Impact of prior mammograms on radiologists and radiographers' detection of different breast cancer lesion types

Introduction: Mammographic interpretation relies on the detection of suspicious abnormal changes, and reference to prior mammograms may support the detection of these changes. However, the influence of prior mammograms on the detection of different breast lesions is unclear. This paper assesses the influence of prior mammograms on the detection of different lesion types in mammograms. Methods: Mammographic test sets comprising different lesion types were interpreted in two stages. In Stage 1, eight radiologists interpreted current mammograms of 72 women (normal: n = 40; cancer: n = 32) with and without access to the prior mammograms. In Stage 2, 13 radiographers interpreted another test set containing 28 mammograms (normal: n = 19; cancer: n = 9) with and without access to the prior mammograms. Radiologists and radiographers' performances in detecting different lesion types with and without prior mammograms were compared separately using a paired t-test. Results: In Stage 1, reference to prior mammograms did not improve sensitivity for architectural distortions (p = 0.48), calcifications (p = 0.85), discrete masses (p = 0.45), and non-specific density (p = 0.22). However, prior mammograms improved the detection of spiculated/stellate lesions (p = 0.05) and diagnostic accuracy for architectural distortions (p = 0.006) and discrete/spiculated/stellate masses (p = 0.01). Prior mammograms had no impact on lesion sensitivity (p > 0.05). In Stage 2, no differences were observed in sensitivity and lesion sensitivity when compared to without prior mammograms for all lesion types (p > 0.05), but prior mammograms improved overall diagnostic accuracy (p ≤ 0.002). Conclusions: Prior mammograms improve the detection of spiculated/stellate lesions but have no impact on the detection of other lesion types when measuring radiologists' and radiographers' performance.