Background: Recently, the numbers of patients with breast cancer have increased rapidly in Japan, and about 40,000 people contract the disease annually today. Therefore, secondary prevention by breast cancer screening has become an important issue. Presently, mammography is used frequently for the diagnosis of breast cancer, but the reading of mammograms is difficult, and physicians must attend training classes specified by the Central Committee on Quality Control of Mammographic Screening and obtain a license to perform screening. Assessment categories are used for the evaluation of mammograms. Although they provide a scale for the diagnosis of breast cancer, the interpretation of mammograms is dependent on the physician's subjective judgment, and the advent of an objective evaluation method is awaited.
Methods: We scored the size, shape, density, margin, border, and internal structure of mammographic images and evaluated the relationships of these scores with lesion categorization.
Results: Since lesions could not be categorized by the analysis of any single item of mammographic images, the items were paired, and a new diagnostic system for breast cancer was prepared. When this system was applied, the diagnostic accuracy was very satisfactory, with a sensitivity of 100% (37/37) and a specificity of 92.9% (65/70) for category 5; 83.6% (51/61) and 97.8% (45/46), respectively, for category 4; and 88.9% (8/9) and 94.9% (93/98), respectively, for category 3.
Conclusion: In this study, findings concerning the margin and internal structure were important for the discrimination of category 5, and those concerning the size, shape, and density made major contributions to the discrimination of category 3.