Automated analysis of rat breast tumors. Comparison of four methods

Chemically-induced malignant rat breast tumors pose diagnostic dilemmas since the majority are well-differentiated, noninvasive papillary lesions that are barely distinguishable from benign papillary lesions. This study compared several automated modalities to see which best separated benign from malignant breast tumors. Thirty-three carcinogen-induced rat breast tumors (13 adenomas, 10 papillary carcinomas and 10 invasive carcinomas) were evaluated by static (image) cytometry (ICM) of integrated optical density, by flow cytometry (FCM) and by two automated morphometric protocols, contextual analysis and single-gland analysis. DNA ploidy analysis, by either ICM or FCM, did not discriminate between the benign and malignant tumors. Contextual analysis correctly identified 11 of 13 benign and 17 of 20 malignant lesions (P less than .01). Single-gland analysis correctly identified all 13 benign and 17 of 20 malignant lesions (P less than .01). No method distinguished invasive from noninvasive carcinomas. The data suggest that architectural features are more important than nuclear features in differentiating benign from malignant rat breast tumors.