%), and 40/40 (100%) with PVU, AVU, and CU, respectively; and reader 2 detected 27/40 (67.5%), (25/40) 62.5%, and 40/40 (100%) with PVU, AVU, and CU, respectively. The dose reduction achieved by omitting the unenhanced acquisition was a mean (SD) of 21.1% (1.2%; P < 0.01). Conclusions: Second-generation abdominal VU data sets, despite a mostly excellent image quality, still cannot replace CU images in clinical practice because of limitations in material subtraction. Advances in Knowledge: 1. Second-generation abdominal VU data sets yield excellent image quality but are still slightly lower than that of CU. 2. More small hypodense liver lesions were detected in VU images than in CU images; however, a significant number of small calcified lesions were not identified in VU images. 3. Second-generation abdominal VU images are still not ready to replace CU images in clinical practice. Implications for Patient Care: 1. Using VU images in abdominal studies makes it possible to reduce the total radiation dose delivered to patients who need multiphasic acquisition by avoiding precontrast scan. 2. Erroneous material subtraction and incomplete abdominal coverage represent a limit in VU data set application in clinical routine.