A retrospective pilot study of the performance of mammographers in interpreting screening mammograms

Purpose: This paper provides pilot data from one Breast Screening Program on whether mammographers have the ability to review mammograms with similar accuracy to screen readers. Methods: The participant group consisted of 11 mammographers (experienced in assessing images for technical adequacy, but no specialised training for detecting abnormalities) and three current screen readers employed at Hunter BreastScreen. Fifty sets of mammograms performed during 2003 were used in the retrospective study. The mammograms were chosen to represent a range of review outcomes. Each participant reviewed each set of mammograms using the BreastScreen Australia assessment scale; these results were then categorised into two groups – Rescreen or Recall. The consensus review outcome of the original screen readers was used as the “Gold Standard”. Patient outcome was assessed by following up on the results of any histology or pathology tests in 2003 or the 2006 screening results. Results: Compared with the Gold Standard outcomes, the three current screen reader participants had sensitivity values of 67%, 83% and 94%; mammographer sensitivity values ranged from 61% to 89%. Specificity for the screen reader participants was 81%, 84% and 87% and mammographers ranged from 45% to 97%. Accuracy for the three screen readers was 76%, 84% and 90% while the mammographers attained between 55% and 86%. Conclusion: Without any training, the sensitivity obtained by three mammographers and the specificity obtained by six mammographers were similar to those of the current screen reader participants. Accuracy rates of the mammographers indicate that screen reading by selected and appropriately trained mammographers may be feasible