The CT's sample volume as an approximate, instrumental measure for density resolution in densitometry of the lung.

ABSTRACT

Ultimately CT-densitometry of the lung should give comparable results on all scanners. One prerequisite for this is the use of the same density resolution. Unfortunately, density resolution is impractical as a performance specifying parameter because it depends on the cellular material scanned. Therefore, another parameter that can be used for scanner and protocol characterization, and that does not depend on a special phantom, would be highly preferable. We investigated how well the CT's nominal sample volume (V), calculated from section thickness and in-plane spatial resolution as specified by the CT manufacturer, can serve as a simple measure, for density resolution. Six CT scanners were studied using foam and lung phantoms. On all scanners we observed for foam an approximately linear relation between density resolution and V-1/2. Density resolution on different scanners varied to some extent. These differences can be interpreted as being caused by deviations of the true sample volume from the nominal value the 95%-confidence interval runs for instance for V = 8 mm3 from 4.6 mm3 to 16.9 mm3. Acceptability of this spread depends on the consequences for parameters of clinical interest, like percentiles and pixel indexes. To evaluate this we used data from a previous patient study on the dependence of histogram parameters on sample volume. With these data it is found that large interscanner differences in histogram parameters are possible for small values of V, as used in thin-section densitometry. For larger values of V, as required for a more adequate density resolution, the differences are much smaller and probably acceptable when compared to other sources of variability in lung densitometry. In conclusion, for sections of at least 2 mm and smooth reconstruction filters, corresponding to V > or = 8 mm3, the CT's nominal sample volume might be used for interscanner and interprotocol comparison of density resolution.