Sci—Fri AM(1): Imaging—04: A Fully Automatic Method for Estimating Breast Density in Digital Mammograms

Percentage mammographic breast density, a measure of the fraction of radio‐fibroglandular tissue in breast, is an indicator for high risk of developing breast cancer. Although the breast cancer development risk with increasing breast density is well established, this information is generally not used clinically due to the lack of quantitative, unsupervised methods to evaluate the breast density. Visual estimates produce only qualitative results and are irreproducible. Hence, semi‐automatic thresholding methods, in which the segmentation is performed by an expert but the percentage are is calculated by computer, are preferred. Here, we present the results of a fully automated method to analyze digital mammograms for breast density assessment. The comparison, using Pearson's product‐moment correlation, for the intra‐observer variability resulted in r 1 =0.89 and r 2 =0.87 for the first data set for two radiologists, respectively. Similarly, the inter‐observer comparison for the two radiologists yielded r 1=0.89 and r 2=0.85 for the first and second data sets. The automatic evaluation algorithm results in comparison to radiologists assessment gave r 1 =0.85 and r 2 =0.90 for the first data set and r 1 =0.86 and r 1 =0.85 for the second set. The significance level for all the coefficients was P < 0.0001. The correlations between the radiologists and automatic estimates imply that the explained variances in each case are comparable with the agreement between the two radiologists. A comparison of this method with the experts' evaluation, obtained using the current standard of semi‐automatic thresholding technique, indicates that the new unsupervised method for breast density estimation is quite accurate, reproducible and ready for clinical use.