Abstract
Aims: To study the visual and automatic measurement of mammographic breast density (MBD) and its implications
in tumor size assessment using distinct imaging techniques.
Methods: Retrospective, observational study of the visual and automatic measurement of mammographic breast
density according to the breast imaging data system (BI-RADS) in 212 patients with invasive unifocal breast cancer,
excluding microinvasive lesions, who did not receive neoadjuvant chemotherapy. Tumor size assessment is compared
using a linear regression according pathologic size with mammographic, US and MR size. The influence of MBD in
each technique of pathologic size was seen by Bland-Altman plot.
Results: Patient’s mean age was 55, 7±9.9 year-old. The mean size of the lesion stablished by mammography was
16.8± 10.4 (4 -70) mm, by US was 13.6±7.2 (5 – 55) mm and by MR 17.2 ±9.9 (5 – 66) mm. Mean pathologic size
was 12.6 ±8.1 (0.3 – 55) mm. Automatic MBD mean was 25.2±16.78. BIRAD assessment with visual and automatic
MBD measurements were correlated with a tendency of tumor size overestimation with visual method. Linear
regression of tumor size according image techniques with pathologic size showed an adjusted r-square of 27.3% for
mammography, 41.8% for US and 51.7% for MR. The best correlation was seen with MR although has a tendency
to overestimate tumor size. Only tumor size assessed by mammography was influenced by MBD. With this technique,
tumor size was best adjusted for those breasts with lower MBD.
Conclusion: Visual measurement overestimates MBD versus automatic measurement according BIRADS categories.
MR is the more accurate breast imaging technique for assessing tumor size independently of the BMD, which only
influences in the mammographic tumor size estimation.