The Influence of Increased Ambient Lighting on Mass Detection in Mammograms

Abstract

Recent research has provided evidence that in reading rooms equipped with liquid crystal displays (LCDs), a measured increase of ambient lighting may improve clinicians' detection performance. In agreement with this research, the American College of Radiology (ACR) has recommended a moderate increase of ambient lighting in mammography reading rooms. This study was designed to examine the effect of a controlled increase of ambient lighting in mammography reading rooms on the diagnostic performance of breast imaging radiologists. Four breast imaging radiologists read 86 mammograms (43 containing subtle cancerous masses and 43 normal) under low (E = 1 lux) and elevated (E = 50 lux) ambient lighting levels on a Digital Imaging and Communications in Medicine-calibrated, medical-grade LCD. Radiologists were asked to identify cancerous masses and to rate their detection confidence. Observer areas under the curve (AUCs) were calculated using a receiver-operating characteristic analysis of fully paired results. Additionally, average observer selection times under both ambient lighting levels were determined. Average radiologist AUCs decreased with elevated ambient lighting (0.78 +/- 0.03 to 0.72 +/- 0.04). Observer performance differences, however, were of the same order of magnitude as interobserver variability and were not statistically significant. Average selection times under increased ambient lighting remained constant or decreased, with the greatest decrease occurring for false-positive (20.4 +/- 18.9 to 14.4 +/- 9.6 seconds) and true-positive (18.0 +/- 13.8 to 12.9 +/- 9.4 seconds) selections. The results agree with those of previous studies in that observer performance differences under a controlled increase of ambient lighting are not statistically significant. On the basis of these findings and ACR guidelines, a moderate increase of ambient lighting in mammography reading rooms is still suggested, but further research with additional cases and observers should be considered.