Interreader Agreement and Diagnostic Confidence in Discriminating Masses and Nonmass Lesions at Breast US

Background Accurate recognition of nonmass lesions (NMLs) at breast US is essential for ensuring consistent diagnostic performance. Purpose To assess agreement in classifying lesions as masses or NMLs at US and to identify features that support confidence in NML identification. Materials and Methods This retrospective single-center study included patients who underwent US-guided biopsy of lesions from January to June 2021. Four breast radiologists participated in the study to assess interreader agreement in classifying lesions as either masses or NMLs. An additional radiologist reclassified NMLs that had conflicting interpretations, categorizing them as either definite NMLs or controversial NMLs. Features associated with definite NML judgments were evaluated by means of logistic regression analysis. Results Among 1067 lesions in 912 women (mean age, 53.4 years ± 12.1 [SD]), 11% (121 lesions) were classified as definite NMLs and 7% (80 lesions) as controversial NMLs. The agreement was moderate to substantial (κ = 0.534-0.638), with a concordance rate of 84%-91%. The odds of being classified as a definite NML were low for lesions smaller than 1 cm or subtle architectural distortion (odds ratio [OR], 0.269 [95% CI: 0.112, 0.647]; P = .003) and high for cases with echogenic foci associated with calcifications at mammography (OR, 2.679 [95% CI: 1.130, 6.354]; P = .03) and microcysts (OR, 3.578 [95% CI: 1.236, 10.360]; P = .02). The presence of corresponding findings at mammography (P = .27) or MRI (P > .99) was comparable between the definite and controversial NML groups. Conclusion Interreader agreement for classifying lesions at US as masses or NMLs was moderate to substantial, with high concordances rates. Smaller lesions or subtle architectural distortions were less likely to be consistently classified as NMLs, whereas lesions with echogenic foci associated with calcifications at mammography and microcysts were more likely to be consistently classified as NMLs. © RSNA, 2025 See also the editorial by Hooley in this issue.