.NTE1Mg.MjYyMzA4

- 4 - INTR0VUCT10N This paper deals with a set of methods spécially developed for an automated cervical screening but of much wider applicability in quantitative cytology. These methods may be used in ail computers able to handle images, more precisely to threshold,digital ize, store, complé¬ ment ,intersect and translate them. One such computer, the TAS, first designed by our group (6) and now commercialized by Leitz easily performs the basic image transformations of mathematical morphology((7), (8), (l3)),like the érosion, the opening and the Golay transformations((4), (1])). Ail our methods have the same principle : instead of measuring parameters on the original set where ail features are mixed, we trans- form the image in a new image where some features are emphasized and the others are eliminated. Each transformation has an image as input and an image as output. Hence several image transformations can be associated to form complex processes called itérative algorithms ((?.), (13)). The power of them will appear clearly by the wide range of applications in automated cytology we will présent : - prescreening at a low resolution (two possibilities) - élimination of false positive signais (an other approach is described in (9)) - automatic séparation of clustered cells - individual analysis of - shape - relation between cytoplasm and nucleus - chromation structure.