Background: AutoPap (currently Focalpoint) is a computerized scanning system for the primary screening of cervicovaginal smears. For smears indicated to require further review, the system provides maps (PapMaps) to identify the most abnormal areas of the smear.
Methods: To study the effect of PapMaps on diagnosis, 481 smears (from 4656 successfully processed smears using AutoPap) were first classified by conventional manual interpretation and then reinterpreted blind by the same cytologist who limited his examination to only the fields of view marked by the PapMaps. The interpretation time with and without PapMaps was measured on another simple random sample of 188 smears.
Results: The interpretation time was reduced significantly, by approximately 40% using PapMaps (average reduction, 136 seconds per smear, 95% confidence interval [CI], 123-150). No cases manually classified as low-grade squamous intraepithelial lesions (LSIL) or worse and 20% of those manually classified as atypical squamous cells of undertermined significance (ASCUS) were judged as completely normal on the basis of fields of visions marked by PapMaps. In none of the latter was a histologically confirmed intraepithelial lesion identified. In relation to detailed diagnosis, the weighted kappa between manual and PapMap-aided interpretation was 0.745 (95% CI 0.687-0.804). Systematic differences were found (symmetry chi-square(10d.f) =19.11, P = 0.039): 48% of smears classified as ASCUS and 23% classified as LSIL by conventional diagnosis were classified as less severe (including normal) on the basis of PapMaps.
Conclusions: Use of PapMaps resulted in a substantial reduction in interpretation time. PapMaps showed good sensitivity (100% for squamous intraepithelial lesions and 80% for ACSUC) for selecting abnormal slides. The latter require full examination: detailed diagnosis based only on areas marked by PapMaps could lead to under-grading.
Copyright 2003 American Cancer Society.