IMAGE SELECTION FOR COMPUTED-TOMOGRAPHY OF THE CHEST - A SAMPLING APPROACH

RATIONALE AND OBJECTIVES. High-resolution computed tomography (HRCT) is useful to characterize the presence and extent of lung abnormalities. Contiguous HRCT images of the chest would require 200 images, which is not for practical due to 1) the extensive examination and reading time, and 2) radiation exposure. This article presents a methodologic framework to select the appropriate number of HRCT images to estimate any quantitative parameter with a desired precision. Alternative sampling strategies are introduced, and the sample size requirements are given. METHODS AND RESULTS. Sample size requirements are developed for the percentage of emphysematous lung using simple random and stratified random sampling. The effect of the number of strata on the sample size requirement also is shown. The marked reduction in the number of HRCT images using different types of sampling plans illustrates the power of sampling techniques. CONCLUSIONS. Proper stratification is critical to reduce the sample size requirement and to avoid missing kev abnormalities, which is particularly critical in the early stages of any disease process when intervention may be most useful. Prior knowledge of the disease is useful in determining the optimum number and location of strata, and can be obtained from available chest radiographs, pulmonary function tests, radionuclide studies, and clinical parameters.