ABSTRACT
Papanicolaou smears from 378 patients were analyzed in a fully automated microscope system using an image analysis technique. The logic of the system subdivided the smears into "normal" (63%), "not possible to analyze in the machine" (17%) and "positive" (20%). The results have been compared with the conventional screening method available today taking into consideration the definition problems of a true diagnosis, cytopathological and histopathological consensus, reproducibility, sample-taking and interpretation mistakes, etc. As 37% (rejected and positive) of the smears in this particular study were analyzed visually after machine prescreening, it could be demonstrated that the false-negative rate among the remaining 63% of the specimens was significantly lower among the machine-screened smears when compared with visual screening if the results were correlated with histopathological consensus. The results clearly indicate that the technology required to build computerized microscope systems, which are able to automatically sort out at least two thirds of Papanicolaou smears is available today. The only limiting factor is the cost-benefit relationship.
