Rachel J. Toomey BSC, Presenter: Nothing to Disclose

John Mark Ryan PhD, Abstract Co-Author: Nothing to Disclose

Mark McEntee, Abstract Co-Author: Nothing to Disclose

Mike Evanoff PhD, Abstract Co-Author: Nothing to Disclose

Anthony Hayes, Abstract Co-Author: Nothing to Disclose

Patrick C. Brennan PhD, Abstract Co-Author: Nothing to Disclose

The fovea centralis region of the human eye is densely packed with photoreceptors and provides an area of high-resolution vision (around 50 cycles/degree for high-contrast imagery). This constitutes a “useful field of view” and is described as occupying approximately 5 degrees of visual angle. We report an observer-performance study that aimed to determine whether improved detection of pulmonary nodules results from presenting chest radiographs with the entire image contained within this “useful visual field”.

A total of 18 highly experienced radiologists assessed 30 chest images for the presence or absence of lung nodules. Viewing distance was constrained to approximately 70cm from the image. Each observer read the same images, presented in random order, under two or all three possible viewing conditions. A counterbalanced methodology was used to reduce potential memory effects. Readings were performed under the following three image-presentation conditions: a control condition of full size and resolution (1500 x 1500 pixels, approximately 25 degrees visual angle); reduced image resolution (300 x 300 pixels) and size-minimized to approximately 5 degrees visual angle; full size, but containing the same spatial information as the size-minimized images (300 x 300 pixels linearly interpolated to 1500 x 1500 pixels). All images were presented within a grey border of the mean digital driving level of the images to minimize any adaptation effects, and to reduce any focusing effects when viewing the interpolated images. Receiver operating characteristic (ROC) analysis was applied to the results of the readings.

When both readers and cases were treated as random, no statistically significant differences existed between any display conditions. When cases were treated as fixed, the interpolated images yielded statistically higher scores than the minimized images (0.7475 vs 0.6579; p≤0.0171).

Presenting chest images so that they may be fully viewed in the foveal vision does not appear to improve the detection of pulmonary nodules. Larger image presentation may yield better performance. The spatial resolution of the images did not impact nodule detection in this study.

The presentation of chest radiographs so that they may be viewed in their entirity within the central 5 degrees of visual angle does not improve, and may detract from, nodule detection.

Toomey, R, Ryan, J, McEntee, M, Evanoff, M, Hayes, A, Brennan, P, The Impact of Retinal Image Size and Spatial Resolution on Chest Nodule Detection.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8007385.html