CT Density Measurements for Characterization of Adrenal Tumors Ex Vivo: Variability Among Three CT Scanners
AJR 2004; 182:671-675.
Stadler A, Schima W, Prager G, Homolka P, Heinz G, Saini S, Eisenhuber E, Niederle B.
OBJECTIVE: Many studies have suggested that Hounsfield measurements on unenhanced CT can reliably differentiate adrenal adenomas from nonadenomas using a scanner-independent threshold level. The purpose of this study was to determine whether establishment of a scanner-independent threshold for differentiation of adenomas from nonadenomas is technically feasible.
MATERIALS AND METHODS: Surgically resected adrenal tumor specimens (total, seven; adenomas, three; nonadenomas, four; size range, 17-76 mm), were placed in an anthropomorphic phantom. Lesion specimens were scanned with one MDCT and two single-detector scanners. Scanning protocols for all three scanners included variations in kilovoltage (140, 120, and 80 [Somatom Plus 4, Somatom VolumeZoom] or 100 [Tomoscan AV] kVp) and slice thickness. Hounsfield measurements were performed on exactly matched slices using regions of interest of a constant size.
RESULTS: The difference in lesion Hounsfield measurements among scanning protocols with 140, 120, and 100/80 kVp was up to 6.2 H for the adenoma group and up to 3.8 H for the nonadenoma group. The comparison of the Tomoscan AV and the Somatom Plus 4 scanners showed a mean difference of 2.6 H at 120 kVp and of 4.6 H at 140 kVp. The differences between the Tomoscan AV and Somatom VolumeZoom scanners were 1.7 and 3.6 H for 120 and 140 kVp, respectively. Between the two Somatom scanners, the divergence was 2.9 and 3.3 H for the two kilovoltage settings. Differentiation between adenomas and nonadenomas was better at lower kilovoltage. Slice thickness did not affect the CT density measurements significantly.
CONCLUSION: Significant differences in CT density measurements of adrenal tumors may occur when different CT scanners or imaging protocols are used. The dependence of measurements on scanner type and scanning technique makes the recommendation of a universal, scanner- and protocol-independent threshold problematic.