[Automation of cytological analysis of cervical smears]

Quality Control (QC) is essential in cytologic laboratories. To reduce or avoid false negatives due to screening or interpretation errors, all cervical smears may be analyzed twice by two different cytopathologists (CP). During rescreening or Quality Control (QC), the second CP may be assisted by an automated device. Today there are different instruments: computerized microscope Ac Cell Series 2000 Pathfinder System), semiautomated or interactive systems (PAPNET, CytoRich or AutoCyte, ACCESS), automated systems (AutoPap 300). These devices, except computerized microscopes, utilize algorithmic image analysis that values single cells using morphological features but has difficulty in analyzing clusters of overlapping cells. For this reason it is better to use thin or monolayer preparations. This approach results in the loss of background and cells useful for the diagnosis and modifies sampling and processing methods. However, the techniques to obtain thin or monolayer preparations may process a higher number of cells than conventional method; moreover, the samples obtained may be valued also by the optic microscope, making cytologic analysis easier. The only device that combines the use of algorithmic image analysis with neural networks is the PAPNET system. Neural networks was inspired by neurobiology and may identify different cellular morphology and overlapping cells. In our laboratory, the PAPNET was proposed in 1995. In the present study, the last rescreening results of 1958 negative cervical smears are reported, analyzed during primary screening from July 1997 to February 1998. During the QC assisted by the PAPNET, 6 false negatives (0.31%), due to cytologic errors in the primary screening were detected. These results confirm the usefulness and the effectiveness of QC assisted by automated devices. However, only the CP evaluates abnormal cells detected by semiautomated systems or analyzes more atypical smears identified by the instruments. The work of CP is difficult: therefore a strict collaboration between clinician and CP to formulate the correct diagnosis is essential.