ABSTRACT
BACKGROUND AND PURPOSE: Mammographic breast density is the strongest known marker of breast cancer risk. Visual breast density assessment is subject to significant intra- and inter-rater variability. The aim of the present study was to test the reproducibility of automatic breast density assessment and to compare the results to the visual assessment. PATIENTS AND METHODS: Serial mammograms of 141 patients were retrospectively reviewed. Breast density was assessed both visually using a BI-RADS four-category breast density scale and with a software tool for volumetric breast density measurement. RESULTS: The intra- and inter-rater reproducibility as well as inter-examination reproducibility were assessed for both techniques by calculating the intraclass correlation coefficient (ICC). The inter-examination reproducibility of the volumetric measurement of breast percent density was 0.91 (ICC; 95 % CI 0.87 - 0.93). There was no difference in the strength of the correlation between patients with a large vs. small difference in compression force. The intra- and inter-rater reproducibility ranged from 0.81 - 0.84 and 0.71 - 0.77, respectively. The inter-examination reproducibility of visual assessment was 0.75 - 0.81. The agreement of visual assessment with volumetric measurement was similar to the agreement among readers. CONCLUSION: Our results indicate that volumetric breast density measurement provides higher reproducibility in serial examinations than visual assessment and may thus be preferable in the longitudinal assessment of breast density and in the measurement of breast density for risk stratification.
