Effect of digital image enhancement on radiological performance

A receiver operator characteristic (ROC) experiment tested the idea that image enhancement techniques can improve the performance of radiologists searching for subtle abnormalities, in this case, faint lung nodules, on a 512 x 512 digital video monitor. Two classes of image processing (IP) operators were selected as the objects of the test, namely, contrast enhancement and edge enhancement. The goal of the experiment was to measure the relative improvement, or lack thereof, which each operator conferred on the radiologists' discrimination performance, as compared to their performance on unenhanced images (baseline condition). Consequently, the primary dependent variable derived from each test of an IP operator was to changing the area under the ROC relative to a baseline performance level. Since the design included an ongoing determination of each observer's baseline, an estimate of the baseline level at the time at which each operator was tested was possible. An additional goal was the comparison of the performance of radiology resident switch that of faculty radiologists