[Considerations on computer-assisted pattern recognition in the endometrium]

Histomorphological diagnosing very much depends on subjective empiric data. In histomorphology of the endometrium nomenclature as well as diagnostic classification lack uniformity and are full of controversy in literature. This results in opposing assessments of dignity of endometrial hyperplasias and prognosis of varying grades of carcinoma differentiation. One of the reasons is the lack of objective and reproducible diagnostic categories, not yet found. Morphometry with so-called computer-assisted interactive pattern recognition systems might offer the solution to numerous problems. Since there are no specific parameters for diagnostic identification of endometrial hyperplasias and carcinomas, only a combination of sensitive and reproducible parameters can allow for diagnostic identification. We report on our first computer-assisted pattern recognition studies and results that appear to provide an adequate basis for leaving descriptive diagnosing and turning to objective diagnosing. Such an objective diagnosing might not only be of clinical and therapeutic significance but also provide a new evaluation of the today's significance -- still full of controversy -- and prospective potency of endometrial hyperplasias. In computer-assisted interactive pattern recognition also lies the possibility of a contribution to objective tumour grading that appears to have high prognostic significance.