Image analysis--derived morphometric differences in fine needle aspirates of ductal and lobular breast carcinoma.

The differentiation of ductal and lobular carcinoma (including lobular carcinoma in situ) of the breast has important prognostic implications; thus, preoperative differentiation of these entities via fine needle aspiration could have a substantial impact on the therapeutic options offered to patients. However, well-defined, reliable light microscopic criteria do not currently exist for their distinction in cytologic preparations, particularly when the differential diagnosis of lobular versus low nuclear grade ductal carcinoma is considered. In this study we utilized a morphometric image analysis system to identify distinct nuclear features useful in the differentiation of these two entities and found that a number of significant differences do exist. Although some features (i.e., nuclear size) may be detected in traditional cytologic preparations, others may not be detected by the human observer, supporting a role for automated morphometric analysis as a useful adjunct in the distinction of lobular and low nuclear grade ductal carcinoma in fine needle aspirates of the breast.

Estrogen receptor immunostaining patterns imaged in breast cancer fine needle aspirates.

Fine needle aspirates (FNA) after estrogen receptor (peroxidase-antiperoxidase) immunostaining were imaged using a SAMBA system. Integrated optical density (IOD) and mean optical density (MOD) was measured in individual nuclei. Receptogram R Analytic software on a remote VAX computer was used to generate log-log scatter, contour and three-dimensional perspective plots of bivariate MOD vs. IOD relationships and for count-dependent Gaussian smoothing of the univariate log histograms. The findings revealed four types of staining patterns among otherwise estrogen receptor immunostaining-positive patients: (I) a discrete, homogeneous subpopulation with unimodal MOD and IOD distributions; (II) coexistent subpopulations of ER+ and ER- cells, revealed by bivariate MOD and IOD distributions; (III) multiple, discrete subpopulations of ER+ cells, revealed in perspective plots of MOD vs. IOD vs. scatter density; and (IV) highly skewed distributions forming a continuum over a broad MOD and IOD range with or without an ER-negative subpopulation. FNA ER-ICA-positive specimens were indistinguishable based upon average nuclear MOD (AV-MOD) or AV-MOD x (% ER-positive cells). Previous evaluation of such patterns in tissue sections revealed failure of tamoxifen response in types II and IV. Staining mosaicism (IV) may correspond to a failure of receptor modulation within defined limits when ER is rendered nonfunctional due to various structural modifications of receptor domains--events that would not affect immunostaining. Failure in type II is ascribed to ascendancy of estrogen-independent ER-negative subpopulations.