Nuclear DNA analysis of oral hyperplasia and dysplasia using image cytometry.

We investigated the value of image analysis in discriminating among oral white lesions with hyperplasia without dysplasia and oral white or white-and-red lesions with moderate or severe dysplasia. Normal oral epithelial tissue was used as a control. Image analysis was applied to 5-micron formalin-fixed sections stained with the azure A-Feulgen reaction for nuclear DNA. For 150-200 cells from each section, 5 nuclear variables were assessed: area, form factor, total stain, average stain and ellipticity. For each variable, 2 measurements were obtained, the mean and the interquartile range, and were used for stepwise discriminant analysis. Using this test, a model of 3 measurements with the most discriminating power was developed. When the jackknife classification test was applied to this model, we could discriminate with 81% accuracy between the 4 groups of tissue studied.