Background: After menopause, declining levels of estrogens may cause vaginal discomfort, or so-called "vaginal atrophy." Evaluation of therapies for vaginal atrophy may be performed using the so-called "maturation index." The maturation index is expressed as the percentage of (para)-basal, intermediate, and superficial epithelial cells in a vaginal smear. Manual assessment of the maturation index is subject to inter- and intraobserver variations. In this study, assessment of the maturation of cells in vaginal smears using automated image analysis was investigated.
Materials and methods: Automated assessment, using a commercially available image analysis system, was performed on hematoxylin-eosin-stained cytospin specimens. A training set was constructed by an experienced cytotechnologist, based upon visual classification of stored grey value images. From this, two discriminant functions (DFs) were calculated capable of classifying cells in one of the three types. These cell classifiers were capable of classifying 97% of the cells correctly. Data from automated assessment were compared with those of classical manual counting. Specimens of 13 mature and 6 atrophic vaginal specimens were assessed in duplicate, both manually and by image analysis, using the DFs.
Results: No significant interobserver effect was found for image analysis, whereas a significant effect was found for manual counting. Both methods were able to distinguish between matured and atrophic specimens.
Conclusions: It was concluded that for assessment of vaginal maturation, the use of automated image analysis systems is recommended. Besides increased reproducibility, image analysis systems yield additional data describing the size and shape of the cytoplasm and nucleus of cells, which might increase discriminating power.