Abstract
OBJECTIVE: The purpose of this study was to assess the accuracy of emphysema measurements obtained from systematic samples of evenly spaced CT images compared with measurements obtained from the entire scan. MATERIALS AND METHODS: Evenly spaced transverse sections from the CT studies of 136 heavy smokers who underwent screening for lung cancer in the National Lung Screening Trial and of 112 subjects who underwent imaging before lung volume reduction surgery were sampled retrospectively. The samples were acquired both by selection of specific numbers of evenly spaced images and by selection of images at specific distance intervals. The percentage of lung pixels with attenuation below specific thresholds was used as an emphysema index. The image sample error was determined as the difference in emphysema index between the image samples and the entire scan. RESULTS: The largest absolute image sample errors in the National Lung Screening Trial cohort with image sample sizes of five, 10, and 20 were 2.2, 0.8, and 0.5 index percentage points, respectively, at 1-mm section thickness (-960 HU threshold), and 2.6, 1.1, and 0.5 index percentage points at 5-mm section thickness (-930 HU threshold). The largest errors in the lung volume reduction surgery cohort for image sample sizes of five and 10 were 5.6 and 2.3 index percentage points at 8- to 10-mm section thickness (-900 HU threshold). Image sample errors were equivalent for the two sampling methods. CONCLUSION: Systematic sampling resulted in very small errors in emphysema quantification and may be useful for decreasing radiation exposure in clinical research studies of emphysema.