Elabbady, Tarek Z.; Babbs, Charles F.; Jackson, Valerie P.; Delp, Edward J.; Chelberg, David M.; Pizlo, Zygmunt; and Geddes, Leslie A., "Clinical ROC Studies of Digital Stereo Mammography" (1995). Weldon School of Biomedical Engineering Faculty Working Papers. Paper 8.
Date of this Version
breast, computer simulation, imaging, mammography, phantom
The objective of this study was to explore and document the diagnostic utility of digital stereo mammography for the detection of localized breast cancer in women. In it we characterized the ability of experienced mammographers, general radiologists, and non-radiologists to detect three types of tumor masses embedded within a heterogeneous background of normal tissue elements in numerically simulated digital mammograms. The simulated mammograms were displayed to the subjects on a high resolution video display, both in stereo mode and in mono mode. Half of the mammograms contained a single tumor, ranging from 0.3 to 0.8 cm in maximal diameter. Each reader rated 120 images (60 in stereo and 60 in mono) as to the probability of abnormality on scale of 1-5. Observer responses were evaluated using receiver operating characteristic (ROC) analysis to characterize any difference in diagnostic performance between the two viewing modes. The synthesized mammograms and the digital display were highly rated by the participant radiologists as promising tools for future research. The results of ROC analysis, however, indicated no significant difference in tumor detection when the same readers utilized the stereo mode versus the mono mode (Az mono = 0.833 versus, Az stereo = 0.826). The results were similar for readers of all 3 experience levels--mammographers, general radiologists, and non-radiologists.