Automated image morphometry of lobular breast carcinoma

Abstract

Objective: To analyze the role of automated image morphometry (AIM) in distinguishing infiltrating lobular carcinoma (ILC) of the breast from benign, borderline and infiltrating ductal carcinoma (IDC).

Study design: Only histopathologically proven lobular carcinoma, ductal carcinoma, borderline lesions and benign breast lesions were selected for the study. There were 19 cases of ILC and 30 cases of IDC, 20 cases of benign lesions (fibroadenoma, 18; fibrocystic disease, 1; and fibroadenosis, 1); 10 cases were borderline lesions (mild epithelial hyperplasia, 3; moderate epithelial hyperplasia, 2; florid epithelial hyperplasia 4; intraductal papillary carcinoma, 1). In all cases hematoxylin and eosin-stained slides were used for AIM. At least 100 cells from each case were subjected to analysis randomly with an image cytometer with Leica Quantimet 600 software (Cambridge, England). Nuclear area, diameter, perimeter, convex perimeter, convex area and roundness were measured in each case with random, unbiased selection of cells and 40 x objectives (one pixel = 0.46 microm). AIM data on the cases were analyzed in relation to final cytologic diagnosis.

Results: All the nuclear morphometric features of ILC were much lower than those of IDC and borderline lesions, whereas nuclear morphometric data on ILC were only marginally more than those on benign cases. ANOVA showed that mophometric data were significant (P < .05) in all the variables between ILC and IDC. However, there was no significant difference between ILC, and borderline and benign cases.

Conclusion: Image morphometry may be useful in distinguishing ILC from IDC on cytologic smears. However, morphometric data may not be helpful in distinguishing benign and borderline lesions from ILC.