Discrimination between precancerous and cancerous lesions of the uterine cervix by DNA measurements on tissue sections.

Morphologically typical uterine cervical biopsies were separated into normal cervices, condylomas and cervical intraepithelial neoplasias (CIN) grades I, II and III. At least 100 nuclei per lesion were measured on 4 micron Feulgen-stained sections using a Zeiss microspectrophotometer, with a variant of the plug method used to compute the nuclear DNA content. DNA distribution histograms were then decomposed into subsets of diploid, tetraploid, octoploid and aneuploid cells. The decomposition, which assumed a log-normal model of polydiploidy distribution, led to the identification of six indices: (1) the percentage of diploid cells, (2) the percentage of tetraploid cells, (3) the percentage of octoploid cells, (4) the percentage of aneuploid cells with DNA contents less than tetraploidy, (5) the percentage of aneuploid cells with DNA contents between tetraploidy and octaploidy and (6) the percentage of aneuploid cells with DNA contents greater than octoploidy. These indices, along with the mean nuclear radius, the 5c exceeding rate and the 2c deviation index, generated a nine-dimensional space. Two methods of discriminant analysis on this space showed discriminating powers of 78.22% and 87.13%, respectively, as compared to the original diagnoses. The most discriminating variable in both analyses appeared to be the percentage of octoploid cells.