Image Analysis Cytology for DNA Determination in Breast and Prostate Cancer
DOI:
https://doi.org/10.3109/02841869109088243Keywords:
Breast cancer, prostate cancer, diagnosis, DNA analysis, statistics, terminology, image analysis cytologyAbstract
Nuclear DNA distribution in fine-needle specimens from 112 breast carcinomas and 45 prostatic tumours was studied. The distributions were described statistically with five separate descriptors, namely mean deviation from 2C, percentage of cells exceeding 2.25C and 4.5C respectively, DNA index and entropy (a mathematical measure of degree of scatter of the DNA content of the nuclei). It was shown that information about DNA index only (‘diploid’ versus ‘aneuploid’) is too limited and the term ‘diploid’ even incorrect and misleading for description of human cancer cell populations. Merely an addition of the percentage of cells exceeding 2.25C allowed separation of a majority of the carcinoma specimens from specimens taken from normal tissue. Moreover, in carcinoma specimens 96% of the patients had between 1 and 100% of the cancer nuclei cell population exceeding 4.5C in contrast to all normal specimens, where this percentage was 0. Entropy was clearly correlated with percentage of Ki-67-positive cells, indicating that it contains information on both scatter of nuclear DNA content and proliferation. Plotting of the scatter (entropy) versus DNA index allowed separation of > 97% of the carcinoma specimens from specimens taken from normal individuals. With the use of all five descriptors of DNA distribution it was possible to separate all breast cancer specimens and all moderately and poorly differentiated prostate cancer specimens from normal specimens. The discrimination between well-differentiated prostate cancer and hyperplasia constituted a special problem in that it was not possible to confirm the diagnosis well-differentiated prostate cancer by objective measurements in 8% of the patients, even when all five descripitors were used for the analysis. None of the verified carcinoma specimens had a DNA distribution that was truly ‘diploid’ in the sense of the distributions of cancer nuclei and normal cell nuclei from the actual organ being quite similar.
