Differential diagnosis of borderline and invasive serous cystadenocarcinomas of the ovary by computerized interactive morphometric analysis of nuclear features.

Whether borderline serous tumors of the ovary can be differentiated from invasive serous cystadenocarcinomas by morphometric analysis of nuclear features of the neoplastic epithelium was examined. Multiple descriptors extracted from nuclear tracings and intranuclear gray-level analysis of argyrophilic nucleolar organizer regions were evaluated in 11 borderline and 18 invasive tumors using computerized interactive morphometric analysis. These descriptors included: nuclear area and perimeter; number, area, and perimeter of argyrophilic nucleolar organizer regions; standard deviations of these findings; and a size distribution of nuclear areas in discrete classes. Multivariate statistical analysis showed the internal consistency of the two groups in terms of physical descriptors and the discriminating value of the parameters of nuclear size and pleomorphism (independently of nucleolar organizer region parameters). The results indicated that interactive morphometric analysis of nuclear features, combined with appropriate statistical methods, could be used to distinguish between these tumors.