Ploidy determination on histologic sections of breast cancer specimens by image analysis using mathematical...

Several mathematical correction algorithms were developed to solve the problem of the unavoidable measurement of fragmented nuclei when determining DNA ploidy on thin histologic slides. These algorithms were designed for model tissue and until now had not been tested thoroughly on malignant human tissue. We evaluated the use of mathematical correction algorithms applied to measurements on thin histologic sections of breast cancer specimens, with strict control of the section thickness. Fifteen cases of breast carcinoma with known ploidy (5 diploid, 5 tetraploid, and 5 aneuploid breast cancer samples) were included. From each tissue block, we made a single-cell preparation and cut a thin histologic section. We evaluated the thickness of each of these sections according to a recently developed protocol and included only sections with a thickness between 5 and 6 microns. We performed DNA measurements with a custom-made image analyzing system equipped with a 100x oil immersion objective. Histograms of tissue section measurements were corrected according to the algorithms of McCready and Papadimitriou and of Haroske et al. We compared these results with the uncorrected histograms and with the histograms of the single-cell preparations. We also measured the single-cell preparations with a commercially available high-resolution image cytometer. The correlation between both image cytometers used was high (r = 0.99). Histogram correction improved results of tissue section measurements in all of the nondiploid tumors when compared with the uncorrected histograms. There were no significant differences between the correction algorithms used (correlation to the single-cell measurements determined by a linear regression; r = 0.98 for both algorithms). No overcorrection of the histograms occurred. We conclude that reliable DNA tissue section measurements are possible on breast cancer specimens and that such measurements will contribute to our understanding of tumor cell kinetics in small tumor cell populations not detected in single-cell measurements.