|| Checking for direct PDF access through Ovid
Mammographic parcnchymal patterns are among the strongest indicators of the risk of developing breast cancer. Risk evaluation through breast patterns may have an important role in studies of the aetiology of breast cancer and for monitoring changes in the breast in evaluating potential risk-modifying interventions. Typically, patterns are assessed by an experienced radiologist according to Wolfe grade, or on a coarse quantitative scale according to percent density. Parenchymal characterization methods, to overcome variability of classification by human observer, are under investigation. These include image segmentation using semi-automatic thresholding and automatic classification through textural and density measures. An important practical question relates to the extent to which information about mammographic pattern is carried by any one of the four views obtained in a typical examination. Specifically, variations of right-left breast symmetry and variations between the two standard views of each breast were tested. The mammograms of 30 premenopausal women, comprising 90 images [30 each of the right cranial-caudal (RCC), left cranialcaudal (LCC) and right medial-lateral oblique (RMLO)] were evaluated. Parameters included both subjective (radiologist classification and interactive image thresholding) and objective (fractal and skewness indices) quantitative measurements of parenchymal pattern. For the parameters tested, a high degree of correlation was observed for measurements on the RCC, LCC and RMLO views. Pearson correlation coefficients between 0.86-0.96 were found for the comparisons of quantitative parameters. The strong correlations suggest that, in the study and application of mammographic density classification, representative information is provided in a single view.