Automatic detection of biological cells

The present research work has dealt with the analysis of biological cell images in general, and more specially with the cervical cells. This work was carried out in order to develop an automaton leading to a better prevention of cancer through automated mass screening. The device has been implemented on Motorola 68.000 microprocessor system. The automaton carries out cell nucleus analysis in several steps. The main steps are: - First: the automaton focuses on an individual cell nucleus among the smear's cell (about 10.000), - Second: it process each nucleus image. The digital processing yields geometrical of the nucleus (area and perimeter) for each cell. These data are stored in a local memory for further discriminant analysis by a microcomputer. In this way smears are classed in two groups: hale smears and uncertain smears. The automaton uses a wired logic for image acquisition and its software algorithms provide image reconstruction. The reconstruction algorithms are general purpose. Tests have proved that they can reconstruct any two dimensional images independently of its geometrical form. Moreover they can make the reconstruction of any image among the several images present in observation field. The processing times registered during the tests (for different cases) were situated, all of them, below three minutes for 10,000 images (each of them formed by an average of 450 pixels). The interest of the method is generality and speed. The only restriction is the primary device sensor (CCD linear array) length. Thus the automaton application can be extended beyond the biological image field. (author)