The Computerized Method for Mammographic Density Classification According to the Fourth Edition of BI-RADS Lexicon

Abstract

     Mammographic density is an important predictor of risk of breast cancer. Typically, the radiologists classify mammographic density according to fourth edition of BI-RADS lexicon which combines the quantitative assessment and qualitative classification. The difference among experiences and expertise of each radiologist causes an error of mammographic density classification. Discrepancies as high as 46% have been reported in mammographic density assessment by two different radiologists. The purpose of this study is to develop the computerized method for classifying mammographic density according to the fourth edition of BI-RADS criteria. The study includes 132 digitized mammograms of women aged 37-81 years (mean 56) from DDSM database. The method consists of four major steps. First, the mammogram image noise is removed. Second, the breast tissue is excluded from background and pectoral muscle. In the third step, image thresholding method based on maximum entropy is used to separate fibroglandular tissue and fatty tissue. Finally, the percentage of fibroglandular tissue in the total of breast tissue area is calculated and classified with BI-RADS criteria. The results show that the overall accuracy of computerized method classification is 75% (99/132). The BI-RADS I reach 90.90% (30/33) correct classification, BI-RADS II 75.75% (25/33), BI-RADS III and IV 66.67% (22/33). The Kappa coefficient and Chi-square test are used to analyze the consistency of this method. The Kappa value (0.67) indicates the good relationship and Chi-square value (7.69,p=0.053) shows no statistically significant difference in the both methods. In conclusion, the computerized method based on maximum entropy would be useful as the radiologist assistant for classifying mammographic density. Nevertheless, this method is suitable for fatty breast but mammographic density classification errors may occur in dense breast.