Search

Purpose To examine the effect that a test set from 1 region of the world has on the performance of expert breast radiologists from that region and from a distant region. Methods Forty-seven Australian and 15 Singaporean breast radiologists interpreted the same test set (ie, 60 mammograms of Australian women consisting of 40 typical and 20 atypical cases) in their respective locations. The radiologists assigned each case a confidence score of 1 to 5, where 1 was typical, 2 was benign, and 3 to 5 indicated identifiable cancer. Localization-response operating characteristic area under the curve (AUC), receiver operating characteristic AUC, location sensitivity, case sensitivity, and specificity were calculated for each radiologist and compared between Australian and Singaporean readers using nonparametric 2-tailed Mann-Whitney U tests. Results Australian readers performed better than did their Singaporean counterparts with respect to localization-response operating characteristic AUC (.738 vs .642, P = .0395) and location sensitivity (.775 vs .660, P = .0185). Discussion Technological advancement has made possible telereporting, in which radiologists report on mammograms originating from other countries. However, those readers' levels of performance might be affected by interpreting images from populations that are different from those with which they are familiar. Conclusion Radiologist performance differed between the groups when localization of lesions was taken into account. This difference might be because of the Australians' reading mammograms from populations with which they were familiar, in contrast to the Singaporeans. [ABSTRACT FROM AUTHOR]

Purpose: This Australian study explores the effects of an educational intervention on first-year medical radiation sciences (MRS) students by examining eye-tracking metrics while they attempted to detect breast lesions on radiologic images before and after an e-learning tutorial. The study also analyzes performance using receiver operating characteristic methodology., Methods: Fourteen first-year MRS students were equally and randomly assigned to a control or experiment group to participate in 2 image-detection sessions. The experiment group completed an online e-learning tutorial between sessions. Eighty mammographic breast images from 20 cases were obtained from a validated online image bank. Of those 20 cases, 30% were normal and 70% demonstrated a single-lesion abnormality., Results: The experiment group demonstrated a 45% increase in the mean number of fixations per case (P = .047), with a 30% increase in sensitivity (P = .022) following the tutorial. The experiment group also demonstrated improved lesion detection overall and a 49% decrease in mean time to first fixation on the lesion (P = .016)., Discussion: This study demonstrates that increased lesion detection is possible after a brief e-learning tutorial. Early changes in the viewing patterns of less-experienced technologists are intriguing, and explanation may be gained from our current understanding of radiologic perception and cognitive neuroscience. The task for the first-year MRS students in this study was only to positively detect a lesion, and we recognize that a higher level of visual processing would be required to critique image quality. Exposure to the brief tutorial, however, might have triggered some learning-related neural changes at an early level of visual processing, representing stimuli relating to task performance., Conclusion: Participants in the experiment group improved their ability to identify breast lesions, which coincided with changes in eye position metrics and error type analysis. The data presented here suggest that the intervention resulted in a significant improvement in detection by the experiment group.