OBJECTIVE
Radiographic exposure has been thought to have little impact on the diagnostic quality of chest computed radiography because of automatic digital control of global optical density. The objectives of this study were to compare images obtained with two different exposures in computed radiography with conventional and asymmetric screen-film images of the chest for the detection of simulated lung nodules by use of receiver operating characteristic analysis and to relate differences in observer performance to parameters of image noise measured for each receptor condition.
MATERIALS AND METHODS
At 110 kVp (fixed), exposures for the two screen-film systems were those necessary to achieve adequate optical densities over the lung and mediastinal regions of an anthropomorphic phantom. The two exposures used for the computed radiographs corresponded to the exposure used for the conventional chest screen-film system and an exposure 22% lower. An anthropomorphic phantom constructed of materials matched to the muscle, lung, and bone attenuation of a muscular adult man was used. Soft-tissue-equivalent plastic nodules of various sizes were secured at multiple sites on the phantom to simulate lung nodules. The chest phantom was imaged in 50 configurations with a total of 70 superimposed nodules. The perceptual performances of five radiologists were compared by use of receiver operating characteristic analysis. The signal-to-noise ratio in the mediastinum and the coefficient of variation of noise were measured for all four image conditions by use of a step wedge technique to provide an explanation for differences in diagnostic accuracy.
RESULTS
We found no significant differences in the detection of lung nodules between the two screen-film systems or between the conventional screen-film images and the standard-exposure computed radiographs. However, there was a significant decrease in nodule detection on computed radiographs obtained at a reduced exposure; this result was associated with a 21% decrease in the signal-to-noise ratio.
CONCLUSION
Our results show that underexposure of computed radiographs decreases the detection of low-contrast objects such as lung nodules. Although consistent global optical density on computed radiographs is achieved over a wide range of exposures, the alterations in signal-to-noise ratio that result from underexposure can reduce the diagnostic quality of computed radiographs.