Quantitative morphometry by image analysis of normal, hyperplastic and cancerous ductal breasts

To analyze a number of quantitative variables in ductal epithelial breast proliferations and test them against diagnoses established by standard histologic and cytologic criteria in order to determine which variables are useful in discriminating between histologic diagnoses.A method was developed for the semiautomatic segmentation and quantitative analysis of breast ductal proliferations using image analysis. This permitted duct profiles and intraductal epithelial proliferations to be analyzed by computer, with calculation of a total of 17 variables--for example, nuclear/cytoplasmic ratio, nuclear circularity, average internuclear distance and lumen area. Four groups were analyzed employing a total of 80 cases with diagnoses established by standard histologic criteria, including normal breast duct (DN = 19 cases), ductal hyperplasia without atypia (DH = 21 cases), ductal carcinoma in situ (DCIS = 20 cases) and invasive ductal carcinoma (INV = 20 cases).Of the 17 quantitative variables tested in this study, 15 showed statistically significant differences in values between DH and DCIS (p < 0.05), and 7 showed statistically significant differences between DN and DH (p < 0.05).By quantitative morphometric analysis of selected variables, a significant statistical distinction can be made between diagnostic groups, paving the way for automatic computerized image recognition for differential diagnosis.SCI(E)PubMed0ARTICLE5255-613