Automated counting of nuclei in series of conventional liver sections to differentiate hypertrophy and hyperplasia.

A method based on image analysis and applicable in pharmacology and toxicology is described that has been designed to meet the statistical requirements for the determination of liver hypertrophy or hyperplasia. An algorithm has been developed to detect and count fluorescent nuclei in Feulgen-stained liver sections (12,500 to 25,000 nuclei or 170 to 350 fields of observation per section within 2-4 hr) by means of fully automatic image analysis with the Leitz Texture Analysis System (TAS) (up to 23 sections in series on the microscope stage). The applicability of this method to conventionally derived, formaldehyde-fixed, tissue sections has been tested on 3-microns sections cut from archived paraplast blocks of already diagnosed materials. Alterations in the liver induced by eight different compounds administered perorally for 28 days to rats have been studied. The morphometric results were found to be specific for a given compound and revealed reproducible and statistically significant dose dependencies of hypertrophy or hyperplasia, or both. Increased frequencies of mitotic figures observed by eye correlated reasonably well with the morphometrically determined hyperplasia, but many hyperplastic livers revealed no mitotic figures. By contrast, the histological diagnoses of an increased degree of hypertrophy were only poorly correlated with the morphometric determinations and, in many instances, were made in livers showing no decrease in the frequency of nuclei. On the other hand, the results from electron microscopy agreed well with the corresponding morphometric analyses: in cases with predominant hypertrophy the proliferation of the smooth endoplasmic reticulum was distinct or striking, or the peroxisomes were altered and increased in number, whereas no obvious ultrastructural changes were seen in cases of morphometrically exclusive hyperplasia.