Emily Eads MD, Presenter: Nothing to Disclose

Jay Alan Baker MD, Abstract Co-Author: Consultant, Siemens AG

Samuel Richard PhD, BSC, Abstract Co-Author: Nothing to Disclose

Huiman Barnhart PhD, Abstract Co-Author: Nothing to Disclose

Ehsan Samei PhD, Abstract Co-Author: Advisory Board, Ion Beam Applications, SA Consultant, Siemens AG Research grant, Siemens AG Research grant, General Electric Company Research grant, Carestream Health, Inc

The human eye is known to be most sensitive to a wavelength of 555 nm (green). This study investigated the potential increase in lesion detection accuracy with color monitor displays compared to conventional grayscale display.

A digital image was acquired of the Contrast Detail Mammography phantom (CDMAM, Artinis), which is an array of cells with discs of varying contrast and diameter. Each cell contains a central reference disc and a disc in one of four corners. The task was to identify the corner containing the second disc. Selected cells from the phantom containing discs of a low conspicuity and a range of disc contrast and diameter were individually displayed on a high luminance color LCD monitor (Eizo, Radiforce RX320, USA) in different colormaps (grayscale, red, green, and blue) while maintaing equal luminance. A total of 26 detection tasks were randomly displayed for each of the 4 colormaps. Each image was shown 5 times by randomly flipping images from left-right or up-down to yield a total of 520 observations per observer. Customized software was used to display images and record observer response. To mitigate the effect of a “learning curve”, each observer started with a training set containing 200 images which were not used in the analysis. None of the 18 observers was colorblind.

Overall, across all four color displays, none of the observers performed best with the grayscale display. Based on raw score, six (6) individuals performed best with green, 6 with red, 4 with blue, and 2 equally best with green and blue colormap displays. A pair-wise comparison of the pooled data for all 18 observers indicated a statistically significant improvement in percentage correct with a green colormap (73.3%) compared to a gray colormap (70.8%) (p-value = 0.009). Higher accuracy was also measured for red (72.5%) and blue (72.6%), however these differences were not statistically significant.  

Accuracy of lesion detection in the contrast detail phantom was found to be improved with a green colormap display compared to conventional grayscale. Results suggest that application of a green colormap to clinical images may improve the detection of subtle lesions.

Adjusting the display color to match the peak sensitivity of the eye (555 nm, green) may result in greater conspicuity and improved perception of subtle lesions in mammograms.

Eads, E, Baker, J, Richard, S, Barnhart, H, Samei, E, Improved Perception of a Mammography Contrast Detail Phantom with Color Display.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11034259.html