[Value of a computer-assisted screening method (PAPNET) for the detection of infectious cervico-uterine smears]

There have been several reports regarding the accuracy of the PAPNET system applied to the screening for cancerous and precancerous lesions. Based on neuronal networks, this computerized tool was initially trained to select atypical cells. It has been approved in the USA for the re-screening of cervical smears for quality assurance. However, its particular behaviour has not been frequently studied when the system is designed to detect frequent infectious organisms. We report the results of re-screening of 42 inflammatory cervico-uterine smears by the PAPNET system. The computerized images were reviewed by two different pathologists, with complete agreement between the two observers in 39 cases. Infectious organisms were detected in only 66% of cases. Trichomonas, mycoses and Gardnerella were diagnosed in 63%, 56% and 87% of cases respectively. No herpetic lesions were identified. The low accuracy of the PAPNET system in the diagnosis of infectious cervico-uterine smears should be taken into account if this system is developed as an exclusive pre-screening method.