Quantitative evaluation of mammographic densities: a comparison of methods of assessment

Differences in the proportion of the breast occupied by mammographic densities have been shown to be associated with differences in breast cancer risk. However, estimation of these densities by radiologists may be subject to error, and it is likely that quantitative measurement will reduce misclassification of densities and strengthen their association with risk of breast cancer. The objective of this study was to compare the extent of mammographic densities estimated subjectively by an experienced radiologist with the measured extent of densities using a digital planimeter. A total of 225 sets of mammograms from women aged 40-49 years and enrolled in the Canadian National Breast Screening Study (NBSS) were selected. The extent of the radiological densities was estimated visually by one radiologist. Independently, the total area of the breast and the areas of density were traced and measured using a digital planimeter. Visual estimations and measurements of mammographic densities were then compared to determine the extent of agreement between the two methods. In general, the two methods showed good agreement (kappa = 0.78). The measured area of mammographic densities tended to be slightly greater than the radiologist's estimations. Both methods were highly reproducible (radiologist-dependent method, kappa = 0.89; quantitative method, r = 0.95, P = 0.0001). Our results indicate that measurement of the area of mammographic density using a quantitative method is reliable, and correlates well with assessment by an experienced radiologist. The method may be useful for identifying women at increased risk of breast cancer.