Objectives. Noncontrast computed tomography (NCCT) has emerged as the diagnostic study of choice in the evaluation of acute flank pain. Recent in vitro studies have suggested that NCCT can be used to predict the composition of urinary stones on the basis of differences in radiodensity (measured in Hounsfield units, HUs). We sought to determine whether the analysis of in vivo urinary stones seen on NCCT could predict their composition.

Methods. Between March 1997 and August 1999, 100 pure stones from patients seen at the Wilford Hall Medical Center in San Antonio, Texas were submitted for analysis. All had been visualized by NCCT before stone passage or retrieval. A General Electric High-Speed Advantage CT scanner evaluated most of these patients by a “flank pain protocol” (ie, helical technique with breath-holding at 120 kV, 200 mA with 5 mm collimation). Each scan was interpreted by one of two staff radiologists who measured the HUs for each stone. A statistical comparison was made between the stone composition and radiodensity. To allow for subsequently observed increases in radiodensity with increasing stone size regardless of composition, the HU value was divided by each stone’s largest transverse diameter in millimeters to give the HU density. A statistical comparison was then made between stone composition and HU density.

Results. No significant difference was noted between the HU values of calcium oxalate and calcium phosphate stones, and thus they were analyzed collectively as “calcium stones.” When the HU values of calcium (n = 87), uric acid (n = 7), struvite (n = 4), and cystine (n = 2) stones were compared, the overlap of ranges precluded accurate identification, and the mean HU values were not significantly different from one another. There was less overlap noted when comparing the HU densities of the stones studied, and no noncalcium stone had an HU density greater than 76 HU/mm. Using one-way analysis of variance, significant differences were noted between the mean HU density of calcium (105 ± 43) and uric acid (50 ± 24) stones (P = 0.006). A trend toward significance was found between the mean HU density of the calcium and struvite stones (53 ± 28, P = 0.073). No significant differences were found among the other stones.

Conclusions. HU density compared with the HU value alone better characterized differences in radiodensities among urinary stones; calcium stones can be distinguished from uric acid stones on the basis of this value. However, neither the HU density nor the mean HU value was able to identify urinary stones in vivo.