Karyometry by automated image analysis. Application to precancerous lesions.

Many works have shown a high frequency of increased nuclear DNA contents in various precancerous lesions. The true nuclear sizes, which are known to be well correlated with DNA values, may be stereologically estimated from nuclear profile data on tissue sections. Therefore, karyometry, which may be effected very efficiently by automated image analysis, offers an attractive choice when studying sections, in which photometric DNA evaluations are often unreliable. Prerequisites are Feulgen-stained slides and a stringent standardization of all histological procedures. Very few karyometric studies have been done on precancerous lesions. Automated image analysis has seldom been used for nuclear profile data gathering on sections. Suitable stereological models are necessary. The best-known are those for spherical nuclei, which allow an estimation of the size-distribution ('unfolding'). A specific model for parallel-oriented spheroids is particularly suitable for epithelial nuclei. It yields many stereological parameter estimations but does not yet allow a size-shape unfolding. We have used the model for parallel spheroids on nasal epithelial biopsies from nickel workers. The nuclear dimensions are usually clearly increased in metaplasia and dysplasia. Some rare dysplastic zones, however, present a major contribution of normal-sized nuclei, and this may have a prognostic significance. Our model may be applied to other epithelia. A new sphere unfolding model has been used on liver sections from Farber-protocol-treated rats. Compared to normal livers, where tetraploid nuclei predominate over diploid ones, transitional cell zones show an overwhelming predominance of diploid nuclei, and hyperplastic precancerous nodules have various degrees of non-modal and high-value ploidies. Karyometry may help to determine which lesions should be considered as possible premalignancies.