Impact of postprocessing on the detection of simulated pulmonary nodules with digital radiography

The authors compared the impact of five postprocessing algorithms on diagnostic performance in the detection of simulated pulmonary nodules on storage phosphor-based digital chest radiographs. Tissue equivalent paraffin nodules (0.5-2.5 cm diameter) were randomly positioned over the chest of a normal volunteer. Receiver operating characteristics (ROC) analysis of a total of 2500 observations by five readers indicated that the default unenhanced image having the appearance of a conventional chest radiograph (ROC area = 0.87 +/- 0.05) was as good as an image with moderate enhancement of medium frequencies (ROC area = 0.85 +/- 0.03), an image with reversed gray scale polarity (ROC area = 0.84 +/- 0.02), an image with reversed gray scale and moderate enhancement of medium frequencies (ROC area = 0.87 +/- 0.03), and an image with a linear rather than a sigmoid gradation curve and incorporating moderate enhancement of medium frequencies (ROC area = 0.87 +/- 0.03). The authors conclude that the specific algorithms they tested had no effect on the detection of pulmonary nodules.