.NTE1Mw.MjYyMzQz

- 3 - INTRODUCTION In the last two décades, a great amount of work has "been done in the field of automated cytology. The main goal was to replace the human observer by a machine, in order to avoid tedious and répétitive cell screening. The two main fields have been blood differential count and the early détec¬ tion of cancer of the cervix uteri. Both of these tasks could be performed by a human being, as is still the case in most laboratories. J.L. BINET suggested a différent material and a différent task. He proposed examining lymphocytes from patients with CLL and asked if it would be possible, by a quantitative morphological analysis of the nucleus structures, to find diffé¬ rences between the lymphocyte populations of normal and abnormal blood "Subsidiary" questions were corrélation with the anatomical stage of the disease and with the immunological nature of the proliferating clone (T- or B- lymphocytes) (6,7,8,9,20). The situation is quite différent from the previous one, where the cytologist or hematologist could answer ail questions himself, and could indicate the criteria in use. This is not at ail the case here. The human eye is to v/eak to notice slight quantitative changes of cells and the refore the cytology of CLL does not yet exist. Thus we can only observe the cells and try to express what we see in a quantitative way. In this paper we will concentrate on the chromatin patterns. Each quantitative description consists of two steps. The first step, the most important one, is to constitute the object to measure. The TV caméra furnishes an image of the measuring mask. This image is nothing other than an array of points with various grey tones. Only by a complex mental procèss can we speak of a lymphocyte in the center of the image, or of chromatin c lumps. A similar process has to be carried out by the image analysis device. For each feature defined, an image