Paired comparison of manual and automated Pap test screening using the PAPNET system

A paired-comparison study of manual and automated (PAPNET) screenings of cervico-vaginal smears (Pap tests) was conducted to determine whether primary PAPNET screening was a reliable alternative to manual screening. A series of 5,037 consecutive Pap tests was first screened by the manual method. Next they were scanned by the PAPNET system, the DAT tapes were reviewed, and using a nonspecific triage protocol, abnormal tests were identified for limited, manual rescreening. False-negative rates (FNR) for each method were calculated and analyzed for due cause. By manual and PAPNET screening, respectively, there were 436 (8.6%) and 250 (4.9%) abnormal results. Manual screening missed 18 abnormals (5 SIL) and PAPNET 202 (7 SIL). The primary, manual screenings relating to the PAPNET false-negative tests were reviewed and revised to normal in 30. Based on the changes in the other 172 tests, cellular features ostensibly missed by the PAPNET system were listed to form part of a specific triage protocol, and were used to scrutinize the companion 172 DAT tapes: 150 tapes were abnormal. The manual FNR for an abnormal (SIL) result was 4% and (8.8%), respectively. Equivalent FNR pre- and postreviews for the PAPNET system were 44.6% (10.6%) and 5.2% (1.3%), respectively. This study discovered that the evaluation of some cell features in monitor-based, video images was the most important factor limiting the application of the PAPNET system as a primary screener. When governed by a specific triage protocol incorporating these features, primary PAPNET screening has the potential to equal the laboratory threshold of manual screening and be a better detector of SIL.