CT DENSITOMETRY OF PULMONARY NODULES

Abstract

Siegelman et al (AJR 135:1, 1980) could distinguish a high percentage of benign pulmonary nodules from malignant nodules by the high CT number of benign nodules. Others have found CT densitometry of nodules less reliable. Scanning lung nodules is technically difficult because of their small size and the large difference in density between them and the surrounding aerated lung. In order to test the reliability of CT number determinations under these conditions, we scanned a phantom containing plastic spheres of different sizes (1”, 1/2”, 1/4”) and densities, simulating soft tissue nodules surrounded by aerated lung. We used both 5-mm and 1.5-mm collimation. We found that the centering of the sphere in the scan plane is the most important variable determining the number of usable voxels for densitometry. If a 1/2” (12.5 mm) sphere is precisely centered in the scan plane, even 5-mm collimation affords a large number of usable voxels. With progressive decentering of the sphere, the fall-off in number of usable voxels is much faster for 5-mm than 1.5-mm collimation. Neither collimator gave a sufficient number of usable voxels for the 1/4” (6 mm) spheres, except for the densest sphere (Teflon, CT number 830). A computer model incorporating the scanning conditions gave an excellent match with the experimental data. We conclude that partial volume effects–the result of the interaction of the edge of the sphere with the 3-dimensional array of voxels–are the most important factors determining the number of usable voxels. With narrow collimation, decentering of the sphere in the scan plane is less important than with wider collimation. With extremely high density, even 1/4” (6 mm) spheres can be scanned with narrow collimation.