Influence of CT image size and format on accuracy of lung nodule detection.

Abstract

To evaluate the effect of reducing image size on observers' ability to detect lung nodules on computed tomographic (CT) scans. Stimuli were 80 single sections from 13 normal chest CT studies. On half of the images, 3-5-mm-diameter nodules were superimposed electronically at random locations. Four observers viewed images in six formats and sizes that ranged from 6 on 1 (133 x 133 mm) to 80 on 1 (40 x 40 mm). The images were viewed at a fixed distance of 55 cm and at an unrestricted, variable distance. With the fixed viewing distance, nodule detection decreased with smaller image sizes. The area under the receiver operating characteristic curve (Az) decreased from 0.857 for the 6-on-1 format to 0.671 for the 80-on-1 format (P = .0001). With a variable viewing distance, Az decreased from 0.884 to 0.834 across all formats (difference not statistically significant). However, there was a significant drop in performance with the smallest images (P < .05). Overall, Az for the fixed and variable viewing distances was significantly different (P < .001). Reducing image size leads to decreased lung nodule detection on CT scans viewed at a fixed distance; however, the observer can compensate for the smaller image by adjusting the viewing distance.