FUNDAMENTAL STUDY OF AUTOMATIC CYTOSCREENING FOR UTERINE CANCER .5. DATA-ANALYSIS FOR IMPROVEMENT OF CYBEST

  • 1 January 1977
    • journal article
    • research article
    • Vol. 21  (4) , 536-538
Abstract
The particles selected by CYBEST [Cytobiological Electronic Screening System] as abnormal cells at the stage of coarse scanning were examined by direct microscopy to determine whether they were actual cells or not. Approximately 20-30% of these were cell clusters and/or clumped leukocytes and red cells or specks of dust. Such incidents interfere considerably with a precise cytologic assessment. To eliminate this problem by a software algorithm, experiments were done setting thresholds for human cell feature values. When the values obtained by machine measurement exceeded the threshold values the particles concerned were automatically disregarded, as interfering incidents, at the stage of coarse scanning. The approach had substantial value in the operation of CYBEST as a prescreening device.