Digital image processing of chest radiographs to compensate for the limitations of video displays

The acceptance of digital workstations as primary diagnostic devices for chest radiographs will be precluded if there is a reduction in the radiologist's accuracy of diagnosis compared to that achieved with conventional screen-film images. Reduction in diagnostic efficacy is believed to be partially due to a reduction in contrast resolution on video monitors. We present the results of a pilot study that tests the ability of the contrast-enhancement algorithm artifact-suppressed adaptive histogram equalization (ASAHE) to compensate for reduced contrast resolution. The ASAHE algorithm is compared to a computed radiographic algorithm that previously delivered observer performance inferior to conventional screen-film images. The algorithms are compared on the basis of five readers interpreting an image set consisting of 45 dllnical cases, 23 of which are confirmed as demonstrating pneumothoraces. Detection efficacy, measured by the area under a receiver operating characteristic (ROC) curve, is not significantly different for the two algorithms. The average ROC curves for the algorithms have different shapes, suggesting that the ASAHE algorithm is affecting diagnostic performance in a way that is not well understood. The results of the pilot study indicate that a test with higher statistical power would need to be performed using this algorithm to form a final estimate of its usefulness.