Quantitative description of chromatin structure during neoplasia by the method of image processing.

Nuclear chromatin is visualized by light microscopy as a mosaic of interchanging regions of low and high optical density (O.D.). The regions of high O.D. are well-defined as chromatin particles; features characterizing these particles enable the description of chromatin structure and the recognition of its changes during neoplasia. This paper presents a method of feature extraction by means of digital image analysis, based on a localization algorithm with locally adaptive thresholding. Although the chromatin particles varied greatly in their O.D., the algorithm enabled the identification of significant numbers of particles, which is essential in characterizing the complex architecture of chromatin. The results obtained by studying neoplastic nuclei and nuclei from control tissue suggest that the appearance of an additional class of chromatin particles, defined by their localization and optical density, is typical of chromatin changes during neoplasia.