An image analysing method for automated cytologic prescreening of cervix carcinoma and its prestages: demonstration and preliminary results.

The image analysing method for automated cytologic prescreening is founded on the evaluation of epithelial cell nuclei in regularly stained and prepared smears. The technical equipment consists of a fast scanning microscope and a software-oriented computer. The device is able to isolate all objects present in the visual fields. Clumps, artefacts, detritus and other disturbing objects are eliminated. Among the cell nuclei the nonepithelial forms (histiocytes, leukocytes, lymphocytes) are distinguished from the epithelial ones. The latter are compared with a representative sample of normal epithelial cell nuclei, exactly classified by cytologists. All objects are described by 18 mathematically defined parameters. The results of classification is a statistical information on the probability of coincidence of an unknown case with the normal object sample. In addition the evaluation program provides the number of automatic isolated cell nuclei. By these informations it is possible to differentiate between positive (tumor-suspicious), negative and inadequate smears.